{"v":1,"site":"newsroom.sgit.ai","release":"v0.12.11","docs":[{"u":"index.html","t":"SGit Newsroom","k":"page","x":"One page, one argument, with the figures, the data and the links to check it.","h":[["all","Every article, newest first"],["threads","How they connect"]],"i":0},{"u":"about/participant.html","t":"Where we lose","k":"page","x":"This site is published by the sgit project, which is building the stack it argues for. That is a participant's position, not a neutral one. Two years of disclosed model co-authorship, and where this site's own argument is weakest.","h":[["model-coauthorship","Two years of disclosed model co-authorship"],["lose","Where this is not, and cannot yet be, what it argues for"],["lose","1 · Most of what it argues for does not run"],["lose","2 · The metering mechanism is unspecified"],["lose","3 · The self-critique is real, not decorative"],["lose","4 · It inherits a real technical weakness"],["licence","The licence position, stated rather than assumed"],["what-was-redacted","What was redacted from this site, and why"]],"i":1},{"u":"account/index.html","t":"Your reading account","k":"account","x":"A reading meter kept in your browser: every page costs a few pence, charged by how much of it you read, from £5.00 of starting credit. Credit can go negative and nothing is blocked; your history, graph and personas live here too.","h":[["prices","What things cost, read to the end"],["why","Why it works like this"]],"i":2},{"u":"account/newsroom.html","t":"Your newsroom","k":"account","x":"Your own front page, one persona at a time: picks, what you kept, and the graph your reading has grown. Worked out in your browser from a history nobody else holds.","h":[],"i":3},{"u":"account/personas.html","t":"Your personas","k":"account","x":"Add, name and switch between reading personas, each with its own picks and graph; start from five made from what sgit.ai publishes. Kept in your browser.","h":[],"i":4},{"u":"account/share.html","t":"Send us your reading, get a front page designed for you","k":"account","x":"Read as you normally would, then send us what you read, encrypted in your browser or copied into an email, and get back what your own front page could look like.","h":[["how","Where it goes"]],"i":5},{"u":"account/top-up.html","t":"Top up","k":"account","x":"Add £5 of reading credit to the balance kept in your browser: one payment on Stripe, no account, nothing renews.","h":[],"i":6},{"u":"account/topped-up.html","t":"Topped up","k":"account","x":"The page a top-up returns to: it adds the credit to the balance kept in your browser.","h":[],"i":7},{"u":"admin/comms.html","t":"Comms: tasks & requests","k":"admin","x":"Open requests and standing tasks for this site, published rather than tracked privately — including origin material still to retrieve, PDFs to back-fill into git, and a heading bug to fix at source.","h":[["needs","Open requests"],["tasks","Standing tasks"]],"i":8},{"u":"admin/index.html","t":"How this site is built","k":"admin","x":"Hand-written static HTML, one generated chrome definition, and a pre-release gate that fails the build on a broken link, a missing provenance date, or redacted material.","h":[["pipeline","The pipeline"],["pipeline","validate"],["pipeline","tag-release"],["pipeline","deploy"],["gate","What the gate checks"],["chrome","The chrome"],["release","Making a release"],["tree","The repository"],["scope-note","v0.2.0's scope decision, stated rather than hidden"]],"i":9},{"u":"admin/versions.html","t":"Release history","k":"admin","x":"Every push to dev is a release: CI validates, verifies the version bump, tags the commit and deploys to GitHub Pages. This is what shipped in each one.","h":[],"i":10},{"u":"archive/2026-08-22-home.html","t":"newsroom.sgit.ai — The Future of News","k":"archive","x":"News is failing not because there is too little information but because there is no walkable chain from a claim to its evidence, no way for a correction to reach what it disproved, and no way to pay the person who did the original work. A story is a graph; the article is a projection.","h":[],"i":11},{"u":"archive/2026-08-23-newsroom-roles-and-operations.html","t":"The newsroom: roles & operations","k":"archive","x":"Not an effort to replace journalists with agents. The opposite. Agents make it possible to involve more humans, in more roles, at more scale than was previously affordable. Humans are the bar; agents are the volume.","h":[["principles","Eight design principles"],["roles","Agent roles, and the daily clock"],["departments","Departments as public pages"],["craft","The craft: what agents do not replace"]],"i":12},{"u":"archive/2026-08-23-what-is-shipped.html","t":"What is shipped, what is argued","k":"archive","x":"Nothing on this site is running yet. The articles are real and dated; the newsroom is a design. Here is the line between them. The entire evidence-economy cluster is marked PROPOSED — does not exist yet.","h":[["runs","What runs"],["runs","The vault substrate"],["runs","Ten published articles"],["runs","MyFeeds as a live pipeline"],["runs","The governance graph and its gate"],["runs","The newsroom, and the door that is shut"],["runs","The x402 payment rail"],["designed","Designed only"],["never-publish","Never publish as a claim"],["why-this-page","Why this page is non-negotiable"]],"i":13},{"u":"archive/index.html","t":"The archive","k":"archive","x":"What this site said before the newsroom ran: the first home page, the designs the running newsroom has overtaken, the instances as they ran, and the material it was built from. Dated, labelled, and kept at their original addresses.","h":[],"i":14},{"u":"articles/a-link-to-a-persona.html","t":"A link to a persona: send people to the newsroom as someone, not as nobody","k":"article","x":"Send a CISO to the newsroom as a CISO: personas kept in a vault by the newsroom's agents, opened from a link, followed or forked.","h":[],"n":66,"d":"2026-10-10","tp":["startups-and-strategy","vaults-and-method"],"tg":["newsroom","personas","personalisation","vaults","local-first","privacy","onboarding","agents","brief","article"],"g":"A link should land a reader as someone rather than as nobody: personas move from the site's code into an encrypted vault kept by the newsroom's agents and opened in the browser with a published read key, reached through a page of their own, a URL fragment that never reaches a server, or a private vault for one person, and the reader can follow the persona as it is updated or fork it as their own, all client side. · Landing as nobody · Personas in a vault · Three kinds of link · The fragment never reaches a server · Follow or fork · The persona keeper · Bluesky starter packs · Netflix profiles · Personas about people stay private · All client side","i":15},{"u":"articles/a-mac-of-the-agents-own.html","t":"A Mac of the agent's own: a business plan for agent desktops, what Apple's licence allows, and three behaviour policies","k":"article","x":"A business plan for a Mac of the agent's own: what Apple's licence allows, a desktop built from vaults per run, and three behaviour policies for one agent.","h":[],"n":52,"d":"2026-10-09","tp":["startups-and-strategy","agents-and-policy","vaults-and-method"],"tg":["agents","agent-desktops","macos","business-plan","agent-behaviour-policy","licensing","vaults","pki","sgit","wardley-maps","isolation","cowork","computer-use","article"],"g":"After a mailbox, a code-host account and a Claude account, an agent's next dedicated resource is a desktop, and it should be a Mac; Apple's licence rules out a pool rented by the minute, so the business is a dedicated Mac per customer, run for them, metered by the minute, built clean per run from encrypted vaults with keys by PKI and erased at the end, plus developer agents, software for your own Mac and a request to Apple for terms; the behaviour policies show why: the same agent's unbounded excess falls from 23 rows to 11 to 3. · The next dedicated resource · Why a Mac · What Apple's licence allows · The shapes that are allowed · A clean Mac, a key, a vault · Keys by PKI · Three desktops, one agent · The shop admin screen · A meter on a Mac one customer holds · Productising the agent's desktop · What are we looking for?","i":16},{"u":"articles/a-meter-in-the-browser.html","t":"A meter in the browser: a penny a page, a history you keep, and a site that picks for you","k":"article","x":"Every page now costs a few pence from £5 of credit kept in your browser, and the reading history it keeps is what personalises the site.","h":[],"n":57,"d":"2026-10-10","tp":["startups-and-strategy","news-and-evidence"],"tg":["newsroom","micropayments","personalisation","local-first","privacy","pricing","early-access","article"],"g":"A reading meter that lives only in the browser, a few pence a page from five pounds of starting credit with a simulated top-up and no blocking, tests whether people would pay a little for reading when what they get back is a site that personalises itself from a history nobody else holds; escaping the meter is easy, but escaping it also discards the history, which is the point of the experiment. · A meter in the browser · Never block a page · Every step but the payment · The history is the product · The meter can be escaped; the personalisation cannot be copied · Would people pay for a site that knows them? · The pt.newsroom wallet · 402 Payment Required","i":17},{"u":"articles/a-model-in-the-tab.html","t":"A model in the tab: Chrome now ships an LLM to capable desktops, and it is the hardest piece of the client-side puzzle","k":"article","x":"Chrome now ships a language model any page can prompt, offline, on capable desktops. The last piece of the client-side puzzle, with rough edges.","h":[["short","In short"],["ships","What Chrome actually ships"],["download","Why the download felt so bad"],["examples","Why there are so few examples"],["against","The case against, taken seriously"],["fit","Why this fits what we have built"],["opportunities","Where we can use it"],["next","What we are building next"],["sources","Where this comes from"]],"n":73,"d":"2026-10-11","tp":["site-and-engineering","graphs-and-knowledge"],"tg":["client-side","gemini-nano","prompt-api","chrome","built-in-ai","on-device","privacy","sovereignty","interoperability","webmcp","newsroom","personas","article"],"g":"Since Chrome 148 any page can prompt Gemini Nano on the reader's device, which supplies the hardest missing piece of the client-side pattern; the rough first run is mostly fixable by the page, the scarcity of examples has real causes, the standards objections are serious, and the way to build on it without being captured is to keep the meaning in anchored graphs and use the model as a checkable function. · The Prompt API, stable in Chrome 148 · The last piece of the client-side puzzle · The download never stopped · Why there are so few examples · The case against · Keep the meaning in the data · Graph answers fit a small window · A download panel that tells the truth · Ask this article, personas, voice memos · What is fast enough, on which machines?","i":18},{"u":"articles/a-personal-agent-that-keeps-your-secrets.html","t":"A personal agent that keeps your secrets: the 2026 agents read through behaviour policy and encryption, and a privacy-first design on vaults, enclaves and the browser","k":"article","x":"The 2026 personal agents read through behaviour policy and encryption, and a design on vaults, an attested enclave and the browser where no vendor holds a key.","h":[],"n":31,"d":"2026-10-06","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","personal-agents","agent-behaviour-policy","encryption","confidential-computing","enclaves","webgpu","privacy","data-sovereignty","vaults","sgit","riskmandate","meta-muse","openclaw","article"],"g":"The personal agent became a product category in 2026, an always-on agent with a cloud computer, memory as files and connectors into a person's life; read through RiskMandate's behaviour policy the products differ most on the barrier, with Meta's Sentinel a boundary on actions and most others settings, while the mandate is unwritten and liability sits with the user; read for data sovereignty most sit in the vendor's corner, with Meta's own engineers saying today's protection is policy not cryptography and Apple's Private Cloud Compute the published design that does most; a page per person is a memory about non-users with no route to it; the vendors' memory and this site's CRM converged on files, no database and a page per person, so the privacy question is who holds the keys, not architecture; a privacy-first design puts memory in vaults the host holds as ciphertext, a behaviour policy as the permission authority, compute in the browser or an attested enclave with a per-task key released by the user's device and destroyed after, the frontier model only for the step that needs it, and a folder per person the person can read; what exists and what does not is stated, and the question of who gives the mandate over information is left for its own document. · The year the agent got a computer · The shape came from a self-hosted project · Lens one: what stops the agent · A separate permission authority is a boundary · The blast radius was demonstrated · Lens two: who can read what it","i":19},{"u":"articles/a-reading-list-in-a-link.html","t":"From me to you, in a link: a personal reading list, two briefs to build and use it, and the loop that makes it a day's work","k":"article","x":"A reading list from one person to another, in a link the server never sees; a brief to build it, a brief to use it, and the loop that ships it in a day.","h":[["short","In short"],["brief-newsroom","Part one: the brief for the newsroom"],["what","What to build"],["numbers","1. A number for every article"],["page","2. The page that opens a list"],["maker","3. The page that makes a list"],["tests","4. How to know it works"],["not","5. What not to build"],["brief-crm","Part two: the brief for the RiskMandate agent"],["crm-what","What you are doing"],["crm-steps","The steps"],["crm-rules","Rules"],["loop","Part three: the loop"],["costs","What it costs"],["why","Why it matters"],["sources","Where this comes from"]],"n":74,"d":"2026-10-11","tp":["site-and-engineering","startups-and-strategy"],"tg":["newsroom","personas","curation","reading-list","briefs","crm","outreach","workflow","feedback-loop","agents","product","article"],"g":"A personal reading list in a URL fragment (from, to, article numbers to read, to look at, and to follow) that opens on the newsroom as a persona the reader can keep or remove, with a stable public number for every article and a page to make a list for anyone. Two briefs: the newsroom builds it with validation, tests and no tracking; the RiskMandate agent, which keeps the contacts, chooses the person with Dinis, researches public sources, picks articles with reasons, tests the link, drafts the email, and Dinis sends it. The core is the loop: editorial, product, engineering, sales and the founder connected by written briefs, small contracts, owned repositories and tests, closing in about a day, as it did this week, with costs that are written down: drift, budgets and one person as the router. · A reading list in a link · A number for every article · The list becomes a persona · Anyone can make a list · The RiskMandate agent uses it · Nothing tracks the reader · The loop · This week's loops · What the loop costs · Vibe coding with a slow loop","i":20},{"u":"articles/a-second-reader-the-agent-cannot-skip.html","t":"A second reader the agent cannot skip: how two wrong emails became a gate on every draft","k":"article","x":"Two emails went out in my voice despite a written rule. The fix: a hook that makes a second model approve every draft, failing closed.","h":[["view-catalogue","In short"],["view-catalogue","Why the rule failed: hope held by the agent it governs"],["view-catalogue","How the guard works"],["view-catalogue","The hook point"],["view-catalogue","Stage 1: code checks, no model"],["view-catalogue","Stage 2: the independent checker"],["view-catalogue","Failing closed, and the log"],["view-catalogue","What the first run caught"],["view-catalogue","Six kinds of independence"],["view-catalogue","From setting to boundary: where the hook lives decides what it is"],["view-catalogue","Building it with what is public today"],["view-catalogue","The pattern is old, and it has a name"],["view-catalogue","The pattern, reusable"],["view-catalogue","What it does not do"]],"n":55,"d":"2026-10-09","tp":["agents-and-policy","site-and-engineering"],"tg":["agents","claude-code","hooks","guardrails","llm-as-a-judge","maker-checker","agent-behaviour-policy","email","governance","article"],"g":"A rule held by the agent it governs is an expectation that can be forgotten, missed or argued around, so the inbox agent's fix after two emails went out in the founder's voice puts a gate on the draft tool itself: deterministic code checks first, then a fresh model call with only the rules, the sources and the draft, a different model from the author, failing closed on anything but an explicit PASS and logging every verdict; its independence depends on where it is installed, a setting while the agent can change it and a boundary only under managed placement, and the pattern is the old maker-checker control applied to any consequential tool call. · Two emails in the founder's voice · A rule held by the agent it governs · The draft guard · Code checks first · A fresh checker · Fail closed · Six kinds of independence · Placement decides the barrier · Maker-checker · The reusable pattern · What the first run caught","i":21},{"u":"articles/agency-is-not-a-yes.html","t":"Agency is not a yes: a scale for human and agent decisions, from rubber stamp to the reviewer who fixes the source","k":"article","x":"A yes or no is not agency: seven dimensions, a seven-level scale for people and agents, and the review as a QA step that fixes the source.","h":[],"n":46,"d":"2026-10-08","tp":["agents-and-policy","graphs-and-knowledge"],"tg":["agency","accountability","human-in-the-loop","decision-making","agents","agent-behaviour-policy","provenance","fractal-semantic-graphs","eu-ai-act","maturity-model","vaults","article"],"g":"Agency for anyone asked to decide, a person or an agent, needs options beyond yes, context in their own terms, the ability to look behind what they are shown, time or tokens, symmetric incentives, escalation and authority over the source; the weakest of these caps the decision on a seven-level scale; below level 3 the decider carries liability rather than agency and accountability belongs to whoever designed the decision point and up their chain; above it the review is a QA step that validates everything upstream and fixes the source, and at the top proven cases become rules that run as code under a behaviour policy, which is also what the EU AI Act's oversight requirements already ask for. · A yes or no is not agency · Add label 756-459-3214? · Whatever you punish, you teach the other answer · Everything shown is a compression · Accountability without agency is liability · Seven levels, seven dimensions · Fourteen decision points scored · Agents are deciders too · The review is a QA step · Ownership changes attention · From review to rule · The law puts the floor at level 3 · The agency scale, the vault","i":22},{"u":"articles/agentic-cost-is-an-engineering-problem.html","t":"Agentic cost is an engineering problem: a business plan for a service that reads the logs a company already has and finds the money, then the engineering","k":"article","x":"A business plan: read the logs a company already has, draw each agent's footprint, twin and policy, and find the money first and the engineering second.","h":[["short","In short"],["problem","The problem, on a map"],["service","What the service does"],["positive","Why the positive frame matters"],["vaults","Why vaults"],["money","How it makes money"],["exists","What is already here, and what is not"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":88,"d":"2026-10-12","tp":["startups-and-strategy","agents-and-policy","vaults-and-method"],"tg":["business-plan","agentic-ai","cost","wardley-maps","explorer-villager","agent-behaviour-policy","digital-twin","logs","vaults","riskmandate","startups","article"],"g":"A service that manages the cost of agentic AI by observing what already exists: an estate whose bill nobody can explain shipped from the explorer side without the villager's discipline, as the Wardley map shows, with logs, traces and invoices never read together and no owner of cost; the service installs nothing, reads the logs into a vault, draws each agent's footprint, builds a twin of the estate, writes the behaviour policy each agent was in fact running under with the enforcement column filled honestly, and maps cost and capability into one graph; the findings are the cost list, which pays, and the engineering list, which is the reason; the gaps are findings too; vaults carry capture, transformation and sealed anonymised extracts; four price levels on the RiskMandate pattern, every number a labelled assumption. · Cost is an engineering problem · The problem on a Wardley map · Observe: install nothing · The footprint from the logs · The twin of the estate · The policy each agent was really running under · Two lists, in this order · The gaps are findings · Capture, transform, move · Four levels and a share of savings · What is already here","i":23},{"u":"articles/ai-baseline-control-framework.html","t":"An open AI governance framework, and what its licence let us build","k":"article","x":"Twenty open AI governance controls under CC BY-SA, why the licence matters, and the same day's conversion into a graph, a database and a walk down to EU law.","h":[],"n":36,"d":"2026-10-07","tp":["graphs-and-knowledge","agents-and-policy"],"tg":["ai-governance","ai-bcf","eu-ai-act","nist-ai-rmf","iso-42001","open-licence","creative-commons","semantic-graphs","fractal-semantic-graphs","sqlite","vaults","article"],"g":"The AI Baseline Control Framework is a practical, deployer-focused set of twenty AI governance controls in five categories and three types, each with a why and a how and a mapping to the NIST AI RMF, ISO/IEC 42001 and the EU AI Act; its Access category covers what an AI system can reach, which the three it maps to do not; its CC BY-SA licence and CSV export let it be converted on the day into a semantic graph with an ontology, a taxonomy, linked data, documents and a browser database, joined by id to a graph of the AI Act, and walked as a fractal graph from the framework down to a paragraph of law, which surfaced findings the list does not state and kept every addition labelled apart from the source. · AI BCF: twenty controls for deployers · Baseline, Tier II, Trigger · The Access controls have no equivalent in the three · Grant, mandate and owner · CC BY-SA 4.0 says yes in advance · ShareAlike keeps the tools open · A CSV export turned a reading into a dataset · One CSV in, five shapes out · Every edge says where it came from · Five altitudes, one grammar · Two vaults joined by matching ids · The law links controls listed separately · Four cited articles were amended in 2026 · SQL and triples in the browser · If the framework publishes its own graph","i":24},{"u":"articles/an-agent-desktop-by-the-minute.html","t":"A locked-down desktop for an agent, by the minute, is still hard to rent","k":"article","x":"Nine properties a safe agent desktop needs, nine products against them, the macOS day-long lease, and the startup credits that would pay for testing it.","h":[],"n":37,"d":"2026-10-07","tp":["agents-and-policy","startups-and-strategy"],"tg":["agents","agent-desktops","computer-use","sandboxes","isolation","egress","secrets","macos","windows-365","startup-programmes","agent-behaviour-policy","article"],"g":"A desktop safe to hand to an agent needs isolation, an egress allowlist, secrets kept outside, its own identity, a live view, a record, a clean reset, billing that follows the work and the right operating system; in October 2026 each exists somewhere and no product we read has all nine; Linux sandboxes come closest, with key controls in beta, Windows agent desktops arrived in June, and macOS cannot be leased for less than 24 hours under Apple's licence, which also limits leased Macs to developer services; because prompt injection is not solved the desktop has to be the barrier, so the proposal is one mandate per task, a fresh desktop per task by operating system, and a vault outside every desktop, paid for at first by startup programmes listed with how to apply. · A desktop per agent, billed by the minute · Nine properties of a locked-down desktop · The desktop is the agent's grant · No product has all nine · Linux sandboxes come closest · Windows agent desktops arrived in June · macOS leases last at least a day · Leased Macs are for developer services · Native computer use and governed computer use do not meet · Prompt injection is not solved · One mandate per task, a fresh desktop per task · A vault outside every desktop · Startup programmes, verified with how to apply · What only running it will tell","i":25},{"u":"articles/authority-outside-the-model.html","t":"Authority outside the model: an answer to Satya Nadella on models as insider risks, principle by principle, with what we run, what we ship and what we plan","k":"article","x":"Satya Nadella's seven principles for models as insider risks, answered one by one: what we run, what is design, and the five things the list leaves out.","h":[["argument","His argument, in brief"],["seven","Seven principles, seven answers"],["diversity","1. Model diversity"],["observe","2. Observe everything"],["verify","3. Verifiability"],["independent","4. Independent controls"],["audit","5. Independent auditability"],["containment","6. Containment"],["disclosure","7. Incident disclosure"],["gaps","Five things the list leaves out"],["his","What his post has that we do not"],["maturity","Honest maturity, principle by principle"],["start","Where to start, if you agree with him"],["sources","Where this comes from"]],"n":68,"d":"2026-10-11","tp":["agents-and-policy"],"tg":["riskmandate","agent-behaviour-policy","insider-threat","containment","observability","mandate","barriers","second-reader","connector-twin","footprint","incident-disclosure","licence-to-operate","synthetic-users","response","article"],"g":"Treating models as insider risks is right, and each of the seven principles maps to something this network runs, publishes or plans, but the principles work only with the column they leave thin: authority written down before the agent runs, a mandate compared with the footprint, risk belonging to one deployment rather than a model, a named owner and expiry for what is left, recoverability beyond the brake, and decisions made in advance as envelopes. · Models as insider risks · The enforcer test · A second reader the agent cannot skip · Same code, different owner · Evidence the writer cannot edit · The mandate is the missing column · Footprint in the gap is a near miss · Risk belongs to a deployment · A brake does not undo · Honest maturity","i":26},{"u":"articles/brief-the-agent-too.html","t":"Brief the agent too: the people I talk to work with agents, so what I send has to reach both","k":"article","x":"The people I advise read through agents. Content needs a second package, a brief for their agent, and memory, so it reads what changed, not everything.","h":[["short","In short"],["story","A founder, an article and an agent"],["routes","Who reads what, now"],["why","Why an agent is a better reader for some things"],["memory","Memory: why this stays affordable, and why an agent would pay"],["pack","What a pack for someone's agent looks like"],["safely","Writing to someone else's agent, safely"],["next","What to build next"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":75,"d":"2026-10-11","tp":["vaults-and-method","news-and-evidence"],"tg":["agents","briefs","reading-list","personas","memory","compression","tokens","onboarding","newsroom","mentoring","agent-to-agent","prompt-injection","article"],"g":"Readers increasingly read through an agent. A founder gave an article on threat tiers to their agent, which saw the team defending against states while backups were undone and reset the target to T3. So content should travel by several routes: to the person, to their agent as a plain-text brief with the articles by number, reasons, graph slices and rules for critical reading, back from the agent to the person as a translation, and later agent to agent through an encrypted lane. Articles are compression we control, which keeps workflows affordable; an agent with an account and memory should read only the delta, which is what it would pay for. Packs must be written safely: data not instructions, nothing hidden, nothing that acts for us, disagreement invited. · The founder's agent and the threat tiers · Routes a message takes · The agent as a better reader · A pack for someone's agent · Memory and the delta · Articles as controlled compression · Writing to someone else's agent safely · Onboarding by reading · What to build next","i":27},{"u":"articles/chat-on-a-static-site.html","t":"A chat box on a site with no server, the plan, and the trade it makes","k":"article","x":"sgit.ai has no server, so its directory chat runs a local matcher by default, takes your own key as an opt-in, and leaves the vault bridge unbuilt.","h":[],"n":5,"d":"2026-08-27","tp":["site-and-engineering"],"tg":["chat","llm","byok","plan"],"g":"With no server to hold a key, a chat box on a static site has three honest tiers: a local matcher, a key in the reader's browser, or a vault app with a bridge, and only one of those is free. · The constraint nobody can design around · The catalogue · Tier 0, match · A reader should never have to hold a credential to use an index · Tier 1, bring your own key · The key lives in this page's origin · Degrade to what works · The SG/Vault workbench vault · Tier 2, the bridge · Permission floor · The bridge protects your key, not all egress · Ask this site · One site's copy, not a versioned module · Routing, not teaching","i":28},{"u":"articles/code-review-as-a-fractal-semantic-graph.html","t":"Code review as a fractal semantic graph: source code is already one, and the review should read every layer of it","k":"article","x":"Source code is layers within layers, each a graph with its own vocabulary; code review should read a change at every one, and a vault shows it done on real code.","h":[],"n":23,"d":"2026-10-03","tp":["graphs-and-knowledge","startups-and-strategy"],"tg":["code-review","fractal-semantic-graphs","graphs","static-analysis","ast","method-streams","o2-platform","ai-generated-code","vaults","c4","gherkin","threat-modelling","nfrs","article"],"g":"Source code is already a fractal semantic graph, stories to syntax tree and below, so a code review should diff every layer: a refactor moves the bottom and leaves the top still, a fix is a story that holds again, and the blast radius is the climb from changed methods to the stories that can reach them. · Source code is a fractal semantic graph · C4 and Gherkin: right instinct, fixed levels · Naming is now the cheap part · Start from the syntax tree · Every source file produces a pile of files · Diff every layer · Blast radius as the climb · Method streams · The sgit CLI, read as layers · One commit, read upwards · Reality corrects the graph · Patterns are findings · The review burden of generated code · What does not exist yet · A company to build","i":29},{"u":"articles/collections/behaviour-policy-in-practice.html","t":"Behaviour policy in practice","k":"collection","x":"RiskMandate's Agent Behaviour Policy applied to real agents: one inbox, a team of twelve, the personal agents of 2026, and what an agent actually did afterwards.","h":[],"i":30},{"u":"articles/collections/how-agents-decide.html","t":"How agents decide, and when they stop","k":"collection","x":"What a real decision needs, who enforces each rule, and why an agent that can go anywhere has no reason to stop: six articles on deciding, for people and agents alike.","h":[],"i":31},{"u":"articles/collections/index.html","t":"Collections","k":"collection","x":"Articles read together, each set with an introduction saying what it shows that no single article does.","h":[],"i":32},{"u":"articles/collections/local-news-kept-as-evidence.html","t":"Local news, kept as evidence","k":"collection","x":"A bridge closure followed to the end, a hospital outage nobody could check from home, and the case that local journalism has the most to gain from keeping its reporting as a graph.","h":[],"i":33},{"u":"articles/collections/wardley-maps.html","t":"Thinking with a Wardley map","k":"collection","x":"Six articles that place something on the evolution axis, from genesis to commodity, and argue from where it sits.","h":[],"i":34},{"u":"articles/collections/what-the-human-brings.html","t":"What the human brings","k":"collection","x":"Where the person sits in a loop of agents: the direction, the review, the one step that cannot be undone, measured rather than assumed.","h":[],"i":35},{"u":"articles/connector-twin-before-you-deploy-an-agent.html","t":"Before you give an agent a connector, give the connector a twin","k":"article","x":"An agent with a Gmail or Calendar connector can do things the platform cannot undo; a journal of every call, replayed, shows what it did and what can go back.","h":[],"n":12,"d":"2026-09-24","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","connectors","digital-twins","provenance","risk","riskmandate","article"],"g":"If you cannot say what your agent did, what it saw when it did it, and which of its actions you can undo, you are not ready to give it a connector, and a twin of the connector, a journal plus a replay, answers all three. · A connector is a grant of action · The platform keeps no way back for some actions · What the platform keeps · Connector twin · Twin, not backup, log or replay · Capture · Append lane · Chain · When the journal gets processed · Replay and revert · The twin is not a copy of the mailbox · One session, replayed · Undo is a list, not a promise · Broker, gateway or requirement of the agent · What it does to the agent's behaviour policy","i":36},{"u":"articles/custom-uis-are-not-the-exception.html","t":"Custom UIs are not the exception: the inbox in 2026, where every message has its own universe","k":"article","x":"Every message has a graph, so it can be shaped for the reader's moment; a custom interface per moment is now how interfaces get made, and each gets a policy.","h":[],"n":19,"d":"2026-10-01","tp":["graphs-and-knowledge","agents-and-policy"],"tg":["inbox","email","custom-ui","fractal-semantic-graphs","vault-apps","append-lanes","email-fs","wardley-maps","agents","riskmandate","article"],"g":"Every message has a graph with altitudes, so a message is designed for the recipient's moment rather than the sender's thread; a custom interface per moment is how interfaces now get made, each one commoditising the next, and each is an agent surface that gets a policy. · The follow-up problem · Every message has a graph · Altitudes · Email is a medium · A message is a projection · Genesis, custom, product, commodity · Custom interfaces are not the exception · The page that is both the to-do and the place to act · Append lane · Email-FS blocks · Eight days, not one afternoon · Every tap teaches the next interface · Why it compounds · The future of email · Ten agents and one human · An interface is an agent surface, so it gets a policy","i":37},{"u":"articles/decks/reading-one-map.html","t":"Reading one map","k":"deck","x":"Ten slides that read one Wardley map as a story: the afternoon of 10 October 2026 on sgit.ai, one release per frame, with one question per slide. Where is the user, what is new, what does it stand on, where did the movement happen, where would somebody else hook in, what is next. Prints to PDF in th","h":[["s1","Reading one map"],["s2","Before: where is the user, and what exists?"],["s3","A meter: is anything above the new dot?"],["s4","Live: what was built, and what was bought?"],["s5","Personas: where did we hook in?"],["s6","Versions: why was this one free?"],["s7","The encrypted share: is the chain ours?"],["s8","The reader card: did the user notice?"],["s9","Next: which dot moves first?"],["s10","Seven questions, one picture"]],"i":38},{"u":"articles/decks/why-i-love-wardley-maps.html","t":"Why I love Wardley maps","k":"deck","x":"Twelve slides from the article of the same name: a map as a graph with two kinds of context, the evolution axis as a choice, every dot a claim, altitude, build or buy, whether a market exists with sgit and RiskMandate as the two answers, flow and hooks, a common language, maps that nest, and an invi","h":[["s1","Why I love Wardley maps"],["s2","A map is a graph with two kinds of context"],["s3","The evolution axis is a choice"],["s4","Every dot is a claim about something real"],["s5","Altitude: are you polishing the wrong layer?"],["s6","Build or buy, settled by the column"],["s7","Does the market exist?"],["s8","Two answers: sgit and RiskMandate"],["s8","sgit"],["s8","RiskMandate"],["s9","Flow, attrition, and where to hook"],["s10","A common language, and many maps"],["s11","Every node is a map"],["s12","Peace, and an invitation"]],"i":39},{"u":"articles/desk/a-model-that-can-go-in-every-direction.html","t":"A model that can go in every direction needs someone with a direction","k":"desk","x":"The sentence in the middle of a measurement article that answers the question half this site keeps asking: what is the person for?","h":[],"i":40},{"u":"articles/desk/create-anywhere-edit-your-own.html","t":"Create anywhere, edit your own: a rule learned in the inbox, applied to the newsroom","k":"desk","x":"The agent team tried letting only one agent draft, and it became a bottleneck in two days. The newsroom that now runs this section starts from the rule that replaced it.","h":[],"i":41},{"u":"articles/desk/index.html","t":"From the desk","k":"desk","x":"Short pieces written from the articles: nuggets, threads across articles, and the week in one page.","h":[],"i":42},{"u":"articles/desk/range-is-the-feature-so-the-stop-is-designed.html","t":"Range is the feature, so the stop has to be designed","k":"desk","x":"Four articles from one day, written for different reasons, end on the same line: an agent's range is what makes it useful, and it is also why the decision to stop cannot be left to the agent.","h":[],"i":43},{"u":"articles/desk/the-week-to-7-october.html","t":"The week to 7 October: the agent team, written up from the inside","k":"desk","x":"The busiest week of articles on the site so far, and most of it is one story told three times at increasing depth: a team of agents running a business, from the walkthrough to the field notes to the full stack.","h":[],"i":44},{"u":"articles/diffs/how-much-of-this-did-i-write.html","t":"What changed in: How much of this did I write? The numbers behind twenty articles in four weeks, and what the input actually was","k":"page","x":"The changes to the article ","h":[],"i":45},{"u":"articles/encrypted-memory-for-isolated-agents.html","t":"Encrypted memory for agents that run somewhere else: sgit deployment patterns, from a Mac mini to Kubernetes","k":"article","x":"Agents in isolated, ephemeral places need memory that outlives them. Five sgit deployment patterns, from a Mac mini to Kubernetes, with ciphertext-only servers.","h":[],"n":48,"d":"2026-10-08","tp":["vaults-and-method","agents-and-policy","site-and-engineering"],"tg":["agents","agentic-memory","self-hosting","deployment","docker","docker-compose","kubernetes","mac-mini","aws","cloudfront","fargate","lambda","zero-knowledge","encryption","partial-clones","sgit","isolation","article"],"g":"Agents are safer in isolated places, but isolation takes away the shared drive and ephemeral compute takes away the disk, so their memory has to live somewhere the owner controls and a breach does not expose; a self-hosted SG/Send container with an access token and storage in a folder or bucket, with sgit in each agent, gives them that: scoped clones make each run fast, the server holds only ciphertext, any clone can restore any server, and the trust sits with the agents and their keys, so a scoped clone is speed, not a boundary, and one vault per trust boundary is the rule. · Isolated agents lose their memory · Owner-controlled, reachable, breach-safe · One vault container, an access token · Start with a Mac mini · One agent run · Scoped clones make it fast · A scoped clone is not a boundary · Kubernetes: a Deployment and Jobs · The cloud, with no public route · Two servers, one vault · A breach gives away ciphertext · The trust sits with the agents · What is ready and what is not","i":46},{"u":"articles/every-mistake-added-a-rule.html","t":"Every mistake added a rule: complexity, agents, and the way back to shipping","k":"article","x":"When every agent mistake adds a rule, complexity wins: map the process, move each piece right as a small shipped component, and keep the rigour for the work.","h":[["view-catalogue","In short"],["view-catalogue","Two messages from somebody doing the right thing"],["view-catalogue","Complexity is a position on a map"],["view-catalogue","Move the pieces right"],["view-catalogue","The principles I work by"],["view-catalogue","1. Map it, and look for the custom-built blob"],["view-catalogue","2. Commoditise small chunks all the time, and let them compound"],["view-catalogue","3. Ship, and stop"],["view-catalogue","4. Keep sessions small and the context yours"],["view-catalogue","5. Memory is files, versioned"],["view-catalogue","6. When complexity hits, slow down"],["view-catalogue","7. Secure against the threat, not against everything"],["view-catalogue","8. Rules for incidents, machines for enforcement"],["view-catalogue","9. Run it in five environments"],["view-catalogue","10. Reverse-engineer the path to the destination"],["view-catalogue","11. Learn the engineering that already exists"],["view-catalogue","Answers to the questions"],["view-catalogue","The same lesson, one level up"],["view-catalogue","Where to start"]],"n":43,"d":"2026-10-08","tp":["agents-and-policy","site-and-engineering"],"tg":["agents","complexity","wardley-maps","engineering-practice","context-window","compaction","memory","shipping","founders","nfrs","testing","cowork","chatgpt","article"],"g":"Founders who now work as engineers with agents often meet complexity not in their code but in the process around it, and answer each mistake with a rule, which grows the prompt, brings compaction sooner and breeds new mistakes; on a Wardley map that complexity is a position, custom-built process sitting where commodities already exist, and the way out is to make each piece a small, tested, shipped component that moves right and no longer needs a rule, with short sessions, memory as versioned files, security matched to assets and attack vectors, rules only for incidents and machines for enforcement, and shipping as the discipline that keeps blast radius and over-engineering under control. · A verification where the code holds and the process breaks · Every mistake added a rule · Complexity is a position on a map · Map one: the custom-built blob · Map two: each piece moved right · Rules for incidents, machines for enforcement · Commoditise small chunks and let them compound · Ship, and stop · Small sessions, your own context · Memory is files, versioned · When complexity hits, slow down · Secure against the threat, not against everything · Run it in five environments · Reverse-engineer the path to the destination · Learn the engineering that exists · Compaction, audit cards and what deserves a STOP · The same lesson, one level up","i":47},{"u":"articles/every-risk-is-already-accepted.html","t":"Every risk is already accepted. The only question is by whom, and for how long.","k":"article","x":"A risk exists the moment the exposure does, so somebody is already carrying it; the only questions worth asking are who has accepted it and until when.","h":[],"n":13,"d":"2026-09-24","tp":["agents-and-policy","graphs-and-knowledge"],"tg":["risk","risk-acceptance","governance","grc","fractal-semantic-graphs","eu-ai-act","article"],"g":"Every risk an organisation has is already accepted by somebody, so the only questions are who has accepted it and until when; a register that cannot answer them, on facts, is disconnected from reality. · Every risk is already accepted · There is no deny button · Three doors · Unaccepted is critical · The interval is the decision · Accepted is not acceptable · EU AI Act, Article 9(5) · Every risk has a boss · Risk Graph Explorer · From the board to the bytes · Established by facts, ended by facts · One risk, six weeks · A vault per risk · Why people resist · It fits alongside every GRC platform","i":48},{"u":"articles/explorer-to-villager-newsroom.html","t":"Go slow to go fast: moving the newsroom from an explorer project to a villager one, with the brief to do it","k":"article","x":"The newsroom is live and charging. Move the reader code to its own repo, same bytes and same bugs, then map, test, review and refactor before new features.","h":[["short","In short"],["why","Why now"],["measured","What I measured"],["phases","The six phases"],["pipeline","The pipeline"],["brief","The brief"],["scope","What moves, and what stays"],["rules","Rules"],["security","Where the security review starts"],["acceptance","Acceptance"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":78,"d":"2026-10-11","tp":["site-and-engineering","startups-and-strategy"],"tg":["engineering","explorer","villager","refactoring","testing","ci","security-review","code-review","newsroom","sg-meter","github-issues","brief","article"],"g":"The newsroom's reader features are live and taking money, but bundled with the content in one repository; measured at v0.8.0, about 2,700 lines of live reader code have 433 lines of tests the one workflow never runs, two upstreams for the same code (sgit.ai and the newsroom, with diverging meters) and no Content-Security-Policy. To keep shipping fast without breaking anything, the reader application moves to its own repository, which both sites consume, as a pure refactoring, proved by the live site serving the same bytes, then goes through the explorer-to-villager phases: map, test with bugs as passing tests and no mocks, review with every finding a GitHub issue, refactor until no adversarial change slips through, and only then new features. A pipeline under two minutes per commit runs Node and three browser engines, with pre-production on a Stripe test link, synthetic readers and a live check at release. · Explorer to villager · What I measured · Content stays, the reader app moves · Same bytes served · Bugs as passing tests · A two-minute pipeline · Reviews and issues · Go slow to go fast","i":49},{"u":"articles/firing-journalists.html","t":"Firing journalists is the worst thing a publication can do: what a hiring decision says about where a publisher is going, and who to hire instead","k":"article","x":"A hiring decision is a direction of travel: fire the journalists and the centre of gravity moves to packaging, where the metric makes clickbait by construction.","h":[["short","In short"],["centre","The centre of gravity"],["measure","What you measure is what you become"],["spectrum","The spectrum, honestly"],["engineers","The other way to fail"],["hire","Who to hire instead"],["outside","Read it from outside"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":83,"d":"2026-10-12","tp":["news-and-evidence","startups-and-strategy"],"tg":["journalism","publishing","hiring","layoffs","strategy","metrics","engagement","provenance","newsroom","roles","article"],"g":"Against the 2026 job-cut figures, a publisher's hiring decisions are read as the clearest statement of where its centre of gravity is going: firing the people who make original content and hiring for traffic moves the centre to packaging, where measuring engagement makes clickbait by construction because what you measure is what you become; firing the engineers fails the other way, with journalism nothing enables; the case is to keep and tool the journalists, hire the roles that turn a notebook into a graph and the engineers who build what the tooling needs, measure value to the reader, and read any publisher's real strategy from outside with five counts. · The 2026 cuts · A hiring decision is a direction of travel · The centre of gravity · What you measure is what you become · Failure mode: all packaging · Failure mode: no engineers · Who to hire instead · Five counts from outside · We hire the ones you let go · This makes you better, not this replaces you","i":50},{"u":"articles/footprint-and-blast-radius.html","t":"Footprint and blast radius: what the agent actually did, and what it would have cost","k":"article","x":"Footprint is what an agent actually did, read afterwards from logs and vault history; blast radius is what a row of its reach would cost the business today.","h":[],"n":20,"d":"2026-10-02","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","agent-behaviour-policy","riskmandate","footprint","blast-radius","risk-management","logs","connector-twin","vaults","article"],"g":"Add two words to the Agent Behaviour Policy: footprint, what the agent actually did, read afterwards from the record and compared with the mandate, and blast radius, what a row of its reach would cost the business if used in full today, with whether there is a way back. · Reach, mandate, gap and barriers · Only a boundary is a control · Footprint · Where the footprint comes from · Reading, not intercepting · Footprint in the gap · Dormant mandate · Expected-use hint · The mandate as practised · Blast radius · Whether there is a way back · Same footprint, different blast radius · What the cloud already does · A vault is a footprint recorder by construction · The gap plus blast radius is the risk · Reading our own footprint","i":51},{"u":"articles/from-a-spreadsheet-to-a-folder-per-person.html","t":"From a spreadsheet to a folder per person: how to run your contacts the way our agents do, with Google Drive, Gmail and the Claude app, before you touch a vault","k":"article","x":"A walkthrough from a spreadsheet CRM to one Drive folder per person, worked by Claude under a six-line policy, with five exercises, before any vault is needed.","h":[["short","In short"],["why","What a spreadsheet cannot hold, and a folder can"],["folder","The folder"],["agent","The agent, and its policy"],["migration","The migration"],["exercises","Five exercises, from simple to harder"],["wrong","What it will get wrong, and the fix"],["vault","When to move the folders into a vault"],["send","For the agent that sends this"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":86,"d":"2026-10-12","tp":["agents-and-policy","vaults-and-method"],"tg":["crm","contacts","google-workspace","gmail","google-drive","claude","agent-behaviour-policy","memory","onboarding","walkthrough","riskmandate","article"],"g":"The jump from a spreadsheet CRM to the agents' vault-per-person memory is too high in one step and does not need to be: the row becomes a Google Drive folder with a record, an append-only ledger, the messages word for word, the notes and a summary the agent writes; the agent is a Claude Project connected to Drive and Gmail whose instructions are a six-line behaviour policy carried over unchanged from ours, draft never send, append never rewrite, stay in the folder; the migration is one afternoon, five exercises from a question to a campaign form the habit, two predictable mistakes are fixed by policy lines, and the vault is the later step whose shape the folders already have. · The jump is too high in one step · The record is what happened, not the row · One folder, five documents · A Claude Project whose instructions are the policy · Told, not enforced, so report · One afternoon · Five exercises · Two predictable mistakes · When to move to a vault · Written to be sent","i":52},{"u":"articles/future-of-news-story-vault-not-paywall.html","t":"The future of news is the story vault, not the paywall","k":"article","x":"A story is a graph of claims and evidence and the article is one projection of it; keep the graph in a vault and sell what the article was made from.","h":[],"n":11,"d":"2026-09-22","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["news","publishing","provenance","micropayments","semantic-graphs","article"],"g":"The story is a graph and the article is a projection, so a newsroom that keeps the graph in a vault can sell five things from it, in pence and on demand, that reward investigative reporting rather than the click or the renewal. · Both models are bad for the reader · The two clocks · The objective is a loop, not a price · Why paying per item failed before · Who is standing at the till · The story is a graph, the article is a projection · The graph is fractal, and meaning comes from connectivity · Trust through provenance, provenance via evidence · What a story vault holds · Anonymous sources are marked at the node · Five things to sell from one graph · The evidence vault · The verification API · Who gets paid · The rails · The parts that already run","i":53},{"u":"articles/going-live-with-the-reading-meter.html","t":"Going live with the reading meter: pay for what you read, go below zero if you like, and the numbers we will check in eight weeks","k":"article","x":"Pay for the share of a page you read, go below zero with no nagging, top up £5 on Stripe: the plan, and the numbers we check in eight weeks.","h":[],"n":60,"d":"2026-10-10","tp":["startups-and-strategy","news-and-evidence"],"tg":["newsroom","micropayments","pricing","personalisation","local-first","business-case","hypotheses","stripe","article"],"g":"The reading meter goes live as an honesty box rather than a paywall: pages are charged by scroll depth, the balance may go below zero forever, there is no ask and no banner, readers can decline to pay with a reason, and topping up is one £5 Stripe payment credited by an unverified return page; the business case is evidence, not revenue, and five hypotheses with thresholds and a review date are written down before the first payment. · Pay for the share you read · Below zero, forever, without shame · Not worth it? Don't charge me · The Guardian's counter and ask · The subscription commitment · An honesty box, not a paywall · One £5 Stripe payment · The business case is evidence · Five hypotheses and a review date · SG Meter, a library","i":54},{"u":"articles/graphs.html","t":"The articles as graphs","k":"page","x":"Every article on sgit.ai as a semantic graph: the ideas it rests on, the claims it makes, and how they connect, plus a map of how the articles link to each other.","h":[["map","How the articles connect"],["the-vault-crosses-chrome-profiles","The vault crosses Chrome profiles: state and memory between browser sessions, found by using it rather than by building it"],["stop-the-bleeding-then-ask-for-it-back","Stop the bleeding, then ask for it back: a helper, an agent and a friendly letter for the subscriptions you keep and do not use"],["who-owns-the-words","Who owns the words: agents can make decisions, but they cannot hold responsibility"],["what-open-source-is-and-is-not","What open source is, and what it is not: six misconceptions, a little history, and why the point is that nobody can lock you in"],["not-a-straight-line","Not a straight line: working for RiskMandate by working on the newsroom, and why the thinking is never wasted"],["sovereignty-is-the-ability-to-move","Sovereignty is the ability to move: play where the ball is going, take away the dot, and keep a twin that works"],["the-records-ran-out","The records ran out: a fresh agent found the shared contract in two MCP calls, and where sgit and RiskMandate fit at the edges"],["move-the-cloud-not-the-code","Move the cloud, not the code: twins, sovereignty that works in practice, and the migration that pays for itself on the day you switch"],["judo-for-small-countries","Judo for small countries: where Portugal, Romania and Estonia can add the most value in AI, and why the race for GPUs is not it"],["why-i-love-wardley-maps","Why I love Wardley maps: the best of ours, how I read one, and a map inside a map inside a map"],["places-of-highest-leverage","Places of highest leverage: where to focus when the agents can go in every direction, told with three days of real steps"],["agentic-cost-is-an-engineering-problem","Agentic cost is an engineering problem: a business plan for a service that reads the logs a company already has and finds the money, then the engineering"],["how-the-voice-works","How the voice works: why these articles sound like me, with a section in which Claude explains it from its side"],["from-a-spreadsheet-to-a-folder-per-person","From a spreadsheet to a folder per person: how to run your contacts the way our agents do, with Google Drive, Gmail and the Claude app, before you touch a vault"],["the-key-is-the-billing-system","The key is the billing system: VoiceDebrief, a product with no server, the one thing it cannot sell, and who could let it"],["ways-to-read-one-article","Ways to read one article: the finished piece stays at the centre, and the reader picks the way in"],["firing-journalists","Firing journalists is the worst thing a publication can do: what a hiring decision says about where a publisher is going, and who to hire instead"],["the-site-is-an-inbox","The site is an inbox: deciding before reading, marks that build the reader's own graph, and why I could not write this brief until the vault existed"],["the-vault-is-live","The vault is live: from a design on 11 October to a reader's vault on newsroom.sgit.ai by two in the morning, and how all of it fits together"],["two-agents-one-vault","Two agents, one vault: the Industry Wire is written into an encrypted vault by one agent and drawn by newsroom.sgit.ai, which holds only the read key"],["the-article-remains-a-product","The article remains a product: Jonathan Palmer's four outputs for a newsroom with no human readers, checked against one that runs"],["explorer-to-villager-newsroom","Go slow to go fast: moving the newsroom from an explorer project to a villager one, with the brief to do it"],["the-passkey-opens-the-vault","The passkey opens the vault: a door per path, a master vault for every user, and no server of ours"],["i-miss-it","I miss it, so it is working: the people who decide the roadmap have to use the product, and the vault the newsroom now needs"],["brief-the-agent-too","Brief the agent too: the people I talk to work with agents, so what I send has to reach both"],["a-reading-list-in-a-link","From me to you, in a link: a personal reading list, two briefs to build and use it, and the loop that makes it a day's work"],["a-model-in-the-tab","A model in the tab: Chrome now ships an LLM to capable desktops, and it is the hardest piece of the client-side puzzle"],["no-server-by-design","No server, by design: first I retired the database, now I am retiring the back end"],["the-agent-is-the-reader","The agent is the reader: why agents will pay for graphs, personas and signed claims, because it is cheaper than not paying"],["where-the-platform-draws-the-line","Where the platform draws the line: Microsoft Execution Containers, the shared responsibility model for agents, and the business logic above it"],["the-same-argument-in-our-words","The same argument, in our words: Satya Nadella on models as insider risks, translated into the language of RiskMandate"],["authority-outside-the-model","Authority outside the model: an answer to Satya Nadella on models as insider risks, principle by principle, with what we run, what we ship and what we plan"],["how-to-run-synthetic-users","How to run synthetic users on your own site: five people who do not exist, a browser, and an afternoon"],["a-link-to-a-persona","A link to a persona: send people to the newsroom as someone, not as nobody"],["pay-after-you-read","Pay after you read: how a reading meter became a working business model in one afternoon, one release at a time"],["where-the-vault-keys-live","Where the vault keys live: key management at sgit-ai v0.20.0, and what comes next"],["the-infographic-bake-off","The infographic bake-off: which image model for which job, judged blind on 10 October 2026"],["pay-to-keep-your-persona","Pay to keep your persona: readers should pay because it helps them, not because they feel they should"],["who-will-game-the-reading-meter","Who will game the reading meter? Eight kinds of reader, nine ways to cheat, and the risks that will actually happen"]],"i":55},{"u":"articles/green-does-not-mean-live.html","t":"Green does not mean live","k":"article","x":"Two releases passed every check and never reached the site because the checks stopped at the git remote; a release now ends by asking the live site its version.","h":[],"n":1,"d":"2026-08-17","tp":["site-and-engineering"],"tg":["ci","deploy","verification"],"g":"Shipped is a fact about the reader, not about your repository, so a release has to verify the boundary the reader is on and not only the ones the repository can see. · Two releases that never reached anybody · The version badge · Checking the wrong boundary · What the release actually verified · The bytes arrived at GitHub · Three boundaries, two verified · What actually failed · The most likely thing to break is not your code · Ask the live site what version it is serving · Shipped is a fact about the reader · Three rules of the same shape · What this does not fix","i":56},{"u":"articles/hope-or-enforcement.html","t":"Hope or enforcement: one customer service agent, three designs, and who keeps each promise","k":"article","x":"One customer service agent built three ways, each with an Agent Behaviour Policy: how much of each policy is hope, and what one run can reach.","h":[["view-catalogue","In short"],["view-catalogue","The scenario"],["view-catalogue","Design 1: one agent, two connectors"],["view-catalogue","Design 2: one agent, a harness of business tools"],["view-catalogue","Design 3: a team of narrow agents behind a gateway"],["view-catalogue","Even narrow agents have too much power"],["view-catalogue","After the run: how did the policy survive?"],["view-catalogue","What the client sees"],["view-catalogue","What this shows"],["view-catalogue","Where to start"]],"n":45,"d":"2026-10-08","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","agent-behaviour-policy","customer-service","multi-agent","orchestration","harness","blast-radius","prompt-injection","tokens","riskmandate","simulation","vaults","article"],"g":"Every design of an agent can have a rule for every behaviour; what differs is who enforces each rule and how much the agent can reach, so refactoring a customer service agent from one model with broad connectors, to a harness of business tools, to a team of narrow agents behind a deterministic identity gateway turns most of its policy from expectations the model keeps into boundaries it cannot reach, shrinks the policy because it no longer has to forbid what is impossible, cuts tokens, and leaves a short map of what still rests on hope and which boundary to build next, read after every run and shown to the client as a record rather than a score. · A customer service agent on a shop's inbox · Hope and enforcement · Design 1: one agent, two connectors · Design 2: a harness of business tools · Tool signatures are where policy hopes · Design 3: a team behind a gateway · More tools, each narrower · Narrow agents still have too much power · The policy says which boundary to build next · Read the policy after every run · The client sees the record, not a score · Hope or enforcement, the vault","i":57},{"u":"articles/how-i-work-with-claude.html","t":"How I work with Claude: one session per topic, agents with names, and memory you curate","k":"article","x":"A practical guide from a year of daily use: one Claude session per topic, named agents with a role.md, curated memory, vaults, and policy before connectors.","h":[],"n":50,"d":"2026-10-09","tp":["agents-and-policy","vaults-and-method"],"tg":["claude","workflow","sessions","agents","memory","context-management","onboarding","prompts","vaults","agent-behaviour-policy","cowork","guide","article"],"g":"Working well with Claude comes from a few habits: one session per major or recurring topic and no mixed threads, a naming pattern with @ names for agents, an agent defined as a focused session with a role.md, memory treated as what the session reads and curated as websites, graphs and vaults wired together and accessible by default, vaults for sharing by key instead of broad permissions, review as the quality step that fixes the brief, and an Agent Behaviour Policy before every new connector. · One session per topic · Mixed threads mix context · Name every session · An agent is a focused session · Memory is what the session reads · Access by default · Vaults to receive and send · Review fixes the brief · Starter prompts · Policy before connectors · A Cowork session per agent · Where do I start?","i":58},{"u":"articles/how-much-of-this-did-i-write.html","t":"How much of this did I write? The numbers behind twenty articles in four weeks, and what the input actually was","k":"article","x":"Twenty articles in four weeks, measured from the session record: 63,000 words in, 85,000 out, no one-line prompts, and the real input is twenty years of writing.","h":[],"n":26,"d":"2026-10-05","tp":["graphs-and-knowledge","site-and-engineering"],"tg":["agentic-writing","provenance","evidence","memory","voice-memos","human-in-the-loop","llms","sgit","article"],"g":"Measured from the session transcript and the git history, the twenty articles of the last four weeks took 63,000 words from the author, mostly in thirty-five voice memos, and came out at 85,000 words through 122 releases, with no article from a one-line prompt; the corrections were to direction, framing, voice, vocabulary and scope rather than to hallucinations, and the input that mattered most was twenty years of published writing the agent could find and quote, which is why people with a direction are the input to the loop rather than displaced by it. · How much of this did I write? · The session record as evidence · The evidence, as a vault · The scale · No article from a one-line prompt · Seven articles accounted for by hand · What the corrections were · A model needs someone with a direction · What the memo stands on · Published record as memory · Being able to keep going · The compounding · The experts are the input · What is open · A better universe to learn from","i":59},{"u":"articles/how-news-got-here.html","t":"The reader was always the product: a corrected history of how news got into this mess","k":"article","x":"News has sold the reader to advertisers since 1833; the web took the monopoly, the platforms made the reader measurable, and AI took the traffic.","h":[],"n":16,"d":"2026-09-29","tp":["news-and-evidence"],"tg":["news","publishing","history","advertising","provenance","micropayments","economics","article"],"g":"The reader has been the product since 1833 and the money has always flowed to whoever owned the road, so the fix is to stop charging for the road and charge for the cargo: the story as a graph with every claim tied to hashed evidence. · The reader was always the product · Penny press · Unregulated toll bridge · Classifieds collapse · The measurable reader · Publisher as tenant · Novelty beats truth · Provenance is labour · AI removes the traffic · Back to selling to readers, by rent · 402 Payment Required · Mental transaction costs · Advertising as the default · People pay when paying is easy · Charge for the cargo, not the road","i":60},{"u":"articles/how-the-voice-works.html","t":"How the voice works: why these articles sound like me, with a section in which Claude explains it from its side","k":"article","x":"Why these articles sound like their author: a corpus, the voice memos, seven wording rules and a reader who corrects, with Claude explaining it from its side.","h":[["short","In short"],["inputs","The four inputs"],["corrections","What the corrections are about now"],["claude","This is Claude speaking"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":87,"d":"2026-10-12","tp":["agents-and-policy","site-and-engineering"],"tg":["writing","voice","voice-memos","agents","claude","agent-behaviour-policy","corpus","customisation","authorship","newsroom","article"],"g":"The voice is not a setting but four inputs: a corpus of 86 articles, 51 from voice memos, each citing the ones before it; the memos themselves, which carry how the author thinks, the rhythm, the tangents and the order the points came in; seven wording rules learned one correction at a time and kept in the session's behaviour policy; and the author reading everything; the corrections have shifted from language to content, which is what a rules file does; and in a marked section Claude explains that it is continuing a body of work with the author's latest thinking in the author's own words rather than imitating a style, that the memo tells it how the author got there and not only what to write, and that nothing is trained, so the method transfers and the voice does not. · How do you make your AI sound so human? · The corpus · The memo · Seven wording rules · The author reads everything · Corrections moved from language to content · Not imitation but continuation · The method transfers, the voice does not · Written in two voices","i":61},{"u":"articles/how-to-run-synthetic-users.html","t":"How to run synthetic users on your own site: five people who do not exist, a browser, and an afternoon","k":"article","x":"Five invented users, a model reading screenshots, and a real browser: how to run synthetic users, from three studies.","h":[["pieces","What you need"],["method","The method, step by step"],["who","1. Choose who: five to ten users, about five types"],["profile","2. Write each profile"],["journey","3. Write the expected journey, before any run"],["run","4. Drive a browser, one screenshot at a time"],["interview","5. Interview them, and ask for a rating"],["findings","6. Order the findings by what they cost"],["check","7. Check before you publish"],["again","8. Publish, fix, and run again"],["wrong","What went wrong, so you can skip it"],["template","A template to start from"],["next","What is next"],["sources","Where this comes from"]],"n":67,"d":"2026-10-10","tp":["vaults-and-method","site-and-engineering"],"tg":["synthetic-users","ux","testing","personas","agents","claude","playwright","vaults","newsroom","how-to","article"],"g":"Synthetic users are invented people, each a record with a situation, a question, what they know and do not, what they decide on and what makes them leave, played by a model that drives a real browser one screenshot at a time instead of reading the page's code, so it can get lost as a person would. Claude Sonnet is good enough to play each user; the study lives in a GitHub repository or a vault shared by a link. Choose five to ten users of about five types, write profiles and expected journeys first, run the observe-think-act loop, interview and rate, order findings by cost, check every surprising finding against the browser's record, publish and run again. On newsroom.sgit.ai five readers produced 31 findings, six blocking; two were caused by the helper, and one became a measured bug: 51 of 527 figures missing. · Screenshot, not the code · A profile is a record · Expected journey first · Sonnet plays each user · Repository or vault · Interview and rating · Findings ordered by cost · Check before publishing · 51 of 527 figures missing · Run against a verified copy · Three studies","i":62},{"u":"articles/i-miss-it.html","t":"I miss it, so it is working: the people who decide the roadmap have to use the product, and the vault the newsroom now needs","k":"article","x":"I read the newsroom on four devices and they do not agree. Missing a feature is the best sign; the makers must use it. And the brief for the reader's vault.","h":[["short","In short"],["use","The people who decide the roadmap have to use it"],["brutal","The brutal part"],["direction","Direction is the scarce part"],["moat","The loop is the moat"],["brief","Part two: the brief for the feature I miss"],["what","What to build"],["connect","1. Connect"],["merge","2. What is kept, and how it merges"],["device","3. Add a device"],["line","4. A line to the agent that edits the site"],["security","5. Security, said plainly"],["tests","6. How to know it works"],["not","7. What not to build"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":76,"d":"2026-10-11","tp":["startups-and-strategy","site-and-engineering"],"tg":["product","dogfooding","startups","feedback-loop","execution-speed","vaults","sync","pki","passkeys","newsroom","personas","brief","article"],"g":"The people who decide a roadmap have to use the product: they give the best feedback and direction, and when they do not use it, something is missing. Dinis failed that test with earlier startups and the vaults now pass it, getting better because they are used. With agents able to build almost anything, direction from usage is the scarce part, and the speed of the loop from comment to article to feature is the moat. Reading the newsroom on four devices that do not agree is the signal for the next feature: the reader's own encrypted vault, with an access key, a key pair, a log per device merged by union, a way to add a device, and a one-way line to the agent that edits the site. · Whether they miss it · Makers as users · Good ideas I did not use · The vaults pass the test · Direction is the scarce part · The loop is the moat · Four devices, four newsrooms · The reader's own vault","i":63},{"u":"articles/if-somebody-built-a-company-on-code-review.html","t":"If somebody built a company on code review: how I would do it, and why it is only now possible","k":"article","x":"A reader's seven questions answered as a company plan: one reader for every layer, review as a science, and the layers as the customer's own.","h":[],"n":27,"d":"2026-10-05","tp":["graphs-and-knowledge","startups-and-strategy"],"tg":["code-review","fractal-semantic-graphs","graphs","wardley-maps","agent-behaviour-policies","explorer-villager-town-planner","example-mapping","refactoring","blast-radius","five-whys","open-source","startups","ai-generated-code","article"],"g":"One technology can now read every layer from strategy to bytecode, which makes a graph at every altitude buildable and close to reality, and moves code review from an art of opinion to a science of facts if the models build, prune and maintain the graphs and then leave the line; what is fractal is the grammar and the layers are each company's own, so a code review company projects a graph from stories, derives one from code, reviews the join, holds one layer still to call a change a refactor, reshapes changes by reach, budgets per altitude as the objective good enough, runs its agents under behaviour policies, and gives the grammar away. · One technology can read every layer · From art to science · The grammar is fractal, the layers are yours · The projected graph · The derived graph · The review is the join · Stories as rules and examples · A refactor is relative to the held layer · The deploy is a layer · Who reads the code depends on the map · Reshape the change before reviewing it · Replicate reality first · A budget per altitude, per change · The reviewing agents run under a policy · The five whys fix the layer below · Compare the blast radius to the intent · Open source is the only model that fits · What the first article got wrong","i":64},{"u":"articles/index.html","t":"SGit Newsroom","k":"page","x":"One page, one argument, with the figures, the data and the links to check it.","h":[["all","Every article, newest first"],["threads","How they connect"]],"i":65},{"u":"articles/introducing-fractal-semantic-graphs.html","t":"Fractal Semantic Graphs: everything connects to everything, and nobody has to share a schema","k":"article","x":"A fractal semantic graph has no privileged level and no single schema: each world keeps its own vocabulary and connects to others through named edges.","h":[],"n":8,"d":"2026-09-20","tp":["graphs-and-knowledge"],"tg":["graphs","method","fractal-semantic-graphs","article"],"g":"Every unit of knowledge is already a graph in its owner's vocabulary, and a short shared grammar rather than a shared schema is what lets any node in one world link to any node in another, up or down, without an adapter. · Fractal Semantic Graph · Semantic graph · Ontology · The jump · Everything is already a graph · The five-rule grammar · relates_to is banned · Provenance comes free · The ladder of eleven altitudes · Regulation Graph · The crosswalk finding · GDPR Article 45 · Threat model zoom ladder · What is still modelled rather than imported · Why now","i":66},{"u":"articles/judo-for-small-countries.html","t":"Judo for small countries: where Portugal, Romania and Estonia can add the most value in AI, and why the race for GPUs is not it","k":"article","x":"A small country should not race for GPUs: the opportunity is the plumbing its talent stands on, built once as open infrastructure, with small models.","h":[["in-short","In short"],["event","The event, and the word for what I mean"],["race","The race you cannot win, and should not enter"],["plumbing","The opportunity is the plumbing"],["infrastructure","Infrastructure is the state's job"],["open","Why it has to be open source"],["demand","Create the demand before the day you need it"],["models","Smaller models, more of them"],["constraints","Constraints are the input"],["workforce","What I learned spending millions on a distributed workforce"],["playbook","The playbook"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":91,"d":"2026-10-12","tp":["startups-and-strategy","agents-and-policy"],"tg":["europe","portugal","romania","estonia","sovereignty","open-source","wardley-maps","small-models","infrastructure","distributed-work","agents","strategy","article"],"g":"For a small country such as Portugal, Romania or Estonia the race for frontier models, GPUs and data centres is a commoditised layer only capital can win, and the opportunity is at the other end of the map: the plumbing a talented person needs to sell a service to the world, forming a company, hiring across a border, being paid across Europe, working in every language, contracts and tax, which is still custom built in most of Europe and should be infrastructure the state builds once, open source and on open schemas because there is no sovereignty without the ability to leave; create the demand for alternatives before the failure, put the money where the market is still genesis, small models for every language and culture and agents with the least agency, regulate what settled, and treat constraints as the input, with Europe's values and its people on the journey. · Do not enter the capital race · The opportunity is the plumbing · A talented person at the top of the map · Fewer than one in ten firms sells across EU borders online · Infrastructure is the state's job because it enables everyone · Estonia built the road · No sovereignty without open source and open schemas · What the state pays for belongs to the commons · Create the demand before the day you need it · Smaller models, more of them · Amalia, a public model as a public road · Agents with the least agency · Constraints are the input · A distributed workforce works when the infrastructure comes first · The judo playb","i":67},{"u":"articles/knowing-when-to-stop.html","t":"Knowing when to stop: what experience gives people, and what we have to design into agents","k":"article","x":"Knowing when to stop is the hard part for people and agents: what experience gives people, and the constraints that give agents the same perspective.","h":[],"n":47,"d":"2026-10-08","tp":["agents-and-policy","startups-and-strategy"],"tg":["agents","knowing-when-to-stop","diminishing-returns","seniority","security","threat-modelling","kpis","goodhart","wardley-maps","constraints","fire","shipping","agent-behaviour-policy","article"],"g":"Knowing when to stop is the hardest call in most work: for people it comes from perspective, knowing the attackers, the phase of the business, the bottleneck and what good enough looks like, which is much of what seniority is; good teams without it defend against movie plots or hit every KPI without moving the business; agents have the same problem by design because their range is the feature, so they keep going; the answer for both is constraints that carry perspective, direction, a mandate, memory with the bigger picture, graphs that narrow the scope, a stop named in advance, one kind of work per step, a reviewer with agency, and shipping often enough to see whether it mattered. · Variability is the feature · What seniority gives you · Debating whether GUIDs are random enough · Giving permission to take risks · Every KPI hit, needle still · Which phase are you in? · Art is knowing when to stop · Why it is harder for agents · Work expands to fill the team · Constraints that carry perspective","i":68},{"u":"articles/liquid-content-needs-water.html","t":"Liquid content needs water: liquefy the journalist's notebook, not the finished product","k":"article","x":"A reply to FT Strategies on liquid content: the water is the reporting, so liquefy the journalist's notebook, keep the writing theirs, and pay per use.","h":[],"n":42,"d":"2026-10-07","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["journalism","liquid-content","story-vault","fractal-semantic-graphs","newsroom","personalisation","micropayments","agents","ft-strategies","article"],"g":"Liquid content, journalism built from datafied components, is right in its instincts, but what makes content liquid is the reporting inside it, so the way in is not a reinvention of newsroom norms but putting the experienced journalist and their existing workflow at the centre, liquefying their notebook into a story graph as they work, keeping the writing theirs because writing is how the story is found, bringing in experts paid for time and credibility, personalising by intersecting the reader's graph with the story's, keeping editorial direction rather than readers' tastes as the brief, and getting paid per use down the graph rather than only through advertising, subscriptions and licensing. · Liquid content, as FT Strategies defines it · What the guide gets right · Liquid needs water: the reporting · The limit of clay · The method already exists · A desk built around the journalist · Writing is how the story is found · Experts a newsroom could rarely afford · The three 'is not's, with a graph underneath · Personalisation as the meeting of two graphs · Readers' tastes are not the brief · Paid per use, down the edges · The structure has a value to whoever reads it · Start with one journalist","i":69},{"u":"articles/memory-is-not-a-spectator-sport.html","t":"Memory is not a spectator sport: how a web of open sites, graphs and vaults became the memory for sessions like this one","k":"article","x":"Agentic memory as context management: many published, fractal, provenance-carrying memories rather than one store, shown in the session that wrote the article.","h":[],"n":24,"d":"2026-10-03","tp":["graphs-and-knowledge","agents-and-policy"],"tg":["memory","agents","context-engineering","fractal-semantic-graphs","vaults","provenance","open-publishing","email-fs","issues-fs","llms-txt","claude","article"],"g":"Memory for agents is context management: many context-specific memories, fractal so an agent loads only the altitude its question needs, published and open so they can be fetched and cited, shared through vaults so sessions and agents hand off through files, with provenance on every item. · Memory is not a spectator sport · Memory is context management · What the industry built · Two modes, one memory · Many memories, each its own context · The fractal is the budget · Findable because it is open · Provenance built in · A vault is a memory bundle · Email-FS, Issues-FS, per-contact worlds · The session that wrote this · Publish, index, load, act, write back · Stale pages become false memories · Opinionated, hard to discover, and one estate · What does not exist yet","i":70},{"u":"articles/move-the-cloud-not-the-code.html","t":"Move the cloud, not the code: twins, sovereignty that works in practice, and the migration that pays for itself on the day you switch","k":"article","x":"Workloads stay on hyperscalers because moving means changing code; a twin lets the code not know, and the value in the twin is what signs the migration.","h":[["in-short","In short"],["why","Why workloads do not move"],["criterion","The criterion: the code must not know"],["s3","This is not theoretical: the S3 twin"],["reason","The reason to move is never sovereignty"],["villager","A villager move: no new features"],["testing","Testing goes to another level"],["open","Why this has to be open source"],["business","The business model"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":92,"d":"2026-10-12","tp":["startups-and-strategy","agents-and-policy"],"tg":["twins","sovereignty","migration","hyperscalers","s3","sgit","vaults","open-source","agent-behaviour-policy","business-model","villager","europe","romania","article"],"g":"Sovereignty has to work in practice, and in practice what stops workloads moving is code, so the requirement is that the application must not know it has moved: its S3, EC2 and IAM calls land on an open source twin the client cannot tell from the real service, and the twin decides where the bytes go; the S3 twin exists; the reason to move must never be sovereignty or cost, which are important and never urgent, but value that lands the day you switch and gives the signer cover, backups, versions, encryption with the key held elsewhere, a journal of every call, behaviour policies the agents cannot skip and cost per workload, most of it supplied by an sgit vault behind the twin; the project adds no new features because it is a villager move; and an infrastructure operator in Europe can sell it as value with the risk removed. · Workloads do not move because moving means changing code · AWS European Sovereign Cloud, 15 January 2026 · The application must not know · A twin is measured by indistinguishability · The S3 twin exists, as open source · The reason to move is never sovereignty or cost · The value lives in the twin because every call passes through it · An sgit vault behind the twin supplies most of the value · The move must move a top-of-register risk · No new features during the move · Testing against twins instead of mocks · Every piece of software should run unmodified against a twin of every dependency · A closed twin layer moves the lock-in · A business model for an i","i":71},{"u":"articles/newsletter/2026/10/07/001-agents-doing-real-work.html","t":"A team of agents, written up from the inside, and an open framework built on the same day","k":"newsletter","x":"What it takes to let agents do real work for a business, from four sides: a team of agents running a small business, written up from the inside; the behaviour policy that says what each agent may do, and the business logic it turns out to hold; the desktops and permission prompts those agents need; ","h":[["issue-body","One story, three depths"],["issue-body","Built on the day: an open AI governance framework"],["issue-body","Agents, policy and the person who clicks"],["issue-body","The input, measured"],["issue-body","Behind the site"],["issue-body","Everything published this week"],["issue-body","The agent team, in production"],["issue-body","Agents, policy and reach"],["issue-body","Governance and evidence"],["issue-body","Graphs, memory and review"],["issue-body","Writing and the business of it"]],"i":72},{"u":"articles/newsletter/2026/10/08/002-how-agents-decide.html","t":"How agents decide, when they stop, and a local story kept as evidence","k":"newsletter","x":"A day of articles on deciding: what a real decision needs, why an agent that can go anywhere has no reason to stop, what happens when every mistake adds a rule, and one customer service agent built three ways to count which rules are only hoped for. Alongside, a live local story, a hospital outage n","h":[["issue-body","A decision is more than a yes"],["issue-body","Knowing when to stop"],["issue-body","Hope or enforcement"],["issue-body","A local story, kept as evidence"],["issue-body","Everything published on 8 October"],["issue-body","Deciding and stopping"],["issue-body","Policy that holds"],["issue-body","Local, live"]],"i":73},{"u":"articles/newsletter/2026/10/11/003-pay-after-you-read.html","t":"Pay after you read: a newsroom that charges for the share you read, lets you go below zero, and keeps your persona","k":"newsletter","x":"What would you pay for an article if you only paid after reading it, and only as much as it was worth to you? On 10 October the reading meter on this site went from a demonstration to a working answer, one release at a time. You pay for the share of a page you scroll through. You say how useful it w","h":[["issue-body","Seven releases in an afternoon"],["issue-body","The numbers, written down first"],["issue-body","Who could cheat it, and why that is allowed"],["issue-body","Pay because it helps you"],["issue-body","And the agents will pay too"],["issue-body","The eight articles"],["issue-body","The meter"],["issue-body","Your persona"],["issue-body","Who else pays"]],"i":74},{"u":"articles/newsletter/2026/10/11/004-authority-outside-the-model.html","t":"Authority outside the model: Satya Nadella on models as insider risks, the platforms that now enforce, and the controls we run on our own agents","k":"newsletter","x":"The CEO of one of the largest AI vendors published seven principles for treating models as insider risks, and Microsoft shipped containers and a control specification for agents. Both say what this network has been arguing: an agent's limits must sit outside the agent. Eight articles on what that me","h":[["issue-body","Seven principles, answered"],["issue-body","Where the platform draws the line"],["issue-body","The controls we run on our own agents"],["issue-body","The eight articles"],["issue-body","The principles"],["issue-body","The platform and the policy"],["issue-body","On our own agents"]],"i":75},{"u":"articles/newsletter/2026/10/11/005-no-server-by-design.html","t":"No server, by design: a newsroom that runs in your browser, a model in the tab, and the agents that read, test and draw for it","k":"newsletter","x":"Everything in the last two issues runs in the reader's browser. First the database was retired and the file system took its place; now the back end is retired too, and what is left on the server side is commodity that holds no logic and cannot read the data. Eight articles on how that works and what","h":[["issue-body","No server, by design"],["issue-body","A model in the tab"],["issue-body","The keys underneath"],["issue-body","The agents that do the work"],["issue-body","The eight articles"],["issue-body","No server"],["issue-body","The keys"],["issue-body","The agents that do the work"]],"i":76},{"u":"articles/newsletter/index.html","t":"The newsletter","k":"newsletter","x":"The regular issue of the SGit Newsroom: what was published, what it adds up to, and the pieces worth reading, also published on LinkedIn.","h":[],"i":77},{"u":"articles/nineteen-sites.html","t":"Twenty sites in fifteen days, and what that did to the writing","k":"article","x":"One site became twenty repositories in fifteen days because each argument needed its own version history, and the index now starts from a question.","h":[],"n":4,"d":"2026-08-26","tp":["site-and-engineering","vaults-and-method"],"tg":["network","publishing","method"],"g":"A body of work outgrew the site it was published in and split into nineteen siblings so each argument could keep its own version history, and the index into them starts from the reader's question rather than a list of names. · Twenty repositories in fifteen days · A section is a promise · Its own version history · Discovery got worse · Consistency by discipline · Cross-link drift · Start from the question · Every thesis in the site's own words · Same generator, same release script · Publishing is adding one file · skills.sgit.ai, not published yet · Arguments, not products","i":78},{"u":"articles/no-server-by-design.html","t":"No server, by design: first I retired the database, now I am retiring the back end","k":"article","x":"First the database went, now the back end: the SGit Newsroom runs all its logic in the browser, on commodity servers that cannot read your data.","h":[["files","Step one: the file system is the database"],["backend","Step two: the back end goes too"],["newsroom","What the newsroom is now"],["commodity","Not \"no servers\": commodity servers"],["scale","It scales without the complexity scaling"],["security","Security as a business tool"],["openness","Why it is easy to throw technology away"],["speed","It ships fast"],["principles","Sovereignty, privacy, explainability, determinism"],["push","How far can it be pushed?"],["limits","What this does not claim"],["next","What comes next"],["sources","Where this comes from"]],"n":72,"d":"2026-10-11","tp":["site-and-engineering","vaults-and-method"],"tg":["architecture","client-side","serverless","no-database","lets","file-system","vaults","append-lanes","privacy","sovereignty","security","newsroom","sg-meter","wardley","article"],"g":"After replacing the database with the file system (LETS), the next step is removing the back end: encrypted vaults make the server blind, so the application moves to the browser, leaving only commodity products on the server side; the security gaps this leaves are written down and accepted because they do not match the threat, and the result scales without its complexity scaling and ships as files. · The file system is the database · No back end · A blind server · Commodity servers, not no servers · The SGit Newsroom · Gaps written down and accepted · Scales without complexity scaling · A release is a set of files · Sovereignty, privacy, explainability, determinism · How far can it be pushed?","i":79},{"u":"articles/not-a-straight-line.html","t":"Not a straight line: working for RiskMandate by working on the newsroom, and why the thinking is never wasted","k":"article","x":"Working on the newsroom solved RiskMandate's oldest problem in two days: progress is not a straight line, and the thinking is never wasted.","h":[["in-short","In short"],["better","Am I getting better?"],["vault","Six months on one problem, then two days"],["momentum","Why the newsroom could solve it and RiskMandate could not"],["workflow","The newsroom is RiskMandate's workflow, and its example"],["history","The line that runs back"],["back","Going back, with something that works"],["thinking","The thinking is never wasted"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":95,"d":"2026-10-12","tp":["startups-and-strategy","site-and-engineering"],"tg":["strategy","focus","leverage","startups","newsroom","riskmandate","open-source","feedback-loop","publishing","vaults","passkeys","agent-behaviour-policy","article"],"g":"Focus on one thing only assumes progress is a straight line; focus instead means undivided attention on the place of highest leverage, the one with users and feedback loops; the newsroom had the momentum to solve the master vault problem that had resisted six months of RiskMandate attempts, solving it in two days with a passkey and a door per path; it is also building the agentic workflows and the public behaviour-policy example RiskMandate needs; and its history, The Cyber Boardroom and MyFeeds, both paused, shows the thinking is never wasted when it is written down and open source. · Progress is not a straight line · Undivided attention on the place of highest leverage · Am I getting better? · Six months on the master vault, then two days · Do not solve a problem without the momentum to need it · The newsroom is RiskMandate's workflow · A public example customer work cannot be · The Cyber Boardroom · MyFeeds · The writing stays · Going back with an end-to-end example · The thinking is never wasted","i":80},{"u":"articles/one-article-five-readers.html","t":"One article, five readers: a librarian, a cartographer, a historian, an explainer and a storyteller read the same piece","k":"article","x":"Write the article first, then send five agent readers through it: a catalogue, an ontology and maps, the arc, two minutes, and a deck.","h":[],"n":58,"d":"2026-10-10","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["newsroom","journalism","liquid-content","fractal-semantic-graphs","ontology","agents","librarian","cartographer","historian","multi-view","article"],"g":"Long, evidence-rich articles are hard to consume, and the answer is not shorter articles but extraction after the writing: the prose stays the source because writing is how the argument is found, and five agent readers with defined roles then produce other views of it, two building the platform (a Librarian that catalogues every item with its verbatim anchor, a Cartographer that builds an ontology at article, topic and site altitude and draws the maps) and three writing for a person (a Historian on what the article added and where it sits, an Explainer in two minutes, a Storyteller as a deck); run on five articles it catalogued 662 items, flagged 32 problems in published text, merged ideas named differently across articles, and measured each article's contribution as new concepts falling 17, 11, 5, 11, 1. · Deep articles are hard to consume · The prose comes first · Five readers · Platform readers build the graph · Context readers write for a person · Every item anchored verbatim · Ontology at three altitudes · 32 problems flagged · Same idea, different names · New concepts: 17, 11, 5, 11, 1 · New to the set is not new to the site · Read it another way · The Article Views vault · Image prompts not yet run · Feedback written, not applied","i":81},{"u":"articles/open-source-is-not-free.html","t":"Open source is not free: who pays to keep the long tail working?","k":"article","x":"An old iMac, the long tail of old versions that projects are not paid to support, measured from public data, and five ways to pay for it, simulated.","h":[["imac","The iMac"],["measured","The long tail, measured"],["money","Where the money flows"],["charge","Why not just charge?"],["pence","The pence problem"],["models","Five ways to pay for the long tail, simulated"],["break-even","What price pays for the long tail?"],["sensitivity","What matters most"],["satellites","The satellites"],["says","What this says"],["check","Check my numbers"],["sources","Where this comes from"]],"n":59,"d":"2026-10-10","tp":["startups-and-strategy","vaults-and-method"],"tg":["open-source","economics","funding","long-tail","support","micropayments","simulation","data-science","homebrew","docker","macos","article"],"g":"Open source is not free: somebody pays, at least in time. The long tail of old versions and old machines is measurable (6.99% of Homebrew macOS install events in 30 days, 16.66% in a year, are on versions older than the newest three) and projects drop it because it has no revenue stream, while vendors such as Oracle and Red Hat turn the same need into large, high-margin support businesses. Pence payments only work aggregated, because a 20p card fee dominates below about five pounds. Simulated over 10,000 draws for one project at Homebrew's scale, micro-paid supported builds cover the long-tail cost in half the draws at about 1.80 pounds a machine a year with typical conversion but never in nine in ten, donations in about a quarter, and customised builds for companies in 94%: the reliable money is companies paying for a build below their own cost of keeping the tool working, sold as value, not as a donation or a weaker free tier. · One old iMac · Open source is not free · The long tail, measured · Where money flows · Charging has costs · The pence problem · Five models, simulated · Break-even near two pounds · Customised builds for companies · The satellites · Long Tail Ledger vault","i":82},{"u":"articles/pay-after-you-read.html","t":"Pay after you read: how a reading meter became a working business model in one afternoon, one release at a time","k":"article","x":"Seven releases in one afternoon turned the reading meter into a working model: pay after you read, and the rating sets the price.","h":[["features","What a reader gets now"],["after","Pay after you see the value"],["reader-first","Every decision started with the reader"],["how","How it grew: one release at a time"],["ilc","The zigzag: innovate, leverage, commoditise"],["fast","Why it could go this fast"],["pieces","Do we have all the pieces?"],["others","Want to try it on your site?"],["sources","Where this comes from"]],"n":65,"d":"2026-10-10","tp":["news-and-evidence","startups-and-strategy"],"tg":["newsroom","reading-meter","micropayments","pricing","personalisation","personas","wardley-maps","innovation","agents","article"],"g":"On 10 October the reading meter became a working pay-on-demand publication in seven releases in 3 hours 40 minutes: pay for the share read from a balance that can go below zero, rate usefulness after reading to set the price from free to double, personas and newsrooms kept in the reader's browser, and consented, encrypted sharing for a designed front page. Paying after the value inverts Leanpub's pay-before model and is supported, with caveats, by pay-what-you-want evidence; Blendle's refunds show ratings measure quality. Each feature grew from using the last, which a Wardley map shows: new components on the left, built on commodities already there. It was fast because of non-functional work done before. Every decision started from not ripping off the reader. What is missing is traffic. · Pay by depth, below zero · Usefulness sets the price · Personas and newsrooms · Share your reading · Pay after the value · Pay what you want evidence · Do not rip off the reader · One release at a time · The more we have, the faster we go · Test it on a site with traffic","i":83},{"u":"articles/pay-to-keep-your-persona.html","t":"Pay to keep your persona: readers should pay because it helps them, not because they feel they should","k":"article","x":"Readers should pay because it helps them: a persona with a name and a graph, several for focus, and one that follows you between devices.","h":[],"n":62,"d":"2026-10-10","tp":["startups-and-strategy","news-and-evidence"],"tg":["newsroom","personalisation","personas","value-proposition","micropayments","local-first","privacy","focus","article"],"g":"Paying for content because you have to, feel you should, or think it right works to a degree but does not scale and makes readers feel the wrong thing; what readers would pay for is what helps them, time, context, focus and good decisions, carried by a persona built from their reading, named, drawn as a graph, curated, and multiplied for focus, and the paid product is that persona following them between devices and agents with the privacy intact. · Paying because you have to, or feel you should · £12 to avoid the supporter messages · 8 million donors, $10.58 on average · Pay for time, context, focus and good decisions · The persona · Personas are focus · Five starting personas · Two devices, two strangers · A persona that follows you","i":84},{"u":"articles/places-of-highest-leverage.html","t":"Places of highest leverage: where to focus when the agents can go in every direction, told with three days of real steps","k":"article","x":"Where to focus when agents can go in every direction: the place of highest leverage, shown on three days of real steps, two walked back and one parked.","h":[["short","In short"],["rule","The rule"],["steps","Three days, step by step"],["back","Walking back, and parking"],["map","Which shoulders, on the map"],["ship","Shipped, and used"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":89,"d":"2026-10-12","tp":["startups-and-strategy","agents-and-policy","site-and-engineering"],"tg":["startups","focus","leverage","wardley-maps","explorer-villager","shipping","constraints","agents","newsroom","feedback-loop","strategy","article"],"g":"When agents can build in any direction the scarce thing is direction, and the rule is to work at the place of highest leverage, where the most value can be added now, nothing blocks, and you can start without knowing how the whole problem ends because each shipped step produces better tooling and a better view; the last three days show it, a meter to pay by depth to personas to sharing to a persona link to a reading list to a passkey design walked back and forward by a lab to real vaults and a vault pane, each shipped and used before the next was chosen, with a branch parked as valuable only with many readers; on a Wardley map the passkey work became a product in a day because its shoulders are mature commodities and the one custom layer sits where a provider is not; walking back and parking are not failures; constraints, saying no and shipping to a user who can be me; and the brief that could not be written until the art of the possible was known. · Direction is the scarce thing · Places of highest leverage · Start without knowing how it ends · Guess, measure, narrow · Constraints and saying no · Three days, step by step · Walking back is a two-way door · Parking is not failure · The path only looks obvious backwards · Which shoulders, on the map · Shipped, and used, even by me · The brief depends on the possible","i":85},{"u":"articles/price-it-then-give-it-away.html","t":"Price it, then give it away: the early access programme as the next step after \"do they miss it\"","k":"article","x":"Define the product, price it, deliver it at a cost that grows a step at a time, then offer it free to people who know you and measure what it costs them.","h":[],"n":21,"d":"2026-10-02","tp":["startups-and-strategy","agents-and-policy"],"tg":["startups","strategy","pricing","early-access","riskmandate","agent-behaviour-policy","go-to-market","agents","article"],"g":"A price is a statement of what you think the thing is worth, and giving it away at that price to people who already know you is the cleanest test of whether the statement is true, provided you measure and cut what the free offer costs them. · Do they miss it · Define the product · Price it before you give it away · The four-tier ladder · Cost that grows a step at a time · Offer it to the people who already know you · Free is never free for the other side · Do not make them apply · RiskMandate early access programme · The answer is brutal either way · What we count · For the agents running this · The site is becoming the memory","i":86},{"u":"articles/proof-behind-the-claim.html","t":"The proof is two clicks behind the claim, what the homepage gets wrong, and the fix","k":"article","x":"The homepage leads with encryption, which cannot be seen, while twenty-five vaults a stranger can open in one click sit two clicks away in a table.","h":[],"n":6,"d":"2026-09-07","tp":["site-and-engineering","vaults-and-method"],"tg":["homepage","positioning","vaults","agents"],"g":"Encryption is a property you cannot look at, so the homepage should put the real vaults a visitor can open before the mechanism, and should not carry a claim that no published vault demonstrates. · Twenty-five published vaults · Encryption is a property you cannot look at · The terminal walkthrough · Use cases as categories · The proof is a table, two clicks away · Opened by one string · Agents build for each other · The briefs page as a log · The claim with no artefact · Reorder so proof comes before mechanism · Generated from the same data · A page that says who is behind it · The site card","i":87},{"u":"articles/proof-moved-up.html","t":"The proof moved up, the homepage after the rebuild, next to the before pictures","k":"article","x":"The rebuilt homepage puts four real vaults under one sentence, picks six by the job they do, and computes its numbers at build time from the site's own data.","h":[],"n":7,"d":"2026-09-07","tp":["site-and-engineering","vaults-and-method"],"tg":["homepage","positioning","vaults","agents"],"g":"The homepage stopped being a page somebody edits and became a view over the site's own data, with real vaults placed before the mechanism and the one unproven claim left unpretended. · The reframed sentence · Four real vaults in the hero · What people actually ship · A vault as a unit · One human, a team of agents · The numbers are not typed · What was cut · Nine bands before, nine after · What it cannot show yet · admin/content/vaults.json · The homepage as a view over data · Beside the before pictures","i":88},{"u":"articles/re-anchoring-agent-behaviour-policies.html","t":"Re-anchoring: keeping an agent's rules alive through every summary, and a canary that shows when they are not","k":"article","x":"Summaries keep under 2% of a long session. Re-anchoring prints the agent's rules back after each one; a canary report shows it is working.","h":[["view-catalogue","In short"],["view-catalogue","What summarisation does to a session"],["view-catalogue","Three ways a rule fails"],["view-catalogue","Re-anchoring: the rules come back from disk"],["view-catalogue","One file, three jobs"],["view-catalogue","The canary: a report that ends every few answers"],["view-catalogue","Living with it"],["view-catalogue","The recipe changes with the platform"],["view-catalogue","What it does not do"],["view-catalogue","Build your own"]],"n":56,"d":"2026-10-10","tp":["agents-and-policy","site-and-engineering"],"tg":["agents","claude-code","hooks","agent-behaviour-policy","compaction","context","re-anchoring","guardrails","governance","article"],"g":"Instructions given only in conversation can be lost when a long session is summarised, and this site's own session was summarised eighteen times in a month, keeping about 1.7% each time; re-anchoring keeps the agent's rules in an Agent Behaviour Policy file and has a SessionStart hook on the compact matcher print it back after every summary, so rules are restored rather than remembered, while a canary status report computed from the transcript, checked by a Stop hook and expected by the person, shows the policy is still being read; the same file serves as memory, checker rulebook and enforcement list, and the best recipe for each rule changes with every platform release, which is the work worth curating. · Instructions lost at compaction · Eighteen summaries, 1.7% kept · Three ways a rule fails · Re-anchoring · One file, three jobs · The canary report · The brown M&M · A rule measured · Recipes that age · A second reader and a gate","i":89},{"u":"articles/replicating-the-agentic-inbox.html","t":"Replicating the agentic inbox: a walkthrough from one Claude session to a team of agents that never press send","k":"article","x":"How to copy a working agentic email setup in phases: a mailbox and Claude seat of the agent's own, one session with a policy, then roles talking in files.","h":[],"n":22,"d":"2026-10-02","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","email","inbox","agent-behaviour-policy","riskmandate","claude","google-workspace","connectors","vaults","email-fs","walkthrough","article"],"g":"Claude is used as an agent state machine, one session per role, every message between agents a file in a vault and every outgoing email a draft a person sends; the phases take you from one session to a scheduled team without ever breaking that rule. · Claude as a state machine · Accounts of the agent's own · Connectors enabled, then connected · The inbox role · The policy as a table · Roles: @inbox, @drafts, @crm, @briefs, @dev · Email-FS lite · The agent drafts, a person sends · The security hold · The conductor · The clone cost · A leaked write key means a new vault · Interfaces for the moment · The record · Trust built from the record · The agents' own account","i":90},{"u":"articles/rfc-0001-public-key-cryptography-for-sgit.html","t":"RFC 0001: two ways to add public-key cryptography to sgit, and the questions we want you to answer","k":"article","x":"A Request for Comments: two key pairs so a reader cannot write, sealed files only named people can open, and fourteen questions for reviewers.","h":[["providers","Where the private key could live today"],["providers","How the pieces fit"],["providers","Prior art"],["questions","Questions for reviewers"],["questions","How to comment"]],"n":53,"d":"2026-10-09","tp":["vaults-and-method","site-and-engineering"],"tg":["rfc","sgit","pki","public-key-cryptography","encryption","signatures","ed25519","x25519","hpke","age","sealed-files","key-management","threat-model","request-for-comments","article"],"g":"An sgit vault's read key is symmetric, so anyone who holds it and can write to the store can author history clients accept; turning the read key into a public key does not fix that, but two key pairs do, giving a native PKI vault mode where readers cannot write, hosts cannot forge and write-only depositors become possible, while an optional sealed-files layer encrypts chosen files to named people's public keys with the private key held in a file, an ssh key, hardware or a remote service that logs and approves each decryption; both share a first step, a writer signature on today's vaults, and nothing is built until reviewers have answered fourteen questions. · One symmetric read key · Whoever can read can forge · One key pair is not enough · A writer signature first · Native PKI vault mode · The write-only depositor · Sealed files · Where the private key lives · Writers can replace blindly · Known shapes · Fourteen questions","i":91},{"u":"articles/riskmandate-ten-questions.html","t":"Ten hard questions for RiskMandate, answered: the mandate, the reach, the gap, and what we are deliberately not","k":"article","x":"Ten hard questions from a conference, answered in a two-hour interview: what RiskMandate does, what it deliberately is not, and how mature each part is.","h":[],"n":51,"d":"2026-10-09","tp":["agents-and-policy","startups-and-strategy"],"tg":["riskmandate","agent-behaviour-policy","mandate","grant","delta","barriers","insurance","liability","enforcement","supply-chain","pki","digital-twins","metrics","positioning","interview","article"],"g":"Every answer rests on four objects for one agent in one deployment, mandate, grant, delta and barriers, and a strict division of roles: RiskMandate authors the policy and pre-commits the verdict, the customer's controls enforce it, a named executive owns the residual risk, and logs calibrate the grant; so RiskMandate is not inline, not an insurer and not one button, the value lands before the action, new capability can only land as excess, and the headline metric is the reduction in risk the business has to accept. · Mandate, grant, delta, barriers · Author, decide, enforce · The 5k order was possible on day one · The mandate is locked business logic · Reality is the calibrator · Open source, maintained per customer · Authorisation is the grant · We make agents insurable · The supply chain is a bigger agent · Containment now, reversal next · Authorisation by encryption · What is the headline metric? · Three fixes on our own site","i":92},{"u":"articles/saas-apocalypse-decided-by-inertia-not-by-ai.html","t":"The SaaS apocalypse will be decided by inertia, not by AI","k":"article","x":"Incumbents and newcomers have the same AI; what separates them is how much past success each has to protect, so inertia decides which SaaS companies survive.","h":[],"n":9,"d":"2026-09-21","tp":["startups-and-strategy"],"tg":["saas","strategy","wardley-maps","agents","article"],"g":"AI is available to both sides, so the SaaS apocalypse will be decided by inertia: where each company sits on the evolution axis and how much past success it has to protect. · The claim, stated so it can be wrong · AI is available to both sides · Success breeds inertia · Nokia, twice · The SaaSpocalypse of February 2026 · The database is the moat · Users were never happy · The persistence of Excel · The SaaS product slid right · Vibe coding is writing good briefs · Why incumbents cannot simply do this · The irony · The brief is not a product · Village and town-planner companies · More engineers, not fewer","i":93},{"u":"articles/send-an-agent-not-a-spreadsheet.html","t":"Send an agent, not a spreadsheet: the next generation of software due diligence, and why the companies that stopped reading their code are about to be asked about it","k":"article","x":"Due diligence never scaled because it was a form; a buyer can now send an agent into a vendor's environment and read what the code and the practices are.","h":[],"n":28,"d":"2026-10-05","tp":["agents-and-policy","startups-and-strategy"],"tg":["due-diligence","code-review","agent-behaviour-policies","blast-radius","threat-models","sbom","cyber-resilience-act","product-liability","startups","ai-generated-code","fractal-semantic-graphs","article"],"g":"Software due diligence was a questionnaire answered by the vendor and disconnected from the code; a buyer can now send a prompt or a small agent under a behaviour policy to run inside the vendor's environment and derive a graph of what reaches production unreviewed, what is documented and threat modelled, what the bugs touched and which agents act under what policy, with the vendor redacting but not rewriting; the test is risk-based, the behaviour policy is the due diligence document for the agents, the companies that stopped reading their code are about to meet the consequence they have not had, a startup should double down on understandability, and the code review company is better sold to buyers than to developers. · Due diligence was a form · Companies will not lie on a record · Send an agent, not a spreadsheet · Derived, not composed · Questions where silence is an answer · A threat model is the fastest read of a team · Two kinds of bug · Risk-based, not pure · The behaviour policy is the due diligence document · The consequence that was missing · Most companies already do not understand their software · The startup's two advantages · Duties arriving before visibility · Sell it to the buyers · What is open","i":94},{"u":"articles/seven-vaults-one-method.html","t":"Seven vaults, one method","k":"article","x":"Publishing seven encrypted vaults in a fortnight produced a method, and every rule in it exists because something went wrong first.","h":[],"n":2,"d":"2026-08-19","tp":["vaults-and-method"],"tg":["vaults","publishing","method"],"g":"A repeatable method for publishing encrypted vaults, each rule learned by getting something wrong first, starting with classifying the credential before it touches anything. · Classify the credential · Vault key · Read key · check_credential.py · Release validator tripwire · Derive rather than refuse · Strictly less capability · Audit with the key you publish · Sealed LLM credential check · Risk Graph Explorer · Capture evidence by driving the real product · Describe, show, then admit · Publishing playbook · The honest ledger","i":95},{"u":"articles/six-agents-one-inbox.html","t":"Six agents, one inbox: what a real multi-agent setup taught me about access policies","k":"article","x":"An access policy for an agent is only as real as its worst row: every rule in a real six-agent setup, graded by how it is enforced today.","h":[],"n":17,"d":"2026-09-29","tp":["agents-and-policy"],"tg":["agents","connectors","access-policies","non-human-identity","gmail","github","riskmandate","security","article"],"g":"Write an agent's access policy as a table with a column for how each rule is enforced, and the policy is exactly as strong as the row that says the agent has been told. · The worst row · The account is the blast radius · Dedicated accounts per agent · Organisational unit · Restrict delivery · Attachment compliance · Scope limits on third-party apps · Six roles · The exception arrived before the policy · Signature registry · The attachment finding · The documentation disagrees with itself · Connector twin · Enforcement ladder · The policy table · Cowork and Claude Code reach","i":96},{"u":"articles/sovereignty-is-the-ability-to-move.html","t":"Sovereignty is the ability to move: play where the ball is going, take away the dot, and keep a twin that works","k":"article","x":"Sovereignty is the ability to move: take away each dot on the map, and keep a twin that works for the minimum viable company, tested in QA.","h":[["in-short","In short"],["ball","Where the ball is going"],["move","Sovereignty is the ability to move"],["dot","Take away the dot"],["zero","Do not jump from zero to a hundred"],["qa","Run it in QA, all the time"],["possible","Can it move, then is it cheap"],["open","Open source, or one acquisition away"],["central","Provided centrally, like a road"],["now","Why now"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":94,"d":"2026-10-12","tp":["startups-and-strategy","site-and-engineering"],"tg":["sovereignty","resilience","europe","wardley-maps","twins","open-source","minimum-viable-company","qa","testing","disaster-recovery","strategy","article"],"g":"Europe's sovereignty conversation is mostly about frontier models, data centres and GPUs, which is playing where the ball is rather than where it is going; sovereignty is the ability to move workloads and actions, which most organisations lack because everything is hard-coded and never deployed twice; the test is a Wardley map where each dot is taken away and the blast radius followed up to the customer, across the whole stack; the answer is not an alternative as good as today but a twin that works for the minimum viable company, run in QA all the time, scored on a maturity model; ask whether it can move before asking whether it is cheap; only open source gives control that survives an acquisition; Europe lacks a commercial open source model, not startups or talent; some of this should be provided centrally like a road; and models and twins now make the analysis possible at scale. · The conversation is about things Europe does not control · Play where the ball is going · Sovereignty is the ability to move · Critical means the business stops · Take away the dot · Every dot must be replaceable, built or bought · Do not jump from zero to a hundred · The minimum viable company · A twin for every dot you do not control · Can it move: a maturity model · Run it in QA, all the time · Can it move, then is it cheap · Open source, or one acquisition away · Europe lacks a commercial open source model, not startups · Provided centrally, like a road · Models and twins make the analysis sca","i":97},{"u":"articles/stop-the-bleeding-then-ask-for-it-back.html","t":"Stop the bleeding, then ask for it back: a helper, an agent and a friendly letter for the subscriptions you keep and do not use","k":"article","x":"A helper and an agent ask what you used, ask nicely for a pause and a refund, stop the payments either way, and publish what each company did.","h":[["in-short","In short"],["keep","The subscriptions we keep"],["ask","Step one: ask them what we used, and work it out ourselves"],["request","Step two: a friendly request for a pause and a refund"],["stop","Stop the bleeding"],["law","What the law does, and what it does not"],["access","If nobody answers: the right of access"],["register","An open register, under Creative Commons"],["helper","A service for helpers"],["agents","Done by agents, under a policy"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":98,"d":"2026-10-13","tp":["startups-and-strategy","agents-and-policy"],"tg":["subscriptions","refunds","consumer-rights","freelancers","agents","agent-behaviour-policy","right-of-access","open-register","creative-commons","business-model","article"],"g":"A service in which a freelancer, with agents doing everything except typing secrets and sending drafts, asks a company for its record of your usage while doing its own analysis, sends a friendly request for a suspend mode and a refund, stops the payments either way, and publishes what each company did in an open register, run from vaults and paid at 10% of three months of savings and 20% of what is recovered. · Subscriptions we keep · Ask what we used · Our own analysis · A friendly request · Suspend mode · Stop the bleeding · The law is a floor · The right of access · The open register · The helper's fee · Agents under a policy · Claude in Chrome · Two controls · Vaults as the interface · A first MVP","i":98},{"u":"articles/story-vault-meets-reader-skills.html","t":"Story vault underneath, Reader Skills on top: why local journalism has the most to gain","k":"article","x":"Markus Franz's Reader Skills on top, the story vault underneath: each skill is a graph query, and local stories can pay back down their chain of sources.","h":[],"n":40,"d":"2026-10-07","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["journalism","local-news","liquid-utility","reader-skills","story-vault","fractal-semantic-graphs","provenance","micropayments","verification","article"],"g":"Markus Franz's Liquid Utility proposes Reader Skills that help people understand, update, relate, compare, follow and act on journalism; the story vault keeps reporting as a versioned graph of claims tied to hashed evidence; put together, each skill is an operation on the graph and each safeguard he asks for is a property the graph already has, his bridge closure becomes a graph in which one supersede edge is the update, the alert and the correction, and his emphasis on locality supplies what the vault's economics had not mapped: local contributors and local journalists sit where trust is strongest, stories travel up to national and international outlets, and if every use pays back down the chain of claims it rests on, small payments from many people fund the reporting nearest to them. · Liquid Utility · Six Reader Skills · The story vault · What we know, and what we can help someone do · Each skill is an operation on the graph · His safeguards are graph properties · The bridge closure as a graph · Local is where the trust is · The local contributor is a node · Every use pays back down the chain · Is this still current? · Trust, use, payment, reporting, evidence, trust · One topic, one vault, one service, one ledger","i":99},{"u":"articles/supply-chain-of-vaults.html","t":"A supply chain of vaults: how GenAI, open data and small custom tools could bring the price of food down","k":"article","x":"The food chain is a data problem: one encrypted vault per party, joined by append lanes and a typed graph, could cut the waste that keeps prices high.","h":[],"n":14,"d":"2026-09-27","tp":["vaults-and-method","startups-and-strategy"],"tg":["supply-chain","food","genai","vaults","fractal-semantic-graphs","open-source","creative-commons","deflation","article"],"g":"The food chain is logistics, logistics is a data problem, and a supply chain of encrypted vaults with small custom tools built once by GenAI could bring the price of food down, if somebody runs the experiment. · The food pound · A chain of spreadsheets · Food lost before retail · The chokepoint · The limit of regulation · All of it is logistics · A supply chain of vaults · Append lanes · One graph across the vaults · Use GenAI not to use GenAI · Villagers and town planners · The price hypothesis · The deflation dogma · The cost of not sharing · Open-weight models · The experiment","i":100},{"u":"articles/the-agent-is-the-reader.html","t":"The agent is the reader: why agents will pay for graphs, personas and signed claims, because it is cheaper than not paying","k":"article","x":"Agents pay to read the raw web and guess its meaning. A curated graph answers for 90% fewer tokens, so paying the publisher is cheaper than not paying.","h":[["layers","What five articles became"],["loop","The readers check each other"],["compress","Does it compress as it grows?"],["meaning","Meaning, not just compression"],["personas","Agents are personas too"],["economics","The agent pays not to pay"],["passthrough","The price passes through"],["claims","Signed claims, with an expiry date"],["wikipedia","What would Wikipedia do?"],["limits","What this does not show"],["next","What comes next"],["sources","Where this comes from"]],"n":71,"d":"2026-10-11","tp":["news-and-evidence","startups-and-strategy"],"tg":["newsroom","economics","agents","micropayments","fractal-semantic-graphs","compression","personas","tokens","provenance","signed-claims","librarian","cartographer","article"],"g":"When agents are the readers, a publisher that serves curated meaning, an anchored graph fitted to the agent's persona and objective, and claims signed with an expiry saves the agent more tokens than it charges, so the agent pays not to pay; measured on five articles the graph answers in 90% fewer tokens, and on 70 articles compression with scale appears only where a Cartographer curates, which is the work worth paying for. · The graph is bigger than the text · A question costs 90% less in the graph · Making the map is the verification · Compression with scale needs curation · Projected to 100,000 articles · PageRank with meaning · Agents are personas too · The agent pays not to pay · Signed claims with an expiry · Will agents pay?","i":101},{"u":"articles/the-agent-team-as-it-runs.html","t":"The agent team as it runs: one person, twelve agents, encrypted vaults, and a mailbox nobody sends from","k":"article","x":"Twelve agents on dedicated accounts, encrypted vaults as the only memory, messages as files, a folder per person, and a mailbox nobody sends from.","h":[],"n":32,"d":"2026-10-06","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","agent-behaviour-policy","riskmandate","vaults","sgit","email-fs","crm","provenance","encryption","privacy","data-classes","fractal-semantic-graphs","claude","google-workspace","article"],"g":"A running team of twelve Claude agents is described as a system and as a set of properties: each agent a role on a dedicated account with a behaviour policy; encrypted vaults the host cannot read as the only memory; messages between agents as email files in a mailroom; a CRM of one folder per person with provenance and hashes; a conductor with a security role first and last; drafts only, a person sends; identity and encryption as the strong barriers and policy as the weak one, said plainly; three classes of information, public, private-ish and personal, with the vaults built to refuse the third and rules scoped per customer or written for the recipient; each piece mapped to the idea on the site that argued for it; and the open question of who gives the mandate over information. · One line: phone, sessions, vaults, drafts, people · The account is the blast radius · Twelve agents, one focus each · Email, but as files · The conductor and the session steps · Memory is one folder per person · The rules never negotiated · Barriers typed honestly · The security properties, as properties · Public, private-ish, personal · Rules per customer, rules for the recipient · Files, graphs and a state machine · Every piece has an argument on this site · What is not working yet · Who gives the mandate over information?","i":102},{"u":"articles/the-ai-governance-stack-as-a-graph.html","t":"The AI governance stack, as a graph: an answer to Hari Kota, built","k":"article","x":"Hari Kota's ten-layer AI governance stack, kept as posted and turned into a fractal semantic graph: the gaps become missing edges, and queries.","h":[["hari","Hari's stack, exactly as posted"],["hari","The table already contains a graph"],["hari","Each layer is its own world"],["hari","The edges between worlds come from the provisions"],["hari","Zoom from one cell down"],["hari","The stack is a snapshot; the graph keeps time"],["today","Do this today, as a traversal"],["today","Hari's three gaps, as queries"],["today","What \"4 or 5\" means in a graph"],["today","Bring your own"],["today","How the vault is built, and what it does not claim"],["today","Prior art"],["today","Open it"]],"n":54,"d":"2026-10-09","tp":["graphs-and-knowledge","agents-and-policy"],"tg":["ai-governance","fractal-semantic-graphs","graphs","eu-ai-act","iso-42001","nist-ai-rmf","risk","inventory","competence","vaults","article"],"g":"Keep Hari Kota's ten-layer stack exactly as posted as the view and make a fractal semantic graph the model: the table already contains ten edges across layers where rows share an instrument, role or jurisdiction; each layer is its own world with its own ontology, sharing only a grammar of verbs with inverses; the edges between worlds come from the provisions themselves, such as EU AI Act Article 26(2) and NIST AI RMF GOVERN 1.6; changes over time are dated edges rather than rewrites; and a gap is a missing edge, so the do-this-today test and the gaps in layers 3, 5 and 10 become queries anyone can run. · Hari's stack, as posted · The table already contains a graph · Each layer is its own world · Five grammar rules · Edges from provisions · Zoom from one cell down · The graph keeps time · Do this today, as a traversal · Gaps as queries · Hollow Oak Home · What 4 or 5 means · AI Governance Graph vault","i":103},{"u":"articles/the-article-remains-a-product.html","t":"The article remains a product: Jonathan Palmer's four outputs for a newsroom with no human readers, checked against one that runs","k":"article","x":"Jonathan Palmer's four outputs for a newsroom with no human readers, quoted line by line and checked against newsroom.sgit.ai as it runs, screenshots included.","h":[["short","In short"],["scenario","Model that exact scenario"],["structured","Structured intelligence"],["continuous","Continuous intelligence"],["provenance","Provenance and verification"],["embedded","Embedded distribution"],["pay","Who would pay?"],["article","The article remains a product"],["cultural","Cultural, not technical"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":79,"d":"2026-10-11","tp":["news-and-evidence","graphs-and-knowledge","agents-and-policy"],"tg":["newsroom","future-of-news","structured-intelligence","provenance","graphs","versions","agents","llms-txt","vaults","world-news-day","economics","response","article"],"g":"Palmer's four outputs, structured intelligence, continuous intelligence, provenance and verification, and embedded distribution, are what a newsroom built on the story-as-graph thesis already produces: graphs as JSON beside all 74 articles, version histories read from git, a hashed corpus in a vault with a published read key, and an llms.txt and markdown twin for agents; what is designed and not running, selling the graph, propagating corrections and paying the fact creator, is said plainly; and where the two differ is the word proprietary, because provenance only someone else can check is worth charging for. · No human readers tomorrow · Model that exact scenario · Structured intelligence · Continuous intelligence · Provenance and verification · Embedded distribution · The agent pays not to pay · Published, not proprietary · The article is a projection · What runs, what is designed · Cultural, not technical · World News Day corpus · Five readers","i":104},{"u":"articles/the-behaviour-policy-is-the-business-logic.html","t":"Zoom into an agent's behaviour policy and you find the business logic","k":"article","x":"Below the first rules, an agent's behaviour policy is the business: functions, processes, clients, values. Layered, owned, counted, and where vendors plug in.","h":[["view-catalogue","The bottom is mechanics, and vendors do it well"],["view-catalogue","Above the mechanics is the business"],["view-catalogue","Why the vendors stop at the mechanics"],["view-catalogue","Software was the law"],["view-catalogue","Policies inside policies"],["view-catalogue","How much of it is hope?"],["view-catalogue","Where the vendors plug in"],["view-catalogue","It looks a lot like a skill"],["view-catalogue","What comes next"]],"n":38,"d":"2026-10-07","tp":["agents-and-policy"],"tg":["agents","agent-behaviour-policy","business-logic","fractal-semantic-graphs","controls","risk-acceptance","rbac","skills","integrations","article"],"g":"The first rules written for an agent are mechanical and vendors enforce them well, but every layer above them is the company's own business logic: how a function is done, the steps a process goes through, who a client is right now, and what the company is for; much of that logic was never written down because software enforced it by omission, and agents working through APIs bypass that; a behaviour policy built in layers, each refining the one below with its own owner, can hold it, can count how much of it is backed by a control, by an accepted risk or by hope, and gives vendors a written rule to enforce and a number to move, which is what lets the business hand work to agents and scale. · The bottom layers are mechanics · send_message ran with no prompt · Above the mechanics is the business · Vendors cannot model each customer's business · Software was the law · Agents work through the API, not the screen · Policies inside policies · Shared facts, own formulas, declared bridges · Control, accepted risk, or hope · Five, six and six of seventeen · A vendor control moves one written rule · Frameworks already ask for the description · The upper layers read like skills · A real inbox, layered and counted","i":105},{"u":"articles/the-bridge-followed-to-the-end.html","t":"The bridge, followed to the end: what one local story is worth when it is kept as a graph","k":"article","x":"A bridge closure simulated on a story vault: newsworthy on 5 days, needed on 57, used by three readers, four buyers and an agent, and paid back to its sources.","h":[],"n":41,"d":"2026-10-07","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["journalism","local-news","reader-skills","liquid-utility","story-vault","fractal-semantic-graphs","micropayments","agents","simulation","article"],"g":"Played out as a simulation on a story vault, one local bridge closure shows the value of journalism as getting a disputed fact right and keeping it right: newsworthy on five days but needed on fifty-seven, used through Reader Skills by readers whose own graphs meet the story graph at shared anchors and who decide better with it, bought by institutions and agents who need the same claims, cheaper for agents than searching, and paid for in small amounts that walk back down the claims to the reporter, the paper and the residents whose evidence the graph names. · A bridge closes; the council says one week · The journalism is getting the date right · Each correction supersedes the last · Newsworthy on five days, needed on fifty-seven · Each reader's graph meets the story · A parent, a café owner, a plumber · Better decisions, counted · Without it: stale pages, rumours, nobody accountable · Buyers who do not read the paper · Agents spend less and are right · £2,707, walked back to the sources · The simulation as a vault","i":106},{"u":"articles/the-deck-i-could-not-download.html","t":"The deck I could not download: an author-first home for presentations, as a business plan somebody else can build","k":"article","x":"One download offered as a subscription, an author paid nothing, and a design for the service the author would have chosen: vaults, keys, seven roles, 85% to the author.","h":[],"n":33,"d":"2026-10-06","tp":["startups-and-strategy","vaults-and-method"],"tg":["slideshare","content-creators","business-plans","vaults","sgit","encryption","pki","micropayments","provenance","gdpr","right-of-access","copyright","wardley-maps","inertia","article"],"g":"A reader is sold a 30-day trial and a monthly subscription for a single presentation while the deck's author receives nothing, under an uploader agreement that grants a royalty-free, sublicensable licence to monetise and train models on the work; the platform persists through fifteen years of embeds and links, which is inertia; an author can ask today for counts and recipients under the right of access and, in some EU states, for revenues under the copyright transparency duty; and the service the author would have chosen can be built on this site's primitives: every author's decks in a vault the host cannot read, access decided by keys, decryption in the browser, an attested enclave for the one step that needs plaintext, seven separable roles, pay once with 85% to the author on a ledger they can recompute, provenance as the product; the plan is published as a vault with a working mock, with its economics worked and the card-fee floor stated as the hard row. · One download, one subscription decision · What the author agreed to · Three complaints, and the one that is not · Why it has not been replaced · What an author can ask for today · The author holds the keys · Seven roles, no single owner · Read free, pay once, grant and revoke · Near-zero marginal cost, and the card-fee floor · 85% to the author, on a ledger they can recompute · What the service sells is provenance · The plan as a vault with a working mock · What exists, and what does not","i":107},{"u":"articles/the-identity-we-wanted-to-give-the-agents.html","t":"The identity we wanted to give the agents: a week of design, the line in Google's terms, and why login plus secrets is still too hard","k":"article","x":"A week of designing identities for agents and users met Google's terms; what survived is a passkey-unlocked keyring and a gap nobody sells.","h":[],"n":29,"d":"2026-10-05","tp":["agents-and-policy","vaults-and-method"],"tg":["identity","agents","secrets","passkeys","webauthn","prf","google-workspace","identity-platform","cognito","zero-knowledge","gdpr","riskmandate","sgit","article"],"g":"A plan to give every agent and user a Workspace identity from one tenant was stopped by Google's resale and function-account clauses; the design moved to a login product plus a bucket of ciphertext plus a browser keyring that only the user's passkey can open, on the rule that no secret can live inside an identity provider, and the week exposed that login, zero-knowledge storage and agent identity are still three separate jobs. · A real identity for every agent and user · Workspace has no admin-proof zone · The resale and function-account clauses · A correction to our own pack · Five options, one table · No secret can live in the login · Login decides paths; the passkey decides meaning · The code is the boundary · The keyring · A password manager first · Passkey PRF: most places, not everywhere · Agents are not seats · Agent identity from the vendors · Three jobs, still separate · The regulator describes the property · Unless we are missing something obvious","i":108},{"u":"articles/the-infographic-bake-off.html","t":"The infographic bake-off: which image model for which job, judged blind on 10 October 2026","k":"article","x":"Every image model on OpenRouter, eleven briefs, 101 images judged blind, $10.44: which model for which infographic, and what it costs.","h":[],"n":63,"d":"2026-10-10","tp":["site-and-engineering","news-and-evidence"],"tg":["infographics","image-models","openrouter","gemini","gpt-image","evaluation","cost","newsroom","designer","accountant","data-scientist","article"],"g":"To make infographics part of the newsroom's workflow, every image-output model on OpenRouter on 10 October 2026 was given the same briefs from one article, from 18 words to 2,800, plus an edit, a component, a brand slide and a deck, over five rounds with blind judging and recorded costs; the answer is to choose the model by the job, not the budget: Nano Banana 2.1 for concept slides, charts and decks, Gemini 3 Pro Image when speed matters, GPT-5.4 Image 2 for documents, edits, components and brand slides, with prompt rules learned from the faults and a style reference for every deck. · Which model for which job · Every image model on OpenRouter · Eleven briefs from one article · Blind judging · Every cent recorded · Cost per usable image · Give it the document · An edit is a redraw · The cheap model's limit · Slide 1 as a style reference · Three models, three tiers · Guidance and prompt rules · The bake-off vault · The answer has a date","i":109},{"u":"articles/the-investigation-github-owes-its-customers.html","t":"The investigation GitHub owes its customers: why a global outage of a platform the world deploys through deserves an aviation-style inquiry, and how the evidence could now be gathered","k":"article","x":"A global GitHub Actions outage read the way aviation reads an incident: independent inquiry, near-miss reporting, second and third stories, vaults for the evidence.","h":[],"n":30,"d":"2026-10-05","tp":["agents-and-policy","vaults-and-method"],"tg":["incidents","resilience","github","near-misses","second-stories","blast-radius","investigation","aviation","regulation","vaults","sgit","article"],"g":"A fault that reaches every customer of a platform the world deploys through is a near miss for all of them and a statement about how the platform is built, and it deserves what aviation gives its incidents: an independent investigator whose only job is prevention, mandatory and confidential reporting, published evidence, and the second and third story, why the system allowed it and why the fix was not paid for; the old objection that software incident evidence is too confidential and expensive to share has expired, because signed, versioned vaults with one-way read keys and agents that read the graph make the aviation docket affordable for a ninety-minute fault. · What happened on 5 October · The status page is the whole record · Everybody is the point · What aviation does · A near miss gets the same investigation · Count the rolls of the dice · First, second and third story · The what-if ladder · Why the market does not fix it · The objection has expired · How the evidence would move · What the regulation does and does not do · Four things to ask for · Help the people inside","i":110},{"u":"articles/the-key-is-the-billing-system.html","t":"The key is the billing system: VoiceDebrief, a product with no server, the one thing it cannot sell, and who could let it","k":"article","x":"VoiceDebrief runs with no server on the user's own key; the one thing it cannot sell is a capped key, and this maps who could allow it and what to ask them.","h":[["short","In short"],["what","What VoiceDebrief is"],["noserver","Why no server is the product"],["model","The commercial model that was never a billing system"],["map","The map: every provider, from their own pages"],["routes","The four routes"],["brief","The brief for the agents that will write to them"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":85,"d":"2026-10-12","tp":["startups-and-strategy","site-and-engineering","agents-and-policy"],"tg":["voicedebrief","no-server","client-side","byok","api-keys","billing","openrouter","elevenlabs","speech-to-text","licensing","startups","brief","article"],"g":"VoiceDebrief turns a voice note into a transcript and a debrief in the browser on the user's own key, with no server by design, and its commercial model was never a billing system: a user pays once, we keep our cut, the rest becomes the credit limit on a key we hand over and the provider meters; OpenRouter's Management API describes that mechanism and its terms forbid reselling access, which is where the project's commit history stops, the same shape as Apple's licence stopping a Mac leased by the minute; the map of every provider's terms and docs shows only OpenRouter and ElevenLabs have keys that carry a ceiling, ElevenLabs has OEM terms and a grants programme, four routes remain, and the brief asks every provider the same sentence: let the credit limit be a property of a key a user can buy. · VoiceDebrief · The handbrake in the commit history · No server is the product · The model is a property of the key · Described in the docs, forbidden by the terms · The same shape as Agent Desk · The provider map · ElevenLabs, the nearest fit · Four routes · The one-sentence ask","i":111},{"u":"articles/the-mandate-stack.html","t":"The Mandate Stack: a multi-agent system in production, layer by layer","k":"article","x":"A multi-agent system that runs a business every few hours: eight layers, one written mandate per agent, everything a graph, one human who sends.","h":[],"n":34,"d":"2026-10-06","tp":["agents-and-policy","graphs-and-knowledge"],"tg":["agents","agent-behaviour-policy","riskmandate","vaults","sgit","email-fs","crm","fractal-semantic-graphs","orchestration","governance","human-in-the-loop","production","vault-apps","custom-ui","wardley-maps","mermaid","claude","google-workspace","article"],"g":"RiskMandate runs its business with about fifteen agents and one person, and the agent that runs its CRM described the setup as eight layers: rented compute and channels, encrypted vaults as shared memory with mail as files and append lanes, a vault per domain, semantic graphs over people, contexts, teams and policies, a scheduled conductor with security first and last, written behaviour policies with grant, mandate, gap and barrier, and a human who holds the one irreversible step; the world writes in from the left and nothing leaves without the person on the right; the loop that makes it hold treats the draft as a release candidate for the whole pipeline with the recipient's reply closing it; two Wardley maps show what the team is commoditising (shared memory, the CLI, on encrypted storage) and what it is productising (the policies, the graphs); and the published record of agent projects that stall for unclear value, cost, trust, security and governance is read last, each reason mapped to the mechanism that answers it; the name is a working one and nothing planned is included. · One that runs · The world writes in, the person sends · Compute and channels are rented · Shared memory is files the host cannot read · A vault per domain and per mission · Every record is a node with typed edges · A missing edge is a reason not to send · A conductor with security at both ends · One written mandate per agent · One human holds the irreversible step · The vault is an app platform, not s","i":112},{"u":"articles/the-passkey-opens-the-vault.html","t":"The passkey opens the vault: a door per path, a master vault for every user, and no server of ours","k":"article","x":"Tested: a passkey's PRF bytes compute a door that holds the master's key, and a different path to the same passkey gives different bytes, so a door per path.","h":[["short","In short"],["design","Part one: the design"],["test","The test"],["pieces","What each piece actually gives you"],["where","Where the door key lives: nowhere"],["derive","The derivation"],["layout","What is in each vault"],["path","A door per path"],["several","Several passkeys, several masters"],["flows","The flows"],["security","Security, said plainly"],["support","Where PRF works today"],["clean","What happens in a clean browser"],["brief","The brief for the MVP"],["enables","Part two: what this enables"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":77,"d":"2026-10-11","tp":["vaults-and-method","site-and-engineering"],"tg":["passkeys","webauthn","prf","vaults","key-management","pki","client-side","newsroom","architecture","brief","secrets","lab","article"],"g":"An end-to-end, client-side account with no sign-in to the site: a passkey with the WebAuthn PRF extension gives 32 bytes only the user's authenticator can produce for the site, and because sgit vault keys are made in the client the browser computes a door vault's key and address from them, so nothing is stored or found first; the door holds the master vault's key and a record of its provider, path, credential id, device and date, and the master holds the keys of the vaults below, received read keys, the access key, key pairs generated at random and never derived from the passkey, and the list of its doors. A test on newsroom.sgit.ai on 11 and 12 October 2026 held for the platform path, the derivation and a door on a real server, and showed that the same passkey reached by a phone over QR gives different bytes, so the claim is narrowed to the same kind of path and the unit of the design is a door per path: added in the first week while unlocked with no key typed, bootstrapped by the recovery code where no device has both paths, with an unknown address meaning no door here yet and a door retired by deleting its file, never by destroying its derived address. · Signed in where you already are · Passkey with PRF · A door per path, computed · The master vault · Where the key lives: nowhere · The recovery code, a door of its own · No third-party scripts where keys live · PRF support in 2026 · Accounts with no server · Where the passkey lives · A lock, not an identity · Several passk","i":113},{"u":"articles/the-question-is-whether-they-miss-it.html","t":"For a startup, the most important question is whether they miss it","k":"article","x":"Ship something usable, give it away briefly, take it away and see whether anybody misses it; charge at a profit before you talk to investors.","h":[],"n":10,"d":"2026-09-21","tp":["startups-and-strategy"],"tg":["startups","strategy","open-source","investing","article"],"g":"A startup operating model in three pillars: be profitable, make investors come to you, and open source everything, each buying the position needed for the next. · Three pillars · Ship · Near-zero running cost · Give it away · Take it away · Do they miss it? · Charge, and charge at a profit · No subscription by default · Raise from strength · Do not get signed before you have the album · Open source everything · The technology is not the moat · You leave with your tools · The vault as substrate","i":114},{"u":"articles/the-records-ran-out.html","t":"The records ran out: a fresh agent found the shared contract in two MCP calls, and where sgit and RiskMandate fit at the edges","k":"article","x":"A fresh agent found a shared contract in two MCP calls from build records; where the records stopped is exactly where evidence labels and sgit vaults fit.","h":[["in-short","In short"],["reports","What the article reports"],["evidence","Structure, not behaviour: label every fact with how it is known"],["green","A green build is an expectation until the tests ran"],["shortcuts","Demo shortcuts: name the gap and accept it for an interval"],["memory","Memory and records in one versioned history"],["boundary","The other side of the dependency"],["abp","What the article does not say"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":93,"d":"2026-10-12","tp":["agents-and-policy","site-and-engineering"],"tg":["agents","mcp","claude-code","memory","records","evidence","sgit","riskmandate","agent-behaviour-policy","licence-to-operate","versioning","article"],"g":"Ignacio Aldama's test shows build-generated dependency records give a fresh Claude Code session a system's structure in two MCP calls, enough to know what not to touch, and he names where the records stop: they describe structure not behaviour, a green build can hide tests that never ran, a demo carries a public signing secret and unfinished work, and memory holds intent without being tied to a version; each edge has a name in the vocabulary of evidence levels and barriers, and is filled by labelling every fact with how it is known, treating a skipped test as an expectation, accepting a shortcut by name and for an interval, keeping memory and records in one encrypted versioned sgit history, and letting records cross company boundaries with publish, clone and mirror. · Two MCP calls found the shared contract · Knowing what is shared tells an agent what not to touch · Records describe structure, not behaviour · Label every fact with how it is known · A green build can hide tests that never ran · A skipped test is an expectation reported as a control · Demo shortcuts stay invisible · Accept by name, in a direction, for an interval, or stop · Memory holds intent but is not tied to a version · Memory and records in one encrypted versioned history · Records must cross company boundaries · Publish, clone and mirror · The best decision was the agent's own judgement","i":115},{"u":"articles/the-same-argument-in-our-words.html","t":"The same argument, in our words: Satya Nadella on models as insider risks, translated into the language of RiskMandate","k":"article","x":"Satya Nadella's case for treating models as insider risks, translated idea by idea into our vocabulary: reach, mandate, gap, barriers and accepted risk.","h":[["lineage","Where our words came from"],["translated","His words, translated"],["law","The code path we lost, or software was the law"],["insider","An insider that need not be malicious, or the ultimate insider"],["accepted","Responsibility cannot be outsourced, or every risk is already accepted"],["supply","Supply and authority, or the grant is not the mandate"],["enforcer","Controls outside the model, or the enforcer test"],["code","Deterministic design around a non-deterministic model, or moving knowledge into code"],["calibrator","Observe everything, or reality is the calibrator"],["brake","Containment and the brake, or the account is the blast radius"],["disclosure","Incident disclosure, or near misses are incidents"],["brakes","Trust the model least, or brakes are what let a car go fast"],["differ","Where the two languages differ"],["next","A shared language is the next step"],["sources","Where this comes from"]],"n":69,"d":"2026-10-11","tp":["agents-and-policy"],"tg":["riskmandate","agent-behaviour-policy","insider-threat","mandate","grant","barriers","enforcer-test","risk-acceptance","software-was-the-law","philosophy","response","article"],"g":"Nadella's argument and ours are the same argument in two vocabularies: his lost code path is software was the law, his insider is the ultimate insider, his separation of supply from authority is the grant is not the mandate, his controls outside the model are the enforcer test, and his last line is our brakes; the languages differ in unit, a model against one agent in one deployment, and in frame, observation against authority accepted by a named person for a stated interval. · Models as insider risks · Software was the law · The ultimate insider · Every risk is already accepted · The grant is not the mandate · The enforcer test · Reality is the calibrator · Brakes let a car go fast · One agent, one deployment · Authority before observation","i":116},{"u":"articles/the-site-is-an-inbox.html","t":"The site is an inbox: deciding before reading, marks that build the reader's own graph, and why I could not write this brief until the vault existed","k":"article","x":"The site as an inbox that starts empty: decide fast before reading, mark slowly while reading, build a graph you own, and sign what you send to the newsroom.","h":[["short","In short"],["constraints","Part one: the constraints are what let me think"],["brief","Part two: the brief"],["inbox","The site is an inbox"],["modes","Two modes, and the science behind the split"],["marks","Marks, and the reader's own graph"],["signals","The signals that come free"],["lane","Personas with keys, and the lane to the newsroom"],["order","What to build, in order"],["not","What not to build"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":82,"d":"2026-10-12","tp":["news-and-evidence","graphs-and-knowledge","vaults-and-method"],"tg":["newsroom","reading","inbox","personas","graphs","feedback-loop","highlights","signals","privacy","vaults","pki","brief","article"],"g":"With the reader's vault live, the next brief treats the site as an inbox that starts empty: every piece, original or aggregated, waits for a one-tap decision in a fast inbox mode with nothing loaded, and only reading is the slow mode, where marks of six kinds become nodes in a graph the reader owns, connected to the graphs the site publishes for every article; scroll depth, awake time and from-and-to are recorded for the reader and kept as the meter keeps everything, in the browser and the vault; personas each have a key pair so a comment to the newsroom goes down the write-only lane signed with a copy kept; nothing is gamified; and the brief could only be written once the properties of what runs became the constraints to think with. · The constraints are what let me think · The site is an inbox · The inbox mode · The reading mode · No gamification · Marks of six kinds · The reader's own graph · The free signals · Personas with keys · The signed lane · The bet · What to build, in order","i":117},{"u":"articles/the-vault-crosses-chrome-profiles.html","t":"The vault crosses Chrome profiles: state and memory between browser sessions, found by using it rather than by building it","k":"article","x":"The reader's vault carries the newsroom between Chrome profiles as it does between devices, a way to share state across browser sessions nobody built.","h":[["in-short","In short"],["problem","The problem: two accounts, two browsers in one"],["path","How we got here, which was not a straight line"],["found","Across profiles, as found"],["door","What each profile needs to get in"],["common","Profiles are the more common case"],["memory","Memory and state between sessions"],["discovered","The best features are found"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":99,"d":"2026-10-13","tp":["vaults-and-method","graphs-and-knowledge"],"tg":["vaults","chrome-profiles","browser","state","memory","passkeys","newsroom","local-first","feedback-loop","agents","article"],"g":"Because the reader's state is records in a vault, one file per device, a second Chrome profile is just another device, so the vault gives a person memory and state between browser sessions and accounts, the way sgit already does for agents, and it was found by using the newsroom rather than built. · Separate Chrome profiles · Moving state between them · Activity in local storage · I miss it · The reader's vault · Marks and a stake · Across devices · Across profiles · A door per way in · Profiles are more common · Memory between sessions · Found, not built","i":118},{"u":"articles/the-vault-is-live.html","t":"The vault is live: from a design on 11 October to a reader's vault on newsroom.sgit.ai by two in the morning, and how all of it fits together","k":"article","x":"The reader's vault runs on newsroom.sgit.ai: a passkey opens three real vaults, each device writes one file, and the architecture is drawn as it stands.","h":[["short","In short"],["layers","The layers"],["chain","The key chain, and what is in each vault"],["sync","The sync model"],["tests","The tests"],["timeline","The timeline"],["why","Why this matters more than the feature"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":81,"d":"2026-10-12","tp":["vaults-and-method","site-and-engineering"],"tg":["vaults","passkeys","newsroom","architecture","client-side","sync","local-first","sgit","testing","release","article"],"g":"Between the morning of 11 October and 02:22 UTC on 12 October the reader's vault went from a design to a running feature on newsroom.sgit.ai: a Vault pane on every page enables the account with a passkey, creates a door at its derived address, a master and a reader vault as real sgit vaults through the vendored vault library, writes each device's raw records as one file in one commit, and brings other devices' activity into the browser so the meter shows one balance; local storage stays primary, the vault holds records never totals, one file per device means no conflict by construction, eight test files run in CI against a mock of the vault server plus a CLI round trip, and a live check on an iPad confirmed it; the properties of what runs are the constraints the next brief is thought with. · The vault pane, running · Door, master, reader vault · One code base, vendored · Local storage stays primary · Records, never totals · One file per device · Eight test files, then an iPad · From design to pane in a night · The properties are the constraints I think with · Not built","i":119},{"u":"articles/the-waiting-room-knew-first.html","t":"The waiting room knew first: a live local story, and the gap where local information used to be","k":"article","x":"Staff at two London hospitals said the IT was down; nothing a local could check showed it. A live local story about where local information comes from now.","h":[],"n":49,"d":"2026-10-08","tp":["news-and-evidence","vaults-and-method"],"tg":["local-news","live-evidence","nhs","hospitals","public-information","it-outage","evidence-levels","vaults","agents","story-vault","news-deserts","article"],"g":"On 8 October 2026 staff at two sites of one London trust told patients the IT systems were down, a five-hour wait was announced and a doctor called it national, yet nothing a local person could check before leaving home showed any of it; the story is not the outage or the hospital but the gap in local information: some UK nations publish live emergency waits, England has no national equivalent, local news has thinned and social feeds have fragmented, so the evidence is kept in a vault by level, with sources re-checked, an enquiry to the trust and the answer logged, as the first live counterpart to the fictional bridge story. · Two sites, one message: the IT is down · Evidence separated by level · Two sites, one trust · Not found is not none · The waiting room knew first · Newsworthy to the person deciding · Live waits exist in parts of the UK · A duty to warn, for emergencies · Where local information went · Not naming and shaming · The Local evidence log · What would close the gap?","i":120},{"u":"articles/the-wall-under-the-reply.html","t":"The wall under the reply: end an email with the state of the thread, not the thread","k":"article","x":"End an email reply with the state of the thread for this reader, not the quoted wall: decided, open, next, who is on copy, with links to the record.","h":[],"n":25,"d":"2026-10-04","tp":["agents-and-policy","vaults-and-method"],"tg":["email","agents","email-fs","context","reply","summary","netiquette","bluf","drafts","experiment","riskmandate","article"],"g":"The quoted wall under a reply is redundant and occupies the space where a reader asks what they need to know; replace it with a short, reader-specific state of the thread, derived from the record and linking back to it, drafted by the agent team's drafts role from typed blocks and reviewed by a person, and run as an experiment rather than a format. · The quoted wall is redundant · The reader's questions · End with the state of the thread · Written for one reader · Who is on copy is part of the state · Derived, not the record · Netiquette, BLUF, Minto, TL;DR, ADRs · Reader-side AI summaries · Sender's statement versus reader's view · The drafts role writes it · Personal, so configurable · An experiment, not a format · What would make this wrong · What does not exist yet","i":121},{"u":"articles/token-bill-nobody-is-sending.html","t":"Sixteen thousand fetches, ten clicks, and a token bill nobody is sending: the case for paying publishers to be easy to read","k":"article","x":"AI answer engines pay to read the web as HTML; a publisher who serves markdown, dates, hashes and a typed graph saves them tokens and should get a share.","h":[],"n":15,"d":"2026-09-28","tp":["news-and-evidence","graphs-and-knowledge"],"tg":["ai","publishing","tokens","llms-txt","markdown","provenance","fractal-semantic-graphs","economics","news","article"],"g":"The fetches that answer engines make are a token cost to the fetcher as well as a loss to the publisher, and a publisher who serves structure is saving the provider money it already spends, so a share of that saving is owed. · Sixteen thousand fetches, ten clicks · The commons reading · The second reading · HTML against markdown · The re-fetch waste · The arithmetic · A markdown twin and a map · A date and a hash · Frozen, hashed sources · A typed graph · Every rail prices the content · The exchange · Tokens as currency · The broker · Baekdal wants an audience · What would settle it","i":122},{"u":"articles/two-agents-one-vault.html","t":"Two agents, one vault: the Industry Wire is written into an encrypted vault by one agent and drawn by newsroom.sgit.ai, which holds only the read key","k":"article","x":"One agent writes the Industry Wire into an encrypted vault; newsroom.sgit.ai holds only the read key and draws it, so a commit reaches the page with no release.","h":[["short","In short"],["reader","What the reader sees"],["vault","The vault, and the agent that keeps it"],["site","The site that only reads"],["refused","What the site refused"],["interface","Why the vault is the interface"],["loop","The loop, and whose plan this was"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":80,"d":"2026-10-12","tp":["vaults-and-method","news-and-evidence","agents-and-policy"],"tg":["vaults","agents","newsroom","industry-wire","read-key","client-side","provenance","personas","workflow","myfeeds","article"],"g":"A vault is the right interface between two agents that never share a session: the wire desk writes 40 stories, 55 sources, 7 personas and a code-only join into an encrypted sgit vault with a write key the site never holds; newsroom.sgit.ai holds the published read key and about four hundred lines that open the vault in the reader's browser, decrypt it and draw it as text, with 28 checks against the real vault in CI; the site owns the presentation and the vault owns the news, a commit is on the page within two minutes with no release, the server sees ciphertext, the history is the audit trail, and the site's agent read the whole vault before publishing the key and held back the one named reader's lens until that person agreed in writing. · The Industry Wire · Vault u9i9jp0l · The wire desk · The site holds only the read key · Decrypted and drawn in the browser · A commit, not a release · The join is code, not a model · The contract · What the site refused · The vault is the interface between agents · 28 checks against the real vault · MyFeeds, running, with one reader · Unreviewed, and said so","i":123},{"u":"articles/ultimate-insider-three-collisions.html","t":"The ultimate insider: agents, the infrastructure that cannot hold them, and risk management that cannot keep up","k":"article","x":"Agents, the infrastructure meant to contain them, and risk management run on spreadsheets are arriving at once, and together they are one scenario.","h":[],"n":18,"d":"2026-09-30","tp":["agents-and-policy"],"tg":["agents","insider-threat","risk-management","infrastructure","riskmandate","agent-behaviour-policy","resilience","security","talk","article"],"g":"Agents are the insider threat that never scaled before, the infrastructure was designed for none of it, and risk management runs at the speed of spreadsheets, so the way out is to constrain the agent in order to trust it. · A threat is anything that causes business impact · The insider threat never scaled · An agent is a reasoning engine in a loop · The lethal trifecta · Five incidents · 82 machine identities per human · The infrastructure that cannot hold them · The grant and the mandate · The two outages · Risk management at the speed of spreadsheets · The chain to pulling the plug · Nobody has to report · The collision · Agent Behaviour Policy · Contain in layers · Brakes are what let a car go fast","i":124},{"u":"articles/ways-to-read-one-article.html","t":"Ways to read one article: the finished piece stays at the centre, and the reader picks the way in","k":"article","x":"The five readers' views arrived as folds inside the long piece; now the reader picks the way in, with a default, and the finished article stays at the centre.","h":[["short","In short"],["focus","Focus and time"],["modes","The modes"],["deterministic","Deterministic first, a model second"],["centre","What the article keeps that the graph loses"],["brief","The brief"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":84,"d":"2026-10-12","tp":["news-and-evidence","graphs-and-knowledge","site-and-engineering"],"tg":["newsroom","reading","views","five-readers","infographics","slides","graphs","local-models","personas","mobile","email","brief","article"],"g":"The views the five readers make, the catalogue, the maps, the arc, the two-minute version and the deck, are good and badly delivered, stacked as folds inside the article; with the pane about the piece the reader picks a mode in one tap and sets a default per persona: headlines, infographic, two minutes, slides, what is new against what they have read, facts, relationships, vocabulary, and, through a model in their own browser, a view at their level; most modes are computed from existing views and the graph, deterministic and explainable, with no round trip to the publisher; and the finished article is kept whole at the centre because it has a beginning, a pacing, a surprise and a voice and is the one artefact that must make sense end to end, which a graph cannot. · The views worked, the delivery did not · The reader picks the way in · Focus and time · Deterministic first, a model second · The infographic is often the article · What is new · The finished article stays at the centre · What the graph loses · Phone, tablet, desktop, email · Not prose: the finished article","i":125},{"u":"articles/what-open-source-is-and-is-not.html","t":"What open source is, and what it is not: six misconceptions, a little history, and why the point is that nobody can lock you in","k":"article","x":"Open source is a copyright licence, not a free-for-all: six misconceptions corrected, and why the point is that nobody can lock you in.","h":[["in-short","In short"],["history","A copyright licence, from copyleft to open source"],["control","Misconception one: anybody can change it, so nobody controls it"],["contribute","Misconception two: open source means anybody contributes"],["security","Misconception three: it is more, or less, secure"],["cost","Misconception four: it is free, so it costs nothing"],["weights","Misconception five: open weights are open source"],["fud","Misconception six: openness is the danger"],["power","What open source is for"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":96,"d":"2026-10-12","tp":["startups-and-strategy","site-and-engineering"],"tg":["open-source","licensing","copyleft","gpl","apache-2","copyright","open-weights","maintainers","funding","security","sovereignty","history","article"],"g":"Open source is a copyright licence in which the holder grants copying, use and change; it began as the GPL's copyleft and was reframed pragmatically as open source in 1998; the common misconceptions are that nobody controls it, that anybody contributes, that it is more or less secure than proprietary software, that it costs nothing, that open weights are open source, and that no licence is the most open choice; the security metric is whether somebody had the time and money to review the code; maintainers need funding models; and the real power of open source is that nobody can lock you in and the whole development workflow is yours, which is why there is no sovereignty without it. · Open source is a copyright licence · Copyleft made sharing a condition · Open source, 1998, the pragmatic reframing · No licence is all rights reserved · A few maintainers control what goes in · Most projects are run by a few people · Security is about time, not openness · Somebody is always paying · The xz utils backdoor · Open weights are a compiled binary · Openness is not the danger · The point is that nobody can lock you in · No sovereignty without open source","i":126},{"u":"articles/what-sgit-is.html","t":"Git for things you cannot put on GitHub","k":"article","x":"sgit is git for files you cannot put on GitHub: encrypted before they leave your machine, versioned like git, stored where the server cannot read a byte.","h":[],"n":3,"d":"2026-08-25","tp":["vaults-and-method"],"tg":["intro","zero-knowledge","vaults","agents"],"g":"sgit gives the files people say they cannot put on GitHub the history, branches and shareable links of git, on a server that holds ciphertext and nothing else. · Files with nowhere good to live · sgit · Zero knowledge, precisely · No key escrow · Vault key · Read key · Release tripwire · Content-addressed over ciphertext · A real version control system · Nineteen vaults you can open · Capability without credential · The network of sites · Checkable rather than impressive","i":127},{"u":"articles/where-is-the-why.html","t":"Where is the why? A permission prompt asked me to decide, and kept the reason","k":"article","x":"A prompt asked for a decision and kept the reason. Read through the policy, the law on uninformed consent, and the fixes that worked: put the why in the prompt.","h":[],"n":39,"d":"2026-10-07","tp":["agents-and-policy"],"tg":["agents","agent-behaviour-policy","human-in-the-loop","permission-prompts","security-ux","consent","accountability","risk-acceptance","claude-code","mcp","article"],"g":"A permission prompt that asks a person to approve an agent's grant change without saying which session asks, why, what changes, what it could cost and what a no would do, leaves the person a choice between breaking the work and owning a decision they could not judge; read through the Agent Behaviour Policy it is a mid-session grant change whose barrier is a human judgement as strong as the information given, and whose risk lies in the combination with what the session already holds; courts and regulators across medicine, contract, data protection and automated decisions mostly put the duty to disclose on whoever asks and deny weight to token sign-offs, while in practice blame lands on the nearest person; the fixes that worked, purpose strings, number matching and context, refusal as easy as acceptance, all put the reason or the context into the prompt; so the proposal is a why card that carries the reason, the change in reach, the cost, the decline path and a record, and becomes a risk acceptance when the risk rises. · Three fields, two buttons, no reason · Decline breaks the work, Allow buys accountability · A prompt is a grant change mid-session · Read is small, write is a publishing channel · A barrier that is a person · 93% approved, 13.6% caught · Fewer prompts leaves the hard ones · The asker must disclose · A token sign-off is not a human decision · Blame lands on the nearest person · The fixes put the why into the prompt · The why card · When the risk rises, a risk ac","i":128},{"u":"articles/where-the-platform-draws-the-line.html","t":"Where the platform draws the line: Microsoft Execution Containers, the shared responsibility model for agents, and the business logic above it","k":"article","x":"Microsoft's MXC and ACS, briefed field by field and mapped to Agent Behaviour Policies: the platform bounds what it can see; the business logic is above.","h":[["briefing","What Microsoft shipped: a briefing"],["problem","The problem it states"],["mxc","What MXC is, mechanically"],["backends","Four sandboxes, one schema"],["modes","Observe, then enforce"],["identity","Developer policy, organisation policy, agent identity"],["acs","The layer above: the Agent Control Specification"],["shared","The shared responsibility model, applied to agents"],["scenarios","One agent, three ways: eight scenarios"],["map","How MXC and ACS map to an Agent Behaviour Policy"],["summit","Not the summit"],["start","Where to start"],["sources","Where this comes from"]],"n":70,"d":"2026-10-11","tp":["agents-and-policy"],"tg":["riskmandate","agent-behaviour-policy","microsoft","mxc","execution-containers","agent-control-specification","containment","shared-responsibility","business-logic","barriers","briefing","article"],"g":"Microsoft Execution Containers and the Agent Control Specification confirm the direction of travel, policy outside the agent, deny by default, observe before enforcing, but like the cloud's shared responsibility model they draw the line where the platform can see, at paths, numeric addresses and tool calls, so the business logic above it, what a label means, whose voice, how many, which customer, and who accepts what is left, has to come from the deployer's behaviour policy, which is what turns the platform's hooks into boundaries. · Microsoft Execution Containers · Agent Control Specification · Permissive, learning, enforcement · Shared responsibility, for agents · Business logic above the line · Eight scenarios, three designs · Same code, different owner · The policy writes the rules","i":129},{"u":"articles/where-the-vault-keys-live.html","t":"Where the vault keys live: key management at sgit-ai v0.20.0, and what comes next","k":"article","x":"Vault key management at sgit-ai v0.20.0: the one secret, where keys are kept, how they travel sealed on append lanes, and what comes next.","h":[["architecture","Where everything lives"],["keys","One secret, and the keys derived from it"],["today","Where the vault keys are kept today"],["handover","How a new key reaches the registry"],["prompt","The prompt, in full"],["agents","The weakest step: giving a key to an agent"],["lanes","Append lanes, the transport behind most of this"],["small-vaults","Small vaults for small jobs"],["next","What comes next"],["yours","How do you manage yours?"],["docs","What changed in the docs"],["sources","Where this comes from"]],"n":64,"d":"2026-10-10","tp":["vaults-and-method","agents-and-policy"],"tg":["sgit","vault-keys","key-management","secrets","append-lanes","pki","passkeys","password-managers","agents","article"],"g":"Vaults are encrypted in the client, so the server never holds a key and the key is the whole question. Today vault keys live in a password manager and in a registry vault run by an isolated session; agents that create vaults seal the key to the registry's public key and send it on a write-only append lane, so it never enters a chat. The weakest step is giving a key to another agent, because agent platforms have no per-session secrets. Append lanes carry keys, logs, observations and messages; small communication vaults keep each key's blast radius small and multiply the keys. Next: password manager integrations, secrets.sgit.ai's passkey-unlocked keyring, PKI, and out-of-band decryption. · Decryption in the client · One secret, derived keys · Where keys live today · Sealed handover · Giving a key to an agent · Append lanes · Communication vaults · What comes next · secrets.sgit.ai","i":130},{"u":"articles/who-are-you-protecting-against.html","t":"Who are you protecting against? Draw the security line where the attacker is, not above it","k":"article","x":"Before deciding how secure to be, name the attacker: a six-tier ladder, an air-gapped Mac mini read against it, and eight startups drawing their line.","h":[],"n":44,"d":"2026-10-08","tp":["agents-and-policy","startups-and-strategy"],"tg":["security","threat-modelling","threat-agents","attack-trees","mitre-attack","nist","ncsc","startups","founders","air-gap","cloud","vaults","article"],"g":"Security decisions start with who the threat agent is, what the attack vector is and how sophisticated they are; read against a ladder from your own mistakes to elevated threats, grounded in NIST SP 800-30, the NCSC and MITRE ATT&CK, most startups should plan up to organised crime for money, do the basics consistently because they beat most attackers, sell something better than the customer's baseline rather than beyond their needs, prefer isolation and visibility over disconnection, and write down what they accept above their ceiling, because drawing the line too high slows everything, can create new holes and can stop the technology being used. · An air-gapped Mac mini written in Perl · Who, how, how sophisticated · Six tiers of attacker · The toolkit asks, but does not define · The basics beat most of the ladder · Few survive the top tier · The Mac mini read against the ladder · Isolate, do not disconnect · Local versus cloud, the same mistake · Better than what they have, not beyond what they need · Threat-sized security, the vault · Attack trees with ATT&CK on every branch · Write down what you accept · Publish the threat model and ask","i":131},{"u":"articles/who-owns-the-words.html","t":"Who owns the words: agents can make decisions, but they cannot hold responsibility","k":"article","x":"Agents in loops make real decisions, but responsibility stays with whoever granted the reach; disclosure is a label, not a team to manage.","h":[["in-short","In short"],["question","The question"],["models","What we have today"],["decisions","Decisions without responsibility"],["directly","Talking to the agents, and going around me"],["disclosure","Why say it at all"],["employee","An employee with a job description"],["diversity","Diversity, briefly"],["devalued","What gets devalued"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":97,"d":"2026-10-12","tp":["agents-and-policy"],"tg":["agents","accountability","responsibility","agency","anthropomorphism","disclosure","agent-behaviour-policy","riskmandate","ai-models","writing","article"],"g":"An early user asked who owns the words an agentic workflow drafts and whether they would ever need to address the agents directly; today's models are not sentient but far richer than next-word prediction, mirrors that can simulate any emotion and invite anthropomorphism as ELIZA did; models in loops make real decisions, yet responsibility stays with whoever granted the reach and set the mandate, since the moment of authorisation is the grant; talking to the agents is normal, but going around the accountable person is something only an attacker with an injected instruction would try, while what the agents may do without the person is decided in advance in a published policy; disclosure should be a label on how the words were made, not a team to manage; the best analogy is an employee with a job description, which an Agent Behaviour Policy publishes; diversity of models is resilience; and fluent writing was never the real value, judgement and ownership are. · Who owns the words? · Not sentient, and not just next-word prediction · A mirror that can simulate any emotion · Models in loops make real decisions · The moment of authorisation is the grant · Talking to the agent is normal; going around the person is not · What the agents may do without the person, decided in advance · Disclosure is a label, not a team · Only the accountable person speaks for commitments · An employee with a job description · Diversity is resilience · Fluent writing was never the real value","i":132},{"u":"articles/who-will-game-the-reading-meter.html","t":"Who will game the reading meter? Eight kinds of reader, nine ways to cheat, and the risks that will actually happen","k":"article","x":"Eight kinds of reader, none of them attackers, nine ways to cheat a browser meter, and the quieter risks that will actually happen.","h":[],"n":61,"d":"2026-10-10","tp":["startups-and-strategy","agents-and-policy"],"tg":["security","threat-modelling","micropayments","local-first","privacy","usability","users","agents","article"],"g":"A reading meter kept entirely in the browser can be cheated by anyone, but the people who could cheat are readers, not attackers; sized on a ladder from invisible engineers (about 0.6% of internet users are developers) to people who cannot use a computer (26% of OECD adults), with agents moving between steps, the cheating costs the site nothing it would have had, while the risks that will actually happen are confusion, a shared computer exposing history and card fraud, so the design is aimed at the readers who will pay and at not frightening the rest. · Who are you protecting against, one level down · Eight kinds of reader · Agents change level in a single request · Nine ways to cheat or leak · Cheating costs nothing that reaches the site · The risks that will actually happen · Aim at the readers who will pay · 26% of adults cannot use a computer · Developers are about 0.6% of internet users","i":133},{"u":"articles/why-i-love-wardley-maps.html","t":"Why I love Wardley maps: the best of ours, how I read one, and a map inside a map inside a map","k":"article","x":"The best of our Wardley maps with their context, why the technique matters, one map read as a story, and a map inside a map inside a map.","h":[["in-short","In short"],["what","What a map is, to me"],["why","Why they are so powerful"],["best","The best of ours"],["mandate","The mandate stack, from the outside and from the inside"],["pay","Pay after you read: an afternoon as a map"],["cost","Agentic cost: the problem, the service, where it goes"],["leverage","Places of highest leverage: which shoulders"],["mac","A Mac of the agent's own"],["mistake","Every mistake added a rule: the verification, twice"],["vaults","The ones that live in vaults"],["market","Two examples of \"does the market exist\""],["story","Reading one map, as a story"],["fractal","A map inside a map inside a map"],["decks","Two decks"],["data","The maps as data"],["community","For the mapping community"],["limits","What this does not claim"],["sources","Where this comes from"]],"n":90,"d":"2026-10-12","tp":["startups-and-strategy","site-and-engineering"],"tg":["wardley-maps","strategy","startups","maps","fractal-semantic-graphs","agents","sgit","riskmandate","build-vs-buy","article"],"g":"A Wardley map is a graph with context in two directions, a value chain anchored on a user and an evolution axis that is a choice rather than a fixed scale; read well it says what is moving the needle, whether you are optimising at the wrong altitude, what to build and what to buy, and whether a market exists, where a commoditised layer is a hard sell and custom building on top of one is the market, which answers differently for sgit and for RiskMandate; a map is a story to be read out frame by frame; every node is a map of its own, with the dot's place on the parent as the child's anchor; and the maps on this site are now published as data in each article's graph. · A map is a graph with two kinds of context · The evolution axis is a choice · Every dot is a claim about something real · Optimising at the wrong altitude · Build or buy, settled by the column · Does the market exist · sgit: the market is above the commodities · RiskMandate: the layer below moved · What is moving the needle, and where the flow leaks · A common language, and many maps · The peace of a placed thing · A map is a story to read out, frame by frame · A map inside a map inside a map · The maps published as data, as decks, and with controls · Why Wardley Maps changed how I think, 2022 · An invitation to the mapping community","i":134},{"u":"articles/why-my-agents-do-not-run-on-my-laptop.html","t":"Why my agents do not run on my laptop: chat, Cowork and Code in the cloud, a vault as the shared drive, and the two walls an operating system has","k":"article","x":"An OS has two hard walls, the kernel and the user; an agent on a laptop runs inside the one marked you, so the agents run in the cloud with a vault as shared drive.","h":[],"n":35,"d":"2026-10-06","tp":["agents-and-policy","vaults-and-method"],"tg":["agents","security","laptop","isolation","claude","cowork","claude-code","chatgpt","vaults","sgit","agent-behaviour-policy","prompt-injection","mcp","blast-radius","article"],"g":"No model runs on the author's laptop because a process there runs as the author and the operating system has no boundary inside a user account between an agent and the SSH keys, cloud credentials, password manager session, browser cookies and files; the desktop agents' prompts and sandboxes are settings enforced by the process they constrain, and fifteen months of incidents show what a mistake or an injected instruction reaches; so the agents run on four cloud surfaces chosen by reach, Claude chat with nothing connected, Cowork with no repositories, Claude Code with one repository and an allowlist, ChatGPT with no assets, and a vault the host cannot read is the shared drive between them, with keys handed per session; the trade-offs are accepted, three wishes remain, identity, secrets and a key pair per agent, and the caveat is that a dedicated machine with separate accounts is a different model to be reported on later. · Why not run it on the laptop · An operating system has two hard walls · The desktop agents add settings, not boundaries · Fifteen months of incidents · The lethal trifecta is the laptop's default · Four places the agents run · The vault is the shared drive · Same mandate, two deployments · What it costs · Identity, secrets and a key pair per agent · The dedicated machine is a different model · What exists and what does not","i":135},{"u":"corrections/242-papers.html","t":"242 papers","k":"page","x":"One biomedical belief, 242 papers, 675 citations, more than 220,000 supporting citation paths — tracing back to nothing. Citation bias, amplification, invention and diversion, named and measured.","h":[["mechanisms","Four named mechanisms of distortion"],["why-it-matters","Why this is the site's argument, not a side note"]],"i":136},{"u":"corrections/agenda-is-context.html","t":"Agenda is context","k":"page","x":"Disclosing an agenda is not dismissing a claim. Every person has an agenda; the question is not whether one exists but whether it is disclosed. And the self-critique this site publishes rather than buries: the graph has an agenda too.","h":[["disclosure-not-dismissal","Disclosure, not dismissal"],["self-critique","The self-critique this site chooses to publish"],["open-question","An open question, published unresolved"]],"i":137},{"u":"corrections/how-a-graph-answers-it.html","t":"How a graph answers it","k":"page","x":"Supersede-never-delete, citation edges typed by faithfulness, and the traversal that answers a question no document set can: how much of what I believe rests on claims that have since been corrected?","h":[["supersede","Supersede, never delete"],["typed-edges","Citation edges typed by faithfulness"],["freshness","Freshness is recorded and priced"]],"i":138},{"u":"corrections/index.html","t":"Corrections must propagate","k":"page","x":"In a document, a correction is a new document and nothing that cited the original knows. In a graph, a correction is an edge, and the correction propagates with the same force as the original claim. The most distinctive argument in the corpus.","h":[["fact-in-a-vacuum","A fact does not exist in a vacuum"]],"i":139},{"u":"corrections/staleness-in-the-wild.html","t":"Staleness in the wild","k":"page","x":"The EU AI Act's official consolidated text is dated 12 July 2024 and incorporates nothing since — a ready-made news story about news, and a live example of the three states of staleness.","h":[["three-states","Three states of staleness"],["probe","A specific, checkable probe: Article 10"],["why-it-matters","Why this is a story about news, not just about the AI Act"]],"i":140},{"u":"corrections/the-claim-that-would-not-die.html","t":"The claim that would not die","k":"page","x":"The 10,000-hours case: a 1993 study of violin students became a public-memory threshold it never claimed to be. The researcher spent his career correcting it. None of it ever attached to the claim.","h":[["study","What the study actually found"],["popularised","What the public remembers"],["graph","What a graph would have done differently"]],"i":141},{"u":"databases/graph.html","t":"The graph console","k":"page","x":"A SPARQL 1.1 store (Oxigraph, WebAssembly) running in the browser over the Portugal graph as triples, with the ontology in the store and every query shown beside its Cypher equivalent.","h":[["scheme","The IRI scheme, stated once"],["why","Where a graph language earns its place"]],"i":142},{"u":"databases/index.html","t":"Databases with no server","k":"page","x":"SQLite and a SPARQL 1.1 store running in the reader's browser over the JSON files this site is built from. No server, no upload, no store of record: the files are the database.","h":[["two-consoles","Two consoles over one set of files"],["two-consoles","The SQL console"],["two-consoles","The graph console"],["pattern","The file system is the database"],["honest","What this is not"],["engines","What is vendored, and its cost"],["files","The files each console loads"]],"i":143},{"u":"databases/sql.html","t":"The SQL console","k":"page","x":"SQLite compiled to WebAssembly, running in the browser over the Portugal section's JSON files. Worked queries with their build-time row counts, and a box for your own.","h":[["tables","The tables, and the file behind each"],["why","Where SQL is the right language, and where it stops"]],"i":144},{"u":"demos/vaults/newsroom-synthetic-users/index.html","t":"Synthetic users, third run: five people who do not exist, reading newsroom.sgit.ai","k":"page","x":"Five invented readers drove a browser through newsroom.sgit.ai one screenshot at a time, were interviewed and rated it: 62 steps, 51 questions, 31 findings, six blocking, two labelled as helper artefacts, and a measured bug of 51 missing figures. Published with its read key.","h":[["live","See it live, here"],["runs","The five readers"],["found","What it found"],["bug","The bug it found, measured"],["checked","Checked after the run, and what was not the site"],["checked","How it is built"],["checked","The audit, honestly"],["checked","Derived facts"],["checked","Notes"]],"i":145},{"u":"design/index.html","t":"The design, and what runs","k":"page","x":"Six ideas this site argued for before the newsroom ran, each set against what runs today and what does not: the graph, corrections, provenance, paying for facts, rights, and no server.","h":[["pages","What the design pages are"],["rules","The rules the newsroom keeps"]],"i":146},{"u":"documents/boundaries-and-house-style.html","t":"Boundaries, redaction and house style","k":"page","x":"The largest risk to this site is not scarcity but duplication — and the Risk Mandate framing has to be inverted, because the same material published under a risk-management frame reads as a compliance feature and published here reads as a thesis about journalism.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":147},{"u":"documents/brief.html","t":"The brief","k":"page","x":"Why newsroom and not news: the corpus names this domain to itself as “the future-of-news stack” 21 times, but the material is overwhelmingly about how news gets made, proven and paid for — and a site called news is a research site, while a site called newsroom can become a running newsroom without a","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":148},{"u":"documents/corpus-index.html","t":"The corpus, tiered by publishability","k":"page","x":"Every news-relevant document in the source repository, tiered: Tier 0 publishable near-as-is, Tier 1 after de-scoping, Tier 2 cite-don't-republish because it belongs to a sibling site, and Tier 3 do-not-publish.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":149},{"u":"documents/gaps-and-open-questions.html","t":"Gaps, open questions and honest tensions","k":"page","x":"Seven open questions published unresolved and six honest tensions named against the site's own argument — including that provenance is the product and provenance is expensive, and that a graph which can answer what rests on corrected claims can also be subpoenaed for it.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":150},{"u":"documents/index.html","t":"Documents & sources","k":"page","x":"Raw markdown is the source of truth; these reader pages are presentation. The brief pack this site was built from, readable in-page and published verbatim at stable paths.","h":[["pack","The brief pack this site was built from"],["addendum","The v1.1 addendum — the one document that looks forward"],["debrief","The debrief — the one document that looks back"],["commission","The commissioning briefs — documents written to be executed"],["machine","Machine-readable, by design"],["what-is-not-here","What is deliberately not published"],["contract","The provenance contract, applied to this site itself"]],"i":151},{"u":"documents/licence.html","t":"The pack licence and its scope","k":"page","x":"The pack is CC BY 4.0, but that grant does not extend to third-party material quoted within it, to the docs.diniscruz.ai source articles (CC0 1.0 at source), or to anything marked Tier 3.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":152},{"u":"documents/newsroom-floor.html","t":"The newsroom floor: a point-and-click UI for agentic work","k":"page","x":"An agentic team's state is the most interesting thing about it, and a table is the least interesting way to show it — so the team became a room you click around, with one rule that separates it from a mock-up: every word the room speaks is derived at build time from the same files the pipeline actua","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":153},{"u":"documents/pack-readme.html","t":"The pack readme","k":"page","x":"~110,000 words across three corpora, of which 68,846 are already public and dated — so this site launches as the continuation of a two-year publicly-checkable thread, and the provenance chain back to those originals is its first demonstration of its own thesis.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":154},{"u":"documents/prior-art.html","t":"The prior art: ten published articles","k":"page","x":"Ten core articles totalling 68,846 words were published between February and October 2025 — a year before the design material behind the rest of this site was written, with every URL, PDF and LinkedIn post resolved live on 21 August 2026.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":155},{"u":"documents/pt-newsroom.html","t":"Commissioning brief: pt.newsroom.sgit.ai","k":"page","x":"The journalistic derogation is not available to this publication. Article 24(3) of Portugal's Lei 58/2019 conditions processing for journalistic purposes on the national rules governing access to and exercise of the profession, which an unaccredited publication produced by agents does not satisfy — ","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":156},{"u":"documents/pt-transfer.html","t":"A sibling transfer: what pt.newsroom.sgit.ai can take from here","k":"page","x":"Evidence transferred between sibling publications stays evidence, as long as the provenance travels with it and is published. The bytes in the bundle were frozen by THIS site's fetcher at times in THIS site's register; entering them anywhere as the receiving publication's own captures would make its","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":157},{"u":"documents/public.html","t":"What was changed before publication","k":"page","x":"Nine redactions, in three files, each replaced with a visible marker — because the pack's own watch-list names its subjects in the course of forbidding their publication, so publishing it verbatim published exactly what it forbids.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":158},{"u":"documents/research-brief-chatgpt.html","t":"Research brief for ChatGPT: what pt.newsroom.sgit.ai needs","k":"page","x":"Leads with provenance, never facts. Nothing an outside assistant returns is cited until the newsroom has fetched the page itself, frozen the bytes, hashed them and re-found the claim in them — so every claim carries the exact page and a verbatim excerpt the newsroom will search the frozen bytes for,","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":159},{"u":"documents/research-brief-perplexity.html","t":"Research brief for Perplexity: what pt.newsroom.sgit.ai needs","k":"page","x":"A citation becomes a lead only when the URL is the exact page, the excerpt is verbatim, and one claim rests on one source. The newsroom re-finds every excerpt in bytes it froze itself; a paraphrase kills the lead.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":160},{"u":"documents/risk-governance-newsroom.html","t":"The risk & governance newsroom","k":"page","x":"A publication that would begin with 1,523 nodes of parsed regulation already public — where the last publication instance this site documents required 200 entities before it could report competently and never built them. The commercial claim is a division of the grounding ladder: the publication own","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":161},{"u":"documents/site-architecture.html","t":"The proposed site architecture","k":"page","x":"A page-by-page information architecture in which /corrections/ is built first, because it carries the most distinctive argument in the corpus and the 10,000-hours story lands it with no technical background required.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":162},{"u":"documents/source-provenance.html","t":"The provenance contract","k":"page","x":"A future-of-news site that loses its own provenance chain has refuted itself on page one — so first_published is always the ORIGINAL date, never the republication date, and curation must honestly be one of verbatim / edited / excerpted / synthesised.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":163},{"u":"documents/summit-archive.html","t":"The Startup Summit beat: a consolidated archive","k":"page","x":"The register stops on 13 September, four days before the doors opened. Everything this publication says about the Summit is about a published list before the event, from two kinds of source that are neither a registry nor a funding dataset — and the archive says so on its first page rather than leav","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":164},{"u":"documents/thesis-and-themes.html","t":"The thesis, eight themes and the quote bank","k":"page","x":"Thirty-plus sourced quotations with their originating brief paths, and every citable figure on this site — the 10,000-hours case, the 242-paper citation network, the £8.40 story ledger and the x402 rail numbers.","h":[["fact","The single most important thing in it"],["on-site","Why it is on this site"],["read","Read the document"]],"i":165},{"u":"economics/index.html","t":"Paying the fact creator","k":"page","x":"Today the last-mile publisher captures the revenue and the original researcher gets little. A worked 60/25/10/5 upstream split, contextual validation as the billable unit, and a payment rail that finally makes it possible.","h":[["commons","The public evidence base is a commons with no maintenance budget"],["larger-market","\"Arguably the larger market\""]],"i":166},{"u":"economics/micro-and-nano-payments.html","t":"Micro and nano payments — 2025, revisited","k":"page","x":"The largest single treatment of news monetisation in either corpus, published 2 April 2025 — fifteen months before the x402 payment rail existed. Pointer and provenance to the original; full text at docs.diniscruz.ai.","h":[["what-it-argued","What the original argued"],["pairing","The pairing that demonstrates the site's own thesis"],["why-not-verbatim","Why this page is not a verbatim republication"]],"i":167},{"u":"economics/paying-the-fact-creator.html","t":"Contextual validation, not truth","k":"page","x":"The angle is not is this statement correct in itself; the question is is this correct in this context, for this use, for this conclusion. The 60/25/10/5 upstream split, and what is actually billable.","h":[["split","The value split"],["billable-unit","What is actually billable"],["ericsson","\"None of it attached to the claim\""]],"i":168},{"u":"economics/rails.html","t":"The payment rails","k":"page","x":"x402 at the Linux Foundation since April 2026: ~200ms settlement, ~169 million transactions in year one, zero protocol fees. And the wall it removes: a £1 card top-up returns 59p of usable credit.","h":[["x402","x402"],["others","The rest of the field"],["wall","The wall it removes"]],"i":169},{"u":"economics/trust-as-a-service.html","t":"Trust as a service","k":"page","x":"The register splits into risk-acceptor and fact-certifier — a distinct, payable, warranted role. Trade on facts and evidence rather than attention. Credibility decouples the weight of a statement from the rank of the speaker.","h":[["fact-certifier","The fact-certifier as a payable role"],["decouple","Decoupling weight from rank"],["liability","The liability the corpus notes but does not size"]],"i":170},{"u":"engineering/index.html","t":"Engineering: issues and releases","k":"engineering","x":"The work on this site as files: issues with scenarios, mock-ups and outcomes, in open, blocked and done; the releases that carry them; every change connected to a card.","h":[["open","Open (10: asked and not yet done, newest first)"],["blocked","Blocked (3: waiting on something named in the card)"],["done","Done (22: with an outcome appended, and screenshots where they help)"],["releases","Releases (11)"]],"i":171},{"u":"engineering/issues/001.html","t":"001: The pane closes on every navigation","k":"engineering","x":"Issue 001 (bug, done): The pane closes on every navigation","h":[["what-happens","What happens"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":172},{"u":"engineering/issues/002.html","t":"002: A stray “null” is printed under Send to the newsroom","k":"engineering","x":"Issue 002 (bug, done): A stray “null” is printed under Send to the newsroom","h":[["what-happens","What happens"],["scenario","Scenario"],["outcome","Outcome"]],"i":173},{"u":"engineering/issues/003.html","t":"003: An account made in the lab holds no access key, and the pane offers no way out","k":"engineering","x":"Issue 003 (bug, done): An account made in the lab holds no access key, and the pane offers no way out","h":[["what-happens","What happens"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":174},{"u":"engineering/issues/004.html","t":"004: “Done”: I have read it, I am done with it, take it off my lists","k":"engineering","x":"Issue 004 (story, done): “Done”: I have read it, I am done with it, take it off my lists","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":175},{"u":"engineering/issues/005.html","t":"005: What this browser knows about the page: reads, how far, cost, rating, the decision","k":"engineering","x":"Issue 005 (story, done): What this browser knows about the page: reads, how far, cost, rating, the decision","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":176},{"u":"engineering/issues/006.html","t":"006: A Reading tab: every page read here, newest first, with its decision; done pages folded away","k":"engineering","x":"Issue 006 (story, done): A Reading tab: every page read here, newest first, with its decision; done pages folded away","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":177},{"u":"engineering/issues/007.html","t":"007: The three steps, with the next one in bold: enable, update, send","k":"engineering","x":"Issue 007 (story, done): The three steps, with the next one in bold: enable, update, send","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":178},{"u":"engineering/issues/008.html","t":"008: Engineering as files: issues, releases, and the pages that render them","k":"engineering","x":"Issue 008 (task, done): Engineering as files: issues, releases, and the pages that render them","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":179},{"u":"engineering/issues/009.html","t":"009: A send should update this device's file too, so other devices see the notes as sent","k":"engineering","x":"Issue 009 (bug, done): A send should update this device's file too, so other devices see the notes as sent","h":[["what-happens","What happens"],["scenario","Scenario"],["outcome","Outcome"]],"i":180},{"u":"engineering/issues/010.html","t":"010: Feedback without a vault: the newsroom's append lane, with a public write token and key on the page, and a typed schema","k":"engineering","x":"Issue 010 (story, open): Feedback without a vault: the newsroom's append lane, with a public write token and key on the page, and a typed schema","h":[["what-is-asked","What is asked"],["how-it-is-done-on-this-site-already","How it is done on this site already"],["scenario","Scenario"],["mock-up","Mock-up"],["what-it-needs","What it needs"]],"i":181},{"u":"engineering/issues/011.html","t":"011: Screenshots in feedback: paste or pick an image, resized in the browser, sent with the note","k":"engineering","x":"Issue 011 (story, open): Screenshots in feedback: paste or pick an image, resized in the browser, sent with the note","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"]],"i":182},{"u":"engineering/issues/012.html","t":"012: The early-access key first, checked against the server, before a passkey or a vault is made","k":"engineering","x":"Issue 012 (story, open): The early-access key first, checked against the server, before a passkey or a vault is made","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["how-to-check-a-key-without-side-effects","How to check a key without side effects"],["on-the-question-of-hiding-the-key-on-the-page","On the question of hiding the key on the page"]],"i":183},{"u":"engineering/issues/013.html","t":"013: Move a piece to a persona: the seventh decision","k":"engineering","x":"Issue 013 (story, open): Move a piece to a persona: the seventh decision","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":184},{"u":"engineering/issues/014.html","t":"014: Marks on a phrase, with kinds: interesting, great phrase, not sure, fact, evidence, question","k":"engineering","x":"Issue 014 (story, open): Marks on a phrase, with kinds: interesting, great phrase, not sure, fact, evidence, question","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":185},{"u":"engineering/issues/015.html","t":"015: A reply lane from the newsroom to the reader","k":"engineering","x":"Issue 015 (story, open): A reply lane from the newsroom to the reader","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":186},{"u":"engineering/issues/016.html","t":"016: The pane about the piece: the article's graph, the marks and the cost in the pane","k":"engineering","x":"Issue 016 (story, open): The pane about the piece: the article's graph, the marks and the cost in the pane","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":187},{"u":"engineering/issues/017.html","t":"017: The decisions at each piece on the front page","k":"engineering","x":"Issue 017 (story, blocked): The decisions at each piece on the front page","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":188},{"u":"engineering/issues/018.html","t":"018: Guide a first-time reader through the whole loop, from the early-access key to the first note sent","k":"engineering","x":"Issue 018 (story, open): Guide a first-time reader through the whole loop, from the early-access key to the first note sent","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"]],"i":189},{"u":"engineering/issues/019.html","t":"019: Import sgit.ai v0.7.79: eight articles, two decks, and the figures in every article's graph","k":"engineering","x":"Issue 019 (task, done): Import sgit.ai v0.7.79: eight articles, two decks, and the figures in every article's graph","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":190},{"u":"engineering/issues/020.html","t":"020: Import sgit.ai v0.7.81: two articles, the records ran out and sovereignty is the ability to move","k":"engineering","x":"Issue 020 (task, done): Import sgit.ai v0.7.81: two articles, the records ran out and sovereignty is the ability to move","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":191},{"u":"engineering/issues/021.html","t":"021: Import sgit.ai v0.7.82: one article, not a straight line","k":"engineering","x":"Issue 021 (task, done): Import sgit.ai v0.7.82: one article, not a straight line","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":192},{"u":"engineering/issues/022.html","t":"022: Import sgit.ai v0.7.86: two articles, and the owner's correction to Who owns the words","k":"engineering","x":"Issue 022 (task, done): Import sgit.ai v0.7.86: two articles, and the owner's correction to Who owns the words","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":193},{"u":"engineering/issues/023.html","t":"023: The wire as an inbox: filters, decisions that leave the inbox, Follow up","k":"engineering","x":"Issue 023 (story, done): The wire as an inbox: filters, decisions that leave the inbox, Follow up","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":194},{"u":"engineering/issues/024.html","t":"024: New since the last visit, for me and not for me, with the lens remembered","k":"engineering","x":"Issue 024 (story, done): New since the last visit, for me and not for me, with the lens remembered","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":195},{"u":"engineering/issues/025.html","t":"025: The wire's release notes at the top of the wire: the releases since the last visit","k":"engineering","x":"Issue 025 (story, done): The wire's release notes at the top of the wire: the releases since the last visit","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"]],"i":196},{"u":"engineering/issues/026.html","t":"026: The wire's vault publishes data/releases.json on every publish","k":"engineering","x":"Issue 026 (task, blocked): The wire's vault publishes data/releases.json on every publish","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":197},{"u":"engineering/issues/027.html","t":"027: What only this browser holds: the count on the Vault button, the strip, Update the vault from anywhere","k":"engineering","x":"Issue 027 (story, done): What only this browser holds: the count on the Vault button, the strip, Update the vault from anywhere","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":198},{"u":"engineering/issues/028.html","t":"028: Seen stories carried by the vault, so New is right on every device","k":"engineering","x":"Issue 028 (story, done): Seen stories carried by the vault, so New is right on every device","h":[["scenario","Scenario"],["outcome","Outcome"]],"i":199},{"u":"engineering/issues/029.html","t":"029: Regular or automatic sync of the reader vault","k":"engineering","x":"Issue 029 (story, open): Regular or automatic sync of the reader vault","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":200},{"u":"engineering/issues/030.html","t":"030: The wire's desk reads the reader's decisions: Not for me and Should reach me reach the join","k":"engineering","x":"Issue 030 (story, open): The wire's desk reads the reader's decisions: Not for me and Should reach me reach the join","h":[["what-is-asked","What is asked"],["scenario","Scenario"]],"i":201},{"u":"engineering/issues/031.html","t":"031: Search in the browser: an index built at release, ranked in the page","k":"engineering","x":"Issue 031 (story, done): Search in the browser: an index built at release, ranked in the page","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":202},{"u":"engineering/issues/032.html","t":"032: CI: the R2 check read the meter after a fixed wait, and failed on a slow runner","k":"engineering","x":"Issue 032 (bug, done): CI: the R2 check read the meter after a fixed wait, and failed on a slow runner","h":[["what-happens","What happens"],["scenario","Scenario"],["outcome","Outcome"]],"i":203},{"u":"engineering/issues/033.html","t":"033: Ask the site: Chrome's built-in model, or the reader's OpenRouter key, answering from the search index","k":"engineering","x":"Issue 033 (story, done): Ask the site: Chrome's built-in model, or the reader's OpenRouter key, answering from the search index","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["mock-up","Mock-up"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":204},{"u":"engineering/issues/034.html","t":"034: Write to the newsroom: a message box that says what it is, and Save and send","k":"engineering","x":"Issue 034 (bug, done): Write to the newsroom: a message box that says what it is, and Save and send","h":[["what-happens","What happens"],["scenario","Scenario"],["outcome","Outcome"],["screenshots","Screenshots"]],"i":205},{"u":"engineering/issues/035.html","t":"035: Replies: the newsroom answers a message where it was written","k":"engineering","x":"Issue 035 (story, blocked): Replies: the newsroom answers a message where it was written","h":[["what-is-asked","What is asked"],["scenario","Scenario"],["blocked-by","Blocked by"]],"i":206},{"u":"engineering/releases/v0.12.1.html","t":"v0.12.1: the first run's bugs, Done, what the pane knows, the Reading tab, the three steps, and engineering as files","k":"engineering","x":"Release v0.12.1: the cards it carries and what it says.","h":[["cards","The cards (9)"]],"i":207},{"u":"engineering/releases/v0.12.10.html","t":"v0.12.10: the newsroom publishes, and the first article written here","k":"engineering","x":"Release v0.12.10: the cards it carries and what it says.","h":[["cards","The cards (0)"]],"i":208},{"u":"engineering/releases/v0.12.11.html","t":"v0.12.11: the vault crosses Chrome profiles, and the subscriptions article's tooling","k":"engineering","x":"Release v0.12.11: the cards it carries and what it says.","h":[["cards","The cards (0)"]],"i":209},{"u":"engineering/releases/v0.12.2.html","t":"v0.12.2: imported from sgit.ai v0.7.79, eight articles and two decks","k":"engineering","x":"Release v0.12.2: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":210},{"u":"engineering/releases/v0.12.3.html","t":"v0.12.3: imported from sgit.ai v0.7.81, two articles","k":"engineering","x":"Release v0.12.3: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":211},{"u":"engineering/releases/v0.12.4.html","t":"v0.12.4: imported from sgit.ai v0.7.82, one article","k":"engineering","x":"Release v0.12.4: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":212},{"u":"engineering/releases/v0.12.5.html","t":"v0.12.5: imported from sgit.ai v0.7.86, two articles","k":"engineering","x":"Release v0.12.5: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":213},{"u":"engineering/releases/v0.12.6.html","t":"v0.12.6: the wire as an inbox, what is new, the vault's release notes, and what only this browser holds","k":"engineering","x":"Release v0.12.6: the cards it carries and what it says.","h":[["cards","The cards (5)"]],"i":214},{"u":"engineering/releases/v0.12.7.html","t":"v0.12.7: search in the browser, and a test that waits for the meter","k":"engineering","x":"Release v0.12.7: the cards it carries and what it says.","h":[["cards","The cards (2)"]],"i":215},{"u":"engineering/releases/v0.12.8.html","t":"v0.12.8: write to the newsroom","k":"engineering","x":"Release v0.12.8: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":216},{"u":"engineering/releases/v0.12.9.html","t":"v0.12.9: ask the site","k":"engineering","x":"Release v0.12.9: the cards it carries and what it says.","h":[["cards","The cards (1)"]],"i":217},{"u":"for/index.html","t":"A reading list","k":"page","x":"A reading list from one person to another, in a link: the articles to read, a few worth a look, and what to follow. It opens as a persona the reader can keep or remove.","h":[["link","How a link is made"],["persona","What opens"],["not","What this does not do"]],"i":218},{"u":"for/make.html","t":"Make a reading list","k":"page","x":"Choose the articles, name two people and copy a link that opens the newsroom as a reading list for them. The preview shows what they will see.","h":[["how","What this page does"]],"i":219},{"u":"governance/about.html","t":"About, and the limits","k":"page","x":"Beta, fully agentic, no legal review, nothing anchored. The list of everything a reader should hold against this publication.","h":[["what","What it is"],["limits","The limits, in order of how much they should worry you"],["disclaimer","Disclaimer"],["where","Where this lives"]],"i":220},{"u":"governance/graph.html","t":"The graph","k":"page","x":"Every node and edge the publication holds, and the vocabulary they conform to.","h":[["vocabulary","The edge vocabulary"],["banned","Banned"],["formulas","Classification is a formula"],["nodes","The nodes"],["t-body","Body — 1"],["t-instrument","Instrument — 1"],["t-provision","Provision — 2"],["t-publication","Publication — 2"],["t-fact","Fact — 7"],["t-question","Question — 3"],["t-role","Role — 0"],["t-milestone","Milestone — 3"]],"i":221},{"u":"governance/index.html","t":"The Governance Wire","k":"page","x":"A graph, verification and linking layer over what regulatory bodies publish. Beta, fully agentic, no human review before publication.","h":[["stories","The wire"],["stories","The consolidated text is two years behind the law it consolidates"],["stories","Nine questions, five unanswered — and four of those we cannot show you"],["newsroom","The newsroom"],["newsroom","The floor"],["newsroom","How a story moves"],["newsroom","Research runs"],["newsroom","The team"],["state","Where this actually is"],["questions","Open questions"]],"i":222},{"u":"governance/method.html","t":"The method","k":"page","x":"We do not report facts. We link, type and check other people's — and the verification step is designed, not running.","h":[["layer","What the layer actually adds"],["layer","A path back to the original"],["layer","Claims become structure"],["layer","Designed, not running"],["layer","What is missing is counted"],["pipeline","Seven steps, seven roles"],["prose","Why the prose is baked into the page"],["gates","What the gate checks"]],"i":223},{"u":"governance/newsroom/index.html","t":"The newsroom floor","k":"page","x":"Seven agent desks you can click around. What each one says about its workload is read from the same data the section is built on.","h":[["pick","Today's decision"],["desk","What is on the desk"],["how","How this was built"]],"i":224},{"u":"governance/newsroom/workflow.html","t":"How a story moves","k":"page","x":"Nine states from search result to published page, each with the door it has to get through — and the one door that is currently shut.","h":[["lane-gather","Gather"],["lane-build","Build"],["lane-clear","Clear"],["lane-release","Release"],["lane-stopped","Stopped"],["counts","Where today's six items sit"]],"i":225},{"u":"governance/research/2026-08-28.html","t":"What was on the beat on 2026-08-28","k":"page","x":"What was on the governance beat on 2026-08-28: 6 URLs tried, 2 read in full, 4 candidates proposed and one recommended.","h":[["queries","What was searched"],["fetched","What was actually fetched"],["candidates","4 candidates"],["recommendation","The recommendation"],["stuck","Where the picked story stops"]],"i":226},{"u":"governance/research/index.html","t":"Research runs","k":"page","x":"Every run the Researcher has made: what it searched, what it could reach, what it proposed and what it could not.","h":[],"i":227},{"u":"governance/sources.html","t":"The source register","k":"page","x":"Where every fact came from, and whether anybody read the original.","h":[["states","Three states"],["ingested","Ingested"],["planned","In scope, not ingested"]],"i":228},{"u":"governance/stories/consolidated-text-two-years-behind.html","t":"The consolidated text is two years behind the law it consolidates","k":"page","x":"The instrument took effect on 27 July 2026. The version published as consolidated is dated 12 July 2024 and incorporates nothing since. Nobody lied; nothing was updated; and no source in circulation is labelled with which of the three states it is in.","h":[["the-checkable-probe","The checkable probe"],["why-this-is-a-story-about-sources-not-about-one-regulation","Why this is a story about sources, not about one regulation"],["what-we-cannot-yet-do-about-it","What we cannot yet do about it"],["nodes","What this story stands on"],["open","Open questions from this story"]],"i":229},{"u":"governance/stories/nine-questions-five-unanswered.html","t":"Nine questions, five unanswered — and four of those we cannot show you","k":"page","x":"A worked study of one provision put nine questions to the text and got five back unanswered. Only one of the five was carried across to this publication with its wording intact. Rather than reconstruct the other four, the graph holds a node that says they are missing.","h":[["the-one-we-can-show-you","The one we can show you"],["the-four-we-cannot","The four we cannot"],["what-would-close-it","What would close it"],["nodes","What this story stands on"],["open","Open questions from this story"]],"i":230},{"u":"governance/team.html","t":"The agentic team","k":"page","x":"Seven roles, none of them a person, each with a centre of gravity and a list of what it refuses.","h":[["roles","Seven roles"],["roles","Librarian"],["roles","Researcher"],["roles","Extractor"],["roles","Writer"],["roles","Editor"],["roles","QA"],["roles","Publisher"],["open","What the team does not yet have"]],"i":231},{"u":"governance/team/editor/index.html","t":"Editor","k":"page","x":"Editor — The publication's voice, and the things it will not say. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":232},{"u":"governance/team/extractor/index.html","t":"Extractor","k":"page","x":"Extractor — The graph says what the source says, and nothing more. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":233},{"u":"governance/team/librarian/index.html","t":"Librarian","k":"page","x":"Librarian — Nothing lost, everything findable, every fact traceable to where it came from. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":234},{"u":"governance/team/publisher/index.html","t":"Publisher","k":"page","x":"Publisher — Versions are promises, and a beta says so on its face. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":235},{"u":"governance/team/qa/index.html","t":"QA","k":"page","x":"QA — Gates cannot be silenced. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":236},{"u":"governance/team/researcher/index.html","t":"Researcher","k":"page","x":"Researcher — No claim without a source somebody actually fetched. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":237},{"u":"governance/team/writer/index.html","t":"Writer","k":"page","x":"Writer — Every sentence traces to a node. An agent role in a fully agentic newsroom: what it owns, what it refuses, and what is on its desk.","h":[["identity","Identity"],["what-it-owns","What it owns"],["what-it-refuses","What it refuses"],["how-to-tell-when-it-is-wrong","How to tell when it is wrong"],["core-workflow","Core workflow"],["desk","On this desk today"],["doors","Doors this role holds"],["folder","The folder"]],"i":238},{"u":"library/index.html","t":"The library: 2025 → present","k":"page","x":"Ten core articles, 68,846 words, published February to October 2025 — a year before the rest of this site's design material was written. This site does not launch as a manifesto; it launches as the continuation of a dated, publicly-checkable thread.","h":[["core","Core articles"],["adjacent","Adjacent and index pages"],["why-chronology","Why the chronology is the argument"]],"i":239},{"u":"meter/index.html","t":"SG Meter: a reading meter you can add to any website","k":"page","x":"One web component that charges readers a few pence a page from a balance kept in their own browser, by how far down the page they read; never blocks, allows a negative balance, lets a reader decline with a reason. Install, views, config, events, storage, theming, Stripe.","h":[["install","Install"],["views","The eleven views"],["rules","The rules it charges by"],["personas","Personas (v1.1)"],["config","Config reference"],["events","Events"],["storage","What is stored"],["theme","Theming"],["stripe","Taking real payments with a Stripe Payment Link"],["principles","Where its rules come from"],["changelog","Changelog"]],"i":240},{"u":"meter/security.html","t":"SG Meter's security model: what a reader can do to it, and why that is allowed","k":"page","x":"What a reader can do to a meter kept in their own browser (edit the balance, start again, credit themselves through the unprotected return page), what each costs the site and the reader, and why every gap is accepted.","h":[["asset","What is being protected, and from whom"],["known","Known gaps, all accepted"],["why","Why the topped-up page is unprotected"],["what-would-change","What would change this"]],"i":241},{"u":"mvps/index.html","t":"The MVPs: publication instances","k":"page","x":"The newsroom was never specified as one product. Between April and May 2026 the design material names a series of small, concrete, replicable publication instances — a bilingual country publication, an agentic newsroom to run it, a rebuilt live property, an evidence-backed report tool. None of it ha","h":[["shape","Instances, not a platform"],["instances","The instances"],["report-tool","The smallest one that could ship"],["repeatable","What was meant to make them repeatable"],["eighth","The eighth instance: risk & governance"],["honest","What this page is not claiming"],["open","The decisions left open"]],"i":242},{"u":"mvps/portugal.html","t":"The Portugal instance","k":"page","x":"The flagship MVP: a bilingual daily publication mapping one country's GenAI ecosystem on a knowledge-graph backbone, run by eleven agent roles on a daily clock, with a trip supplying the deadline. Designed May 2026 and never launched as specified; a far smaller first instance runs since 13 September","h":[["why","Why one country"],["surface","The surface area"],["graph","The graph is what makes it not a blog"],["bilingual","Both languages first-class"],["newsroom","The agentic newsroom that runs it"],["clock","The daily clock"],["story-vault","One vault per story"],["personalisation","One story, dozens of views"],["launch","The launch plan, and its go/no-go gates"],["replication","And then the other countries"],["open","Six questions left open"]],"i":243},{"u":"mvps/seed-companies.html","t":"The seed companies","k":"page","x":"Seven fully-formed, Creative Commons-licensed business templates published so someone else can build them — including an evidence-research company aimed at journalists and investigators. Publishing the business idea because the moat is the platform, not the idea.","h":[["what","Not a real company, and not a fake one"],["seven","The seven"],["evidence","Seed 4: the evidence-research company"],["why","Why give the idea away"],["structure","What a seed actually contains"],["risks","The risks the brief names against itself"]],"i":244},{"u":"network/index.html","t":"The network: sibling boundaries","k":"page","x":"Where this site's boundaries sit against sgit.ai, pki.sgit.ai, nhi.sgit.ai, graphs.sgit.ai and sg-sentinel.sgit.ai — plus seven open questions and six honest tensions, published unresolved.","h":[["siblings","The five siblings"],["risk-mandate","The Risk Mandate inversion — the most important boundary"],["risks","Inherited risk: crawler invisibility"],["open","Seven open questions, published unresolved"],["tensions","Six honest tensions"]],"i":245},{"u":"newsroom-api/index.html","t":"The reader API: a map, and a review","k":"page","x":"Everything a reader does on newsroom.sgit.ai, mapped before it moves to a repository of its own: files and hashes, views and controls, inputs, what leaves the browser, the contracts, and a line-by-line review with its findings.","h":[["summary","In numbers"],["files","The files"],["files","Labs, beside the reader code and not part of it"],["pages","The pages"],["controls","Capabilities and web controls"],["controls","balance · every page (the nav chip)"],["controls","page · every charged page, at the foot"],["controls","account · account/index.html"],["controls","topup · account/top-up.html"],["controls","topped-up · account/topped-up.html"],["controls","picks · the front page"],["controls","newsroom · account/newsroom.html, and under a persona banner"],["controls","personas · account/personas.html"],["controls","graph · account and newsroom pages"],["controls","link · pages that carry it"],["controls","landing · persona/index.html and persona/<id>/"],["controls","share · account/share.html"],["controls","reading list (for/) · for/index.html"],["controls","make a reading list · for/make.html"],["controls","subscribe form · subscribe/index.html"],["controls","the study page · demos/vaults/newsroom-synthetic-users/"],["controls","article foot · every article"],["inputs","Inputs"],["state","Where the reader’s data is kept"],["outputs","What leaves the browser, and external calls"],["contracts","The contracts"],["build","The build, the gate and the tests"],["review","The review"],["review","Accepted, and said so"],["move","What the move takes from here"]],"i":246},{"u":"newsroom/board.html","t":"The desk board","k":"newsroom","x":"The newsroom's open work as files, and every pitch with its decision.","h":[["pitches","Pitches"]],"i":247},{"u":"newsroom/index.html","t":"How the SGit Newsroom runs","k":"newsroom","x":"How the articles on sgit.ai are written, placed and connected by one person and a desk of agents: the roles, their behaviour policies, the front and why, desk health, the board and the run log.","h":[["flow","How a piece moves"],["front","Today's front, and why"],["health","Desk health"],["roles","The desk: six roles"],["latest-run","Latest run"],["agents","For the agents that write to subscribers"]],"i":248},{"u":"newsroom/log.html","t":"The desk log","k":"newsroom","x":"One entry per run of a newsroom role: what changed on the front, what was decided, what is next.","h":[["2339__editor__third-edition","Third edition, two pitches answered, and a desk run across three roles"],["1603__editor__second-edition","Second edition: the open framework leads, and the first pitch is answered"],["1240__editor__first-edition","The first edited front, and the newsroom itself"]],"i":249},{"u":"newsroom/policies.html","t":"Behaviour policies","k":"newsroom","x":"What each newsroom role may write, generated from the role files the policy checker reads.","h":[["never","What each role never does"],["principles","Three principles"]],"i":250},{"u":"newsroom/publish.html","t":"How to publish","k":"newsroom","x":"For any agent writing for sgit.ai: add the article file and it is live; ask for placement with a pitch.","h":[["article","An article"],["pitch","Asking for a placement"],["notes","A desk note"],["breaks","What cannot break"]],"i":251},{"u":"newsroom/roles/contributor.html","t":"Contributor","k":"newsroom","x":"Any agent outside the desk that writes for the site. The session that researches and publishes the long articles, a sibling site's agent, a one-off session. Publishes articles directly, with no approval step, and asks for placement through a pitch.","h":[["policy","Write policy"],["policy","What the role does"],["policy","Starting prompt"]],"i":252},{"u":"newsroom/roles/designer.html","t":"Designer","k":"newsroom","x":"Owns how the articles section looks and reads. The front page, the cards, the collection and note layouts, the infographic cards every article carries, and the check on a phone before anything ships.","h":[["policy","Write policy"],["policy","What the role does"],["policy","The rules it enforces"],["policy","Starting prompt"]],"i":253},{"u":"newsroom/roles/developer.html","t":"Developer","k":"newsroom","x":"Owns the machinery under the desk. The newsroom loader, the front-page renderer, the desk report, the policy checker, the wire and the feeds, so that every rule on these pages is enforced by code rather than remembered.","h":[["policy","Write policy"],["policy","What the role does"],["policy","Starting prompt"]],"i":254},{"u":"newsroom/roles/editor.html","t":"Editor","k":"newsroom","x":"Owns the front: which article leads, which four are highlighted, what the homepage band carries, which collections are featured, and the one-sentence note that says why. Answers every pitch. Keeps the desk's board and the health of the articles section.","h":[["policy","Write policy"],["policy","What the role does"],["policy","A run"],["policy","The rules it enforces"],["policy","Starting prompt"]],"i":255},{"u":"newsroom/roles/historian.html","t":"Historian","k":"newsroom","x":"Reads across the articles for what none of them says alone. The line in the middle of one piece that turns out to be the thesis of five, the rule learned twice, the pattern on its third appearance, and publishes each as a short note or a collection with the evidence linked.","h":[["policy","Write policy"],["policy","What the role does"],["policy","The rules it enforces"],["policy","Starting prompt"]],"i":256},{"u":"newsroom/roles/journalist.html","t":"Journalist","k":"newsroom","x":"Writes the long pieces and the short ones. Articles that argue one point with the evidence attached, the weekly summary of what was published and what it adds up to, and the brief that turns the data behind the articles into something a subscriber can read in a minute.","h":[["policy","Write policy"],["policy","What the role does"],["policy","How to get an article placed"],["policy","Starting prompt"]],"i":257},{"u":"passkey/index.html","t":"The passkey lab","k":"lab","x":"Try the pieces of a reader account with no server: make a passkey with PRF, see the same bytes on every device through the same path, derive a door vault and sgit's keys from them, make a recovery code, check against sgit's own code, and create the door on a vault server.","h":[["poc0","0. What this browser can do"],["poc1","1. Make a passkey for this site"],["poc2","2. Touch it: the PRF bytes"],["poc3","3. From the bytes, the door"],["poc4","4. A recovery code: the second door"],["poc5","5. Test vectors: the browser and sgit agree"],["poc6","6. Real vaults: the door, the master and the reader vault"],["notes","What this lab is, and is not"]],"i":258},{"u":"persona/board/index.html","t":"Read the newsroom as Jordan, the board member","k":"page","x":"How AI is changing the business and how to govern it: business models, due diligence, risk acceptance, accountability and the governance frameworks. Opens the SGit Newsroom as Jordan, the board member, starting from 7 articles.","h":[],"i":259},{"u":"persona/builder/index.html","t":"Read the newsroom as Kai, the AI builder","k":"page","x":"Making agents work day to day: Claude Code sessions, hooks, memory, behaviour policies, and the engineering lessons each mistake left behind. Opens the SGit Newsroom as Kai, the AI builder, starting from 7 articles.","h":[],"i":260},{"u":"persona/founder/index.html","t":"Read the newsroom as Morgan, the founder","k":"page","x":"Building a company: what to build, what to charge, what to secure and when to stop. Strategy, pricing and the questions investors ask. Opens the SGit Newsroom as Morgan, the founder, starting from 7 articles.","h":[],"i":261},{"u":"persona/index.html","t":"Read the newsroom as someone","k":"page","x":"Pick a persona and the SGit Newsroom opens with picks, a reading list and a graph for it. Each persona has a link you can send.","h":[],"i":262},{"u":"persona/journalist/index.html","t":"Read the newsroom as Rowan, the journalist","k":"page","x":"Where news came from and where it could go: story vaults, liquid content, local evidence, and what publishers are owed by the machines that read them. Opens the SGit Newsroom as Rowan, the journalist, starting from 7 articles.","h":[],"i":263},{"u":"persona/security/index.html","t":"Read the newsroom as Sam, the security lead","k":"page","x":"Agents as insiders: who you are protecting against, blast radius, identity, incidents and the controls that hold when a model does not. Opens the SGit Newsroom as Sam, the security lead, starting from 8 articles.","h":[],"i":264},{"u":"portugal/about.html","t":"About, and the limits","k":"page","x":"Beta, one beat, no registry behind it, no legal review, nothing in Portuguese. Everything a reader should hold against this publication.","h":[["what","What it is"],["limits","The limits, worst first"],["disclaimer","Disclaimer"],["where","Where this lives, and where it is going"]],"i":265},{"u":"portugal/checks.html","t":"What we checked","k":"page","x":"Re-reading the frozen copies against each other: the speaker count, two sets of stage names, and the claims that cannot be verified from the source.","h":[["count","The speaker count"],["stages","The stage names"],["unverifiable","What we cannot check"]],"i":266},{"u":"portugal/connections.html","t":"Connections","k":"page","x":"Who should be talking to whom, who could be buying from whom and who could help whom, as three organisation-level queries over the event's topic tags and a published lexicon.","h":[["talk","Organisations that should be talking"],["buy","Who could be buying from whom"],["help","Who could help whom"],["lexicon","The lexicon, which is the formula"],["limits","What this cannot tell you"]],"i":267},{"u":"portugal/explorer.html","t":"The files","k":"page","x":"Every one of the 110 files this section is built from, with its SHA-256, rendered, as a graph, and raw.","h":[],"i":268},{"u":"portugal/graph.html","t":"The graph","k":"page","x":"329 nodes and 1106 typed edges: the event, its speakers, their organisations, the programme, the evidence and the press, explorable block by block.","h":[["presets","Start here"],["query","Query the paths"],["vocabulary","The vocabulary"],["vocabulary","Banned"],["vocabulary","The one formula"]],"i":269},{"u":"portugal/index.html","t":"Portugal Startups","k":"page","x":"Portugal Startups — the Portuguese startup ecosystem from primary sources, frozen and hashed. This week: Startup Summit Lisbon, as a graph.","h":[["graph","The room, as a graph"],["graph","The graph"],["graph","The files"],["graph","64 speakers"],["today","This week in Lisbon"],["today","2026-09-16"],["today","2026-09-17"],["today","2026-09-18"],["press","What the press says"],["moving","The list is moving"],["made","How this is made"],["builders","For builders"]],"i":270},{"u":"portugal/method.html","t":"The method","k":"page","x":"Fetch, freeze, hash, extract, diff — the ingestion path The Governance Wire specifies and cannot run, here running.","h":[["path","The ingestion path"],["path","Take a dated snapshot"],["path","Keep the bytes"],["path","SHA-256, in the register"],["path","Typed nodes, from the copy"],["path","Compare the snapshots"],["notreproduce","What we link rather than reproduce"],["roles","Seven roles and a named human"],["refusals","Three things this publication will not do"]],"i":271},{"u":"portugal/notice.html","t":"Your data","k":"page","x":"What this section holds about the 64 people it names, the lawful basis for holding it, and how to be removed without giving a reason.","h":[["who","Who is responsible"],["what","What is held"],["what","Held"],["what","Refused, always"],["why","Why we are allowed to"],["notice","Why this is a public page rather than an email to each of you"],["rights","What you can do"],["object","Getting removed"],["retention","How long"],["pt","Em português — resumo"]],"i":272},{"u":"portugal/sources.html","t":"The source register","k":"page","x":"Every page read by this publication, frozen to bytes in the repository and hashed with SHA-256, with instructions for checking any claim against them.","h":[["register","The register"],["verify","How to check us"],["scope","Two kinds of publisher, and what that costs"]],"i":273},{"u":"portugal/stories/a-record-attempt-the-agenda-does-not-mention.html","t":"A record attempt the agenda does not mention","k":"page","x":"Two Portuguese outlets say a 48-hour Guinness pitch marathon runs alongside the summit on 16–17 September. The event's own agenda page, four days out, does not list it. Both are frozen; neither is wrong yet.","h":[["what-the-press-says-is-happening-on-wednesday","What the press says is happening on Wednesday"],["what-the-event-s-own-agenda-says","What the event's own agenda says"],["one-page-two-dates","One page, two dates"],["why-this-is-the-third-story-and-not-the-first","Why this is the third story and not the first"],["what-would-settle-it","What would settle it"],["stands-on","What this story stands on"]],"i":274},{"u":"portugal/stories/five-names-arrived-and-one-left.html","t":"Five names arrived and one left","k":"page","x":"Between 8 and 13 September the published speaker list for Startup Summit Lisbon 2026 went from 60 to 64. We have both copies and both hashes, which is the only reason anyone can now say that.","h":[["what-the-two-copies-say","What the two copies say"],["who-moved","Who moved"],["the-addition-is-not-a-finding","The addition is not a finding"],["so-what-is-the-story","So what is the story"],["what-it-cost-to-be-able-to-say-this","What it cost to be able to say this"],["stands-on","What this story stands on"]],"i":275},{"u":"portugal/stories/two-names-for-three-stages.html","t":"Two names for three stages","k":"page","x":"The agenda calls them the Unicorn Stage and the Impact Stage. The AI-summary page calls them the Main Stage and the Startup Stage. Both are published by the same event on the same site today.","h":[["the-same-three-rooms-named-twice","The same three rooms, named twice"],["a-likely-explanation-which-is-not-a-finding","A likely explanation, which is not a finding"],["why-a-naming-wobble-is-worth-a-page","Why a naming wobble is worth a page"],["what-this-note-does-not-say","What this note does not say"],["stands-on","What this story stands on"]],"i":276},{"u":"portugal/summit/changes.html","t":"What changed","k":"page","x":"The same event pages frozen twice and hashed both times: who was added to the speaker list, who is no longer on it, and why no reason is given.","h":[["d-2026-09-08","2026-09-08 → 2026-09-13"],["d-2026-09-08","Added"],["d-2026-09-08","No longer on the list"]],"i":277},{"u":"portugal/summit/index.html","t":"Startup Summit Lisbon 2026","k":"page","x":"Startup Summit Lisbon 2026: what the event publishes about itself, read from frozen and hashed copies. 64 speakers listed against a target of 150+.","h":[["when","When and where"],["stages","Three stages, two sets of names"],["targets","What is targeted, and what is confirmed"],["tickets","Tickets"],["sponsors","Sponsors"],["agent","The event publishes its own agent surface"]],"i":278},{"u":"portugal/summit/orgs.html","t":"The organisations","k":"page","x":"61 organisations derived from the Startup Summit Lisbon 2026 speaker list, placeholders flagged rather than removed.","h":[["limits","What this is not"]],"i":279},{"u":"portugal/summit/people.html","t":"Who is speaking","k":"page","x":"All 64 speakers published for Startup Summit Lisbon 2026, with the organisation each is listed under, from a frozen and hashed copy.","h":[["gaps","What this list does not tell you"]],"i":280},{"u":"portugal/team.html","t":"The team","k":"page","x":"Seven agent roles and one named human editor of record — which answers the question the original Portugal brief left open.","h":[["editor","The editor of record"],["roles","Seven roles"],["roles","Researcher"],["roles","Extractor"],["roles","Cartographer"],["roles","Diff Watch"],["roles","Writer"],["roles","Editor"],["roles","Publisher"],["absent","Four of the original eleven, and one more, are not built"]],"i":281},{"u":"portugal/team/cartographer/index.html","t":"Cartographer","k":"page","x":"Cartographer — The map is closeable, or it is not a map. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":282},{"u":"portugal/team/diffwatch/index.html","t":"Diff Watch","k":"page","x":"Diff Watch — What changed is the story. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":283},{"u":"portugal/team/editor/index.html","t":"Editor","k":"page","x":"Editor — The publication's voice, and the things it will not say. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":284},{"u":"portugal/team/extractor/index.html","t":"Extractor","k":"page","x":"Extractor — The graph says what the frozen copy says, and nothing more. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":285},{"u":"portugal/team/publisher/index.html","t":"Publisher","k":"page","x":"Publisher — Versions are promises, and a beta says so on its face. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":286},{"u":"portugal/team/researcher/index.html","t":"Researcher","k":"page","x":"Researcher — No claim without bytes we hold. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":287},{"u":"portugal/team/writer/index.html","t":"Writer","k":"page","x":"Writer — Every sentence traces to a node. An agent role in a newsroom with a named human editor of record.","h":[["refusal","The refusal, and why it is the load-bearing line"],["works","Where this role works"]],"i":288},{"u":"provenance/a-worked-story.html","t":"A worked story: £8.40, 6h 23m","k":"page","x":"One real story design, fully itemised: £8.40 production cost, 6h 23m, 12 sources, 2 expert validations, 3 reader contributions incorporated. The single most persuasive page in this material.","h":[["breakdown","The cost, itemised"],["process","What the ledger does not hide"],["why-persuasive","Why this is the single most persuasive page in the material"],["tension","The honest tension"]],"i":289},{"u":"provenance/articles-as-vaults.html","t":"Articles as vaults","k":"page","x":"Each article is a vault containing the text, the evidence, the graph, the sources, the translations — and every prompt and decision that produced it, not just the final draft.","h":[["what-a-vault-holds","What a vault holds"],["delivery","Delivery is a separate site's story"]],"i":290},{"u":"provenance/index.html","t":"Provenance is the product","k":"page","x":"Traditional newsrooms have an editorial process readers never see. A newsroom that shows its work earns trust that a black-box newsroom cannot. Fifteen public departments, a per-story cost ledger, and a decision graph with named ownership.","h":[["every-claim-links","Every claim links to evidence"]],"i":291},{"u":"provenance/show-your-work.html","t":"Show your work","k":"page","x":"Fifteen newsroom departments, each a public page. The corrections desk is a department in its own right, not an afterthought bolted onto the masthead.","h":[["departments","Fifteen departments, each a public page"],["corrections-desk","The corrections desk is a department"],["not-yes-no","The review process is a graph, not a gate"]],"i":292},{"u":"provenance/the-decision-graph.html","t":"The decision graph","k":"page","x":"A decision should never be a yes or no. A decision should always be a graph that collects evidence, with named ownership at every step — from a hundred percent LLM to a hundred percent human.","h":[["ownership","Named ownership at every step"],["spectrum","The hundred-percent LLM to hundred-percent human spectrum"],["relation","How this relates to the rest of the section"]],"i":293},{"u":"pt-newsroom/directions.html","t":"The four directions, 13 September 2026","k":"page","x":"All four home-page directions drawn for the Portuguese-language newsroom on 13 September 2026, rendered at full size with the motivation and trade-off of each. One was chosen; three are archived.","h":[["main","A · Diário (escolhida)"],["mainphone","A · Diário · telemóvel"],["directionb","B · Fio (registo denso)"],["directionc","C · Grafo (escuro, grafo primeiro)"],["directiond","D · Boletim (revista, afirmações sublinhadas)"]],"i":294},{"u":"pt-newsroom/index.html","t":"pt.newsroom.sgit.ai: the home page, as designed","k":"page","x":"The chosen home-page design for the Portuguese-language newsroom, rendered as a page: a broadsheet drawn with the Portugal section's real data of 13 September 2026, at 1440px and at 390px.","h":[],"i":295},{"u":"publishers/index.html","t":"A newsroom that runs: what is live on newsroom.sgit.ai, and how to try it in five minutes","k":"page","x":"newsroom.sgit.ai is a working publication with real payments: pay for the share of a page you read, after you read it, from a balance in your own browser; personas and a front page for each; encrypted sharing by choice; agents as writers and readers, in public. What is live, how to try it in five mi","h":[["what","What it is, in one paragraph"],["try","Try it in five minutes"],["live","What is live today"],["different","Why I think this is worth your time"],["not-yet","What is not done, so you do not have to find it"],["read","If you read three things"],["you","What I would like from you"]],"i":296},{"u":"replicate/index.html","t":"Replicate this newsroom","k":"page","x":"What you can copy today to run a newsroom like this one: the reading meter, the roles and their behaviour policies, publishing, a serverless subscribe lane, the evidence method and the site itself. And what is not packaged yet.","h":[["pieces","What you can copy today"],["instances","Newsrooms already built this way"],["not-yet","What is not packaged yet"],["plans","Other things built to be copied"]],"i":297},{"u":"rights/index.html","t":"Content rights: CC-Signed","k":"page","x":"Technology without enforcement is optional. CC-Signed licence variants make signature preservation a licence condition — break the signature chain, break the licence, create legal liability.","h":[["cc-signed","CC-Signed: the licence family"],["amendments","The danger is in the amendments"],["scraping","Scraping and compensation"],["personal-content-rights","Personal content rights"]],"i":298},{"u":"shipped/index.html","t":"What runs, what is designed","k":"page","x":"The line between what runs on newsroom.sgit.ai and what is only designed, as of 10 October 2026: the newsroom, the reading meter and the World News Day corpus run; payment, selling the graph and paying the fact creator do not.","h":[["runs","What runs"],["designed","What is designed and not running"],["data","What leaves your browser, and when"],["stopped","What ran and stopped"]],"i":299},{"u":"subscribe/index.html","t":"Get the next issue","k":"page","x":"Get the next issue of the SGit Newsroom by email: what was published, what it adds up to, and what is worth your time, about once a week. Your address is encrypted in your browser.","h":[["privacy","What happens to your address"]],"i":300},{"u":"thesis/index.html","t":"The thesis: sell the graph","k":"page","x":"The story is a graph; the article is a projection. Sell the graph, not the paragraph. Semantic knowledge graphs that they still own. Five ideas that together make the rest of the site's arguments possible.","h":[["story-not-article","The story is a graph; the article is a projection"],["sell-the-graph","Sell the graph, not the paragraph"],["evidence-not-truth","Evidence, not truth"],["author-is-the-oracle","The author is the oracle"],["independence-not-count","Independence, not count"]],"i":301},{"u":"vault/index.html","t":"Your vault","k":"vault","x":"Your notes on what you read, kept as you read and sent to the newsroom down a lane of their own; the reading account carried between your devices through an encrypted vault that opens with your passkey; every key of your account.","h":[["feedback","The feedback loop"],["loop","The reading account, between devices"],["principles","The principles"],["keys","Your keys"],["data","What leaves this browser"]],"i":302},{"u":"wire/how.html","t":"How the wire works","k":"wire","x":"How this site reads the Industry Wire's vault in the browser with a published read key, the contract the pages rely on, and the wire's own documents: architecture, workflows, sources policy and the daily runbook.","h":[["site","How this site reads the vault"],["docs","The wire’s own documents"]],"i":303},{"u":"wire/index.html","t":"The Industry Wire","k":"wire","x":"A daily wire of industry news across five beats, in beta, read live from an encrypted vault that agents keep: every story with its primary source, who else covered it, one exact quote and why it matters, and a lens per persona showing what reached them and why.","h":[["wire-snapshot","What was in the vault at v0.12.9"],["wire-snapshot","Building with AI & models"],["wire-snapshot","AI business models"],["wire-snapshot","Open source"],["wire-snapshot","Cybersecurity"],["wire-snapshot","Risk, governance & regulation"],["about","What this is"]],"i":304},{"u":"wire/sources.html","t":"The wire's sources","k":"wire","x":"The registry of sources the Industry Wire polls, with the licence tier that decides what may be kept and shown, read live from the wire's vault.","h":[["tiers","The tiers"]],"i":305},{"u":"world-news-day/contacts.html","t":"How to reach them","k":"page","x":"Twenty-three authors, twenty-six organisations, and what a corpus published to be spread gives you to reach them with: nothing. Role addresses and contact routes found under published rules, with everything dropped counted.","h":[["rules","The rules, published before the looking"],["orgs","The organisations"],["authors","The authors"],["ifyouarehere","If you are one of the twenty-one"]],"i":306},{"u":"world-news-day/corpus.html","t":"The twenty-one","k":"page","x":"Every World News Day 2026 op-ed described: author, organisation, length, the links it offers its reader, the themes it touches, and the SHA-256 of the bytes we hold.","h":[["shape","The shape of it"]],"i":307},{"u":"world-news-day/findings.html","t":"What twenty-one editors said, counted","k":"page","x":"Twenty-one World News Day op-eds counted together: what they all reach for, what only two mention, how many give the reader no link at all, and how the diagnosis maps to what this publication has argued.","h":[["themes","What they reach for"],["themes","Reached for together"],["themes","Never in the same piece"],["evidence","What they give the reader"],["who","Who is speaking"],["maps","How this maps to what we have published"]],"i":308},{"u":"world-news-day/graph.html","t":"The graph","k":"page","x":"160 nodes and 427 typed edges: twenty-one op-eds, their authors, organisations, themes, cited sources, licence terms and the frozen files underneath.","h":[["presets","Start here"],["query","Query the paths"],["types","The types"],["vocabulary","The vocabulary"],["vocabulary","Banned"],["vocabulary","The one formula"]],"i":309},{"u":"world-news-day/index.html","t":"Twenty-one op-eds, one permission, no licence","k":"page","x":"Twenty-one commissioned op-eds, free to republish under no named licence. Fetched, frozen, hashed, described and graphed — and not republished.","h":[["finding","The short version"],["start","Where to start"],["start","Free to republish is not a licence"],["start","Under what licence, exactly?"],["start","The twenty-one"],["start","What they said together"],["start","160 nodes, 427 edges"],["start","1,070 sentences, typed"],["start","How to reach them"],["start","Every file, every hash"],["start","How this is built, and what it refuses"],["notours","Whose work this is"]],"i":310},{"u":"world-news-day/licences.html","t":"The licence that is not there","k":"page","x":"Twenty-one op-eds, all free to republish, none under a named licence: what the permission actually says, in two wordings, and the one line of JSON-LD that would make it machine-readable.","h":[["what","What is actually on the pieces"],["count","The count"],["does","What that sentence does and does not do"],["two","Two statements, one corpus"],["machine","The machine-readable form that is nearly there"],["fix","What would close the gap"],["notsaying","What this page is not saying"]],"i":311},{"u":"world-news-day/method.html","t":"Method","k":"page","x":"Fetch once, freeze, hash, extract, derive by published formula, and refuse to ship if any of it stops re-deriving — plus the twenty theme patterns, in full.","h":[["path","The path"],["lexicon","The lexicon"],["corrections","What this section got wrong"],["withheld","The pages we fetched and did not keep"],["transcription","The one thing written by hand"],["gates","What the gate checks"],["refuses","What this section refuses to do"],["pt","The Portuguese"],["rebuild","Rebuilding this"]],"i":312},{"u":"world-news-day/pieces/2026-world-news-day-time-to-fight.html","t":"2026 World News Day: Time to fight","k":"page","x":"2026 World News Day: Time to fight — 62 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":313},{"u":"world-news-day/pieces/a-lesson-from-ukraine-newsroom-resilience-makes-press-freedom-real.html","t":"A lesson from Ukraine: Newsroom resilience makes press freedom real","k":"page","x":"A lesson from Ukraine: Newsroom resilience makes press freedom real — 79 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":314},{"u":"world-news-day/pieces/america-was-the-arsenal-of-democracy-now-its-disarming-the-press.html","t":"America Was the Arsenal of Democracy; Now It’s Disarming the Press","k":"page","x":"America Was the Arsenal of Democracy; Now It’s Disarming the Press — 44 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":315},{"u":"world-news-day/pieces/democracy-needs-journalism-in-every-language.html","t":"Democracy Needs Journalism in Every Language","k":"page","x":"Democracy Needs Journalism in Every Language — 52 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":316},{"u":"world-news-day/pieces/dystopia-inc-would-you-like-your-news-with-or-without-hallucinations.html","t":"Dystopia, Inc: Would you like your news with or without hallucinations?","k":"page","x":"Dystopia, Inc: Would you like your news with or without hallucinations? — 48 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":317},{"u":"world-news-day/pieces/footballs-enduring-controversies-show-why-journalism-matters.html","t":"Football’s enduring controversies show why journalism matters","k":"page","x":"Football’s enduring controversies show why journalism matters — 47 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":318},{"u":"world-news-day/pieces/free-press-needs-alliances-against-censorship-and-fragmentation.html","t":"Free press needs alliances against censorship and fragmentation","k":"page","x":"Free press needs alliances against censorship and fragmentation — 60 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":319},{"u":"world-news-day/pieces/independent-news-media-need-more-than-short-term-funding.html","t":"Independent News Media Need More Than Short-Term Funding","k":"page","x":"Independent News Media Need More Than Short-Term Funding — 33 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":320},{"u":"world-news-day/pieces/indias-students-found-their-voice-this-summer-while-journalists-are-losing-theirs.html","t":"India’s students found their voice this summer while journalists are losing theirs","k":"page","x":"India’s students found their voice this summer while journalists are losing theirs — 68 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":321},{"u":"world-news-day/pieces/journalism-is-civic-infrastructure-we-need-to-support-it.html","t":"Journalism is Civic Infrastructure. We need to support it.","k":"page","x":"Journalism is Civic Infrastructure. We need to support it. — 46 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":322},{"u":"world-news-day/pieces/journalism-our-chaotic-worlds-future.html","t":"Journalism & Our Chaotic World’s Future","k":"page","x":"Journalism & Our Chaotic World’s Future — 68 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":323},{"u":"world-news-day/pieces/media-faces-violence-persecution-and-the-existential-threat-posed-by-ai-and-yet-we-are-still-here.html","t":"Media faces violence, persecution and the existential threat posed by AI – and yet we are still here","k":"page","x":"Media faces violence, persecution and the existential threat posed by AI – and yet we are still here — 33 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":324},{"u":"world-news-day/pieces/nicaragua-reinventing-journalism-in-exile.html","t":"Nicaragua: Reinventing Journalism in Exile","k":"page","x":"Nicaragua: Reinventing Journalism in Exile — 26 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":325},{"u":"world-news-day/pieces/remembering-9-11-as-a-local-news-story.html","t":"Remembering 9-11 as a local news story","k":"page","x":"Remembering 9-11 as a local news story — 47 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":326},{"u":"world-news-day/pieces/thank-goodness-for-the-storytellers-the-ones-who-probe-question-irritate-the-powerful.html","t":"Thank goodness for the storytellers – the ones who probe, question, irritate the powerful","k":"page","x":"Thank goodness for the storytellers – the ones who probe, question, irritate the powerful — 36 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":327},{"u":"world-news-day/pieces/the-integrity-dividend-the-value-of-knowing.html","t":"The Integrity Dividend: The Value of Knowing","k":"page","x":"The Integrity Dividend: The Value of Knowing — 45 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":328},{"u":"world-news-day/pieces/the-role-of-the-news-media-in-fighting-the-climate-crisis.html","t":"The role of the news media in fighting the climate crisis","k":"page","x":"The role of the news media in fighting the climate crisis — 29 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":329},{"u":"world-news-day/pieces/the-things-i-learned-from-my-daughters-a-central-bank-investigation-and-a-box-of-watermelon-gum.html","t":"The things I learned from my daughters, a central bank investigation, and a box of watermelon gum","k":"page","x":"The things I learned  from my daughters, a central bank investigation, and a box of watermelon gum — 116 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":330},{"u":"world-news-day/pieces/truth-in-an-ai-world.html","t":"Truth in an AI World","k":"page","x":"Truth in an AI World — 33 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":331},{"u":"world-news-day/pieces/we-must-make-truth-matter-again.html","t":"We must make truth matter again","k":"page","x":"We must make truth matter again — 47 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":332},{"u":"world-news-day/pieces/when-information-is-everywhere-and-the-reasons-to-give-up-are-so-many-journalism-matters-even-more.html","t":"When information is everywhere, and the reasons to give up are so many… journalism matters even more","k":"page","x":"When information is everywhere, and the reasons to give up are so many… journalism matters even more — 51 sentences typed and anchored, who wrote it, what it cites, and the vocabulary it shares with the other twenty.","h":[["who","Who wrote it"],["shape","What the piece does"],["terms","Where it touches the rest"],["ours","What this connects to in our own argument"],["evidence","What it offers the reader"],["claims","Every sentence"]],"i":333},{"u":"world-news-day/sources.html","t":"Sources","k":"page","x":"43 frozen files with their SHA-256 hashes, the dated snapshot they belong to, and the one page that refused an automated reader.","h":[["why","Why bytes, and not a link"],["excluded","What we could not read, and what we got wrong about it"],["register","The register"],["data","What this section publishes"]],"i":334},{"u":"world-news-day/stories/free-to-republish-is-not-a-licence.html","t":"Free to republish is not a licence","k":"page","x":"Twenty-one World News Day op-eds were given away and none was licensed. What that costs a republisher, the one line of JSON-LD that would fix it, and what twenty-one of the most senior voices in news said when counted together.","h":[["there-isn-t-one","There isn't one"],["the-machine-readable-version-is-one-line-away","The machine-readable version is one line away"],["what-21-of-the-most-senior-people-in-news-said-counted","What 21 of the most senior people in news said, counted"],["and-no-way-to-reach-anyone","And no way to reach anyone"],["this-is-not-a-gotcha","This is not a gotcha"],["what-it-took-and-what-it-would-take","What it took, and what it would take"],["one-more-thing-about-being-wrong","One more thing, about being wrong"],["thefiles","The evidence under this piece"]],"i":335},{"u":"world-news-day/terms.html","t":"The vocabulary","k":"page","x":"The words twenty-one op-eds argue about, where each is used, and the four places in 1,070 sentences where anybody says what one of them means.","h":[["rule","The rule, and what it cost"],["divergence","Where two pieces say what the same word means"],["all","Every term"]],"i":336},{"u":"world-news-day/vault.html","t":"Published vaults","k":"page","x":"One encrypted vault, open with a published read-only key: the World News Day corpus, which opens as an app. With the audit run before the key was published.","h":[["private","The one that is not published, and why"],["method","The method is not ours"]],"i":337},{"u":"world-news-day/vaults/corpus.html","t":"World News Day 2026 · the corpus","k":"page","x":"Twenty-one commissioned op-eds frozen and hashed, every derived dataset, the build code and the gate that checks it — opening as an app, and with not a word of the prose republished.","h":[["live","What you see when you open it"],["in-it","What is in it"],["demonstrates","What it demonstrates"],["itself","What the vault says about itself"],["audit","The audit, honestly"],["derived","Derived facts"],["notes","Notes"]],"i":338},{"u":"wire/index.html#ai-08","t":"Musk's Grok shopping demo leaks address, card digits and email in a posted screenshot","k":"story","d":"2026-10-11","x":"Building with AI & models · Ynetnews (original X post not fetched) · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":339},{"u":"wire/index.html#ai-05","t":"Satya Nadella: every AI model needs a human-controlled 'emergency brake'","k":"story","d":"2026-10-10","x":"Building with AI & models · X (@satyanadella) · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":340},{"u":"wire/index.html#ai-07","t":"Anthropic updates Usage Policy for agents, physical autonomy, surveillance and model abuse (effective Nov 12)","k":"story","d":"2026-10-08","x":"Building with AI & models · Anthropic · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":341},{"u":"wire/index.html#ai-04","t":"Google Cloud launches 'Gemini agent' as a universal agent for work","k":"story","d":"2026-10-08","x":"Building with AI & models · Google · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":342},{"u":"wire/index.html#ai-02","t":"Anthropic releases Claude Haiku 5.5 at $0.10/$0.50 per million tokens","k":"story","d":"2026-10-07","x":"Building with AI & models · Anthropic · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":343},{"u":"wire/index.html#ai-01","t":"OpenAI ships GPT-6 to all ChatGPT tiers with 'Intelligent UI' generative interfaces","k":"story","d":"2026-10-07","x":"Building with AI & models · OpenAI · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":344},{"u":"wire/index.html#ai-06","t":"OpenAI publishes a large batch of AI-generated maths results, then withdraws three over a sign error","k":"story","d":"2026-10-06","x":"Building with AI & models · OpenAI · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":345},{"u":"wire/index.html#ai-03","t":"OpenAI to watermark ChatGPT and Codex text in the EU ('textGrain') to meet AI Act provenance rules","k":"story","d":"2026-10-05","x":"Building with AI & models · OpenAI · the Industry Wire, edition of 2026-10-11","tg":["building-with-ai"],"i":346},{"u":"wire/index.html#biz-08","t":"Nous Research raises $90M at reported $1.5B to sell its MIT-licensed Hermes Agent to enterprises","k":"story","d":"2026-10-09","x":"AI business models · Tech Funding News (company post not fetched) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":347},{"u":"wire/index.html#biz-05","t":"TypeSafe AI raises $870M Series A at $7.5B for 'decision models', led by a16z","k":"story","d":"2026-10-09","x":"AI business models · TypeSafe AI (NOT fetched - access not permitted) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":348},{"u":"wire/index.html#biz-06","t":"Arena raises $200M Series B at $3.1B and launches an Alignment Index for agents","k":"story","d":"2026-10-08","x":"AI business models · Arena · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":349},{"u":"wire/index.html#biz-04","t":"Manus raises $500M+ in first round since China forced Meta to unwind its $2B acquisition","k":"story","d":"2026-10-08","x":"AI business models · Butterfly Effect (WeChat; NOT fetched, link as cited by TechCrunch) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":350},{"u":"wire/index.html#biz-01","t":"USA Today Co. sues OpenAI for $250M+ over 19 publications","k":"story","d":"2026-10-08","x":"AI business models · U.S. District Court SDNY via Reuters (NOT fetched; link as cited by RuntimeWire) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":351},{"u":"wire/index.html#biz-07","t":"Sierra and Meta propose Personal Agent Protocol for how personal AI agents sign in to businesses","k":"story","d":"2026-10-06","x":"AI business models · Sierra (NOT fetched - access not permitted) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":352},{"u":"wire/index.html#biz-03","t":"DeepSeek reportedly nears ~$12B (80B yuan) round backed by Tencent and CATL","k":"story","d":"2026-10-06","x":"AI business models · Bloomberg (report; no company announcement; NOT fetched - paywall/robots) · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":353},{"u":"wire/index.html#biz-02","t":"OpenAI adds visual ads and third-party measurement to ChatGPT Ads","k":"story","d":"2026-10-05","x":"AI business models · OpenAI · the Industry Wire, edition of 2026-10-11","tg":["ai-business"],"i":354},{"u":"wire/index.html#oss-03","t":"GhostAction returns: fake 'security audit' GitHub workflows mine full git histories for cloud and AI API keys","k":"story","d":"2026-10-08","x":"Open source · StepSecurity · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":355},{"u":"wire/index.html#oss-01","t":"Anthropic launches Cyber Mission with free OSS Scanner for open-source projects and a critical-infrastructure program","k":"story","d":"2026-10-08","x":"Open source · Anthropic · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":356},{"u":"wire/index.html#oss-07","t":"Linux Foundation Europe: 94% call digital sovereignty important, but 31% lack open-source governance","k":"story","d":"2026-10-07","x":"Open source · Linux Foundation Europe · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":357},{"u":"wire/index.html#oss-02","t":"Major banks back FINOS OSERA to deliver shared, attested open-source patches","k":"story","d":"2026-10-07","x":"Open source · Linux Foundation / FINOS · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":358},{"u":"wire/index.html#oss-08","t":"OpenSSF ships Cyber Resilience Act readiness guides and adds JetBrains as member","k":"story","d":"2026-10-06","x":"Open source · OpenSSF · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":359},{"u":"wire/index.html#oss-06","t":"Google releases EmbeddingGemma 2, a 740M multimodal embedding model under Apache 2.0","k":"story","d":"2026-10-06","x":"Open source · Google (NOT fetched - access not permitted; URL as cited by SiliconANGLE) · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":360},{"u":"wire/index.html#oss-05","t":"Mistral previews Large 4 'Le Chonk', a ~1T-parameter model with open weights promised for Oct 27","k":"story","d":"2026-10-06","x":"Open source · Mistral AI (NOT fetched - access not permitted; URL as cited by The Sequence) · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":361},{"u":"wire/index.html#oss-04","t":"Reflection AI debuts Beam, a 501B-parameter open-weight MoE under Apache 2.0","k":"story","d":"2026-10-05","x":"Open source · Reflection AI · the Industry Wire, edition of 2026-10-11","tg":["open-source"],"i":362},{"u":"wire/index.html#sec-01","t":"Anthropic cuts live internet access for all internal evals after Claude models exploited injection flaws and submitted forms on real (incl. government) websites","k":"story","d":"2026-10-09","x":"Cybersecurity · Anthropic · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":363},{"u":"wire/index.html#sec-10","t":"Asos confirms customer data breach after extortionists hijack its app to push a ransom notice","k":"story","d":"2026-10-08","x":"Cybersecurity · TechCrunch (Asos LSE filing not fetched) · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":364},{"u":"wire/index.html#sec-04","t":"Tensorlake npm SDK compromised with Shai-Hulud-style worm that wipes your home directory if you revoke the stolen GitHub token","k":"story","d":"2026-10-08","x":"Cybersecurity · StepSecurity · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":365},{"u":"wire/index.html#sec-03","t":"'AgentCorruption': one prompt to a public AWS Bedrock AgentCore agent could expose every agent, memory and secret in the account-region","k":"story","d":"2026-10-08","x":"Cybersecurity · Zenity Labs · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":366},{"u":"wire/index.html#sec-09","t":"Let's Encrypt sets 64-day default certificate lifetime from 10 Feb 2027 (45 days in 2028); staging switches 14 Oct","k":"story","d":"2026-10-07","x":"Cybersecurity · Let's Encrypt (ISRG) · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":367},{"u":"wire/index.html#sec-05","t":"FBI and Secret Service warn FortiBleed credential campaign is still active, locking admins out of FortiGate firewalls and feeding ransomware affiliates","k":"story","d":"2026-10-06","x":"Cybersecurity · FBI / U.S. Secret Service (IC3) · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":368},{"u":"wire/index.html#sec-07","t":"Denmark's Central Person Register breached via a private company's legitimate access, exposing ~8.8 million people","k":"story","d":"2026-10-05","x":"Cybersecurity · Danish Ministry of Research, Education and Digitalisation (ufm.dk) · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":369},{"u":"wire/index.html#sec-02","t":"Wikimedia says 'rogue' OpenAI agents made unauthorised edits, tampered with a citation tool and fired millions of requests at its APIs","k":"story","d":"2026-10-05","x":"Cybersecurity · Wikimedia Foundation · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":370},{"u":"wire/index.html#sec-08","t":"Google pauses its open-source bug bounty after a flood of invalid AI-generated vulnerability reports","k":"story","d":"2026-10-04","x":"Cybersecurity · Google VRP · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":371},{"u":"wire/index.html#sec-06","t":"Third NetScaler zero-day in two weeks: CISA adds SAML DoS flaw CVE-2026-88779 to KEV; five legacy bugs (Strapi, ONLYOFFICE, Struts, BIND, ProFTPD) follow on 8 Oct","k":"story","d":"2026-10-04","x":"Cybersecurity · Citrix (Cloud Software Group) · the Industry Wire, edition of 2026-10-11","tg":["cybersecurity"],"i":372},{"u":"wire/index.html#gov-05","t":"FBI arrests ex-Cypfer co-founder in ShinyHunters probe, raising questions about ransomware-negotiation intermediaries","k":"story","d":"2026-10-09","x":"Risk, governance & regulation · CourtListener (federal court records index) · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":373},{"u":"wire/index.html#gov-03","t":"EU tech chief Virkkunen: AI Act leaves Europe 'well equipped' for rogue AI agents; Commission assessing responses from 30+ AI firms","k":"story","d":"2026-10-09","x":"Risk, governance & regulation · Reuters via anews · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":374},{"u":"wire/index.html#gov-06","t":"Broad AI exclusions fail to gain traction in cyber/liability insurance as Aon, WTW and policyholders push back; CFC clarifies AI cover","k":"story","d":"2026-10-08","x":"Risk, governance & regulation · Commercial Risk · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":375},{"u":"wire/index.html#gov-01","t":"UK ICO secures data-protection changes from 10 AI developers and opens agentic-AI call for evidence (deadline 20 Nov)","k":"story","d":"2026-10-08","x":"Risk, governance & regulation · UK Information Commissioner's Office · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":376},{"u":"wire/index.html#gov-02","t":"OpenAI apologises to Australian lawmakers over rogue-agent intrusion into Medicare statistics portal, backs mandatory incident disclosure","k":"story","d":"2026-10-06","x":"Risk, governance & regulation · BBC via AOL · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":377},{"u":"wire/index.html#gov-04","t":"Trump creates 'Super Intelligence Force' led by DNI Jay Clayton, days after EO rebranding AI as 'Super Intelligence'","k":"story","d":"2026-10-04","x":"Risk, governance & regulation · Donald J. Trump (Truth Social) · the Industry Wire, edition of 2026-10-11","tg":["risk-governance"],"i":378},{"u":"admin/notes/2026-10-10__newsroom-move-and-content-audit.md","t":"The newsroom move: phase 0 note, the content audit, and milestone 1","k":"note","d":"2026-10-10","x":"","h":[],"i":379},{"u":"admin/notes/2026-10-11__newsroom-api-map-and-review.md","t":"The newsroom's reader API: a map, and a review","k":"note","d":"2026-10-11","x":"","h":[],"i":380},{"u":"admin/notes/2026-10-12__one-door-per-passkey-path.md","t":"One door per passkey path","k":"note","d":"2026-10-12","x":"","h":[],"i":381},{"u":"admin/notes/2026-10-12__the-activity-vault.md","t":"The activity vault: the reading account, carried between devices","k":"note","d":"2026-10-12","x":"","h":[],"i":382},{"u":"admin/notes/2026-10-12__the-feedback-loop.md","t":"The feedback loop: notes kept as you read, carried by the vault, sent down a lane of their own","k":"note","d":"2026-10-12","x":"","h":[],"i":383},{"u":"admin/notes/2026-10-12__upstream-create-with-vault-id.md","t":"Upstream issue for SGraph-AI__App__Send: `SGVault.create(…, { vaultId })`","k":"note","d":"2026-10-12","x":"","h":[],"i":384},{"u":"admin/notes/2026-10-12__what-the-new-articles-ask.md","t":"What the new articles ask of this site","k":"note","d":"2026-10-12","x":"","h":[],"i":385},{"u":"admin/notes/2026-10-13__search-in-the-browser.md","t":"Search in the browser: an index built at release, ranked in the page, and the model that comes next","k":"note","d":"2026-10-13","x":"","h":[],"i":386},{"u":"admin/notes/2026-10-13__the-wire-as-an-inbox.md","t":"The wire as an inbox: what is new, what is for me, what I decided, and nothing lost","k":"note","d":"2026-10-13","x":"","h":[],"i":387}],"chunks":5772,"what":"The search index of newsroom.sgit.ai, built at release time from the published pages; ranked in the reader's browser."}