Home / Articles / Simple on top of years: why a startup has to find something very easy, very narrow and able to scale, why building from the bottom up, from provenance, understanding and determinism, ended in a simple product anyway, and how a website in five designs opened from a QR code sits on a CI pipeline, zero-knowledge sharing, sgit, vaults, keys and agentic teams, each learnt on its own
Simple on top of years: why a startup has to find something very easy, very narrow and able to scale, why building from the bottom up, from provenance, understanding and determinism, ended in a simple product anyway, and how a website in five designs opened from a QR code sits on a CI pipeline, zero-knowledge sharing, sgit, vaults, keys and agentic teams, each learnt on its own
By Dinis Cruz · 2026-11-19 · article v1.0.0 · newsroom v0.14.174 · startupsstrategysovereign-websitesgitagenticdeterminismarticle
Abstract: My view of startups has always been that you have to find something very easy: a narrow use case, at scale, that somebody can just use. What is odd is how I got to one. For the last couple of years I have been building from the bottom up, starting from provenance, from understanding and from determinism, because I do not want the black magic. You would expect that to end in something complicated. It ended in something simple: a person scans a QR code and sees their own website in five designs, and if they say yes it becomes theirs, with a team of agents inside it. That simple thing sits on layers I learnt one at a time, each for its own reason: a CI pipeline and tests on every change, almost the first thing I built; a file-sharing service where the server never holds the key, which taught me pricing and payments and also that "the technology worked, the use case was not there yet"; sgit, Git's model encrypted; vaults a web page can open; keys and secrets as QR codes; agentic teams with briefs and behaviour policies; and a factory that turns one commit into a site. None was built for this. Connecting them is the product, and the learning is the moat.
In short
- A startup needs something very easy: narrow, at scale, that somebody can just use.
- I built from the bottom up: provenance, understanding, determinism, no black magic.
- It ended in something simple anyway: your website, five designs, from a QR code.
- Each layer under it was learnt on its own, for its own reason.
- Connecting them is the product; the learning is the part nobody can copy.
What a startup needs
I have said it for a long time: a startup has to find something super easy. A narrow use case, one that can scale, that somebody can just use, without a meeting, a manual or a migration. The ideas that fail are rarely bad ideas; they ask too much of the person before they give anything back.
I learnt that the hard way this year too. SG/Send, the file-sharing service that became sgit, had paying customers within a month of its first commit, and by 5 March its own briefs said it plainly: "Users buy credits, test once, and don't come back." The technology worked. The use case was not there yet (from sending a file to sgit).
From the bottom up
The odd thing is the direction I came from. I have been starting from the bottom: from provenance, where everything came from and who changed it; from understanding, being able to explain every step; and from determinism, the same inputs giving the same result. I do not want the black magic, the parts that work but nobody can say why.
You would expect that to end in something complicated, a platform for engineers. It ended in the opposite. A person scans a QR code and sees their own website, in five designs, made by agents (your website, from a vault). If they say yes, it becomes theirs, with a team of agents inside it (a website with a team inside). That is about as easy as it gets.
The layers
None of the layers was built for this. Each was learnt on its own, for its own reason, and that is the point:
- CI, tests and a release on every change was almost the first thing I built, and everything since runs on it: this newsroom releases about ten times a day because every change is checked (the second run).
- Zero knowledge: SG/Send's first specification, on 8 February, said the decryption key is never sent to the server. That rule is why a demo can now sit in a vault that nobody but its recipient can read.
- Selling it: taking SG/Send to market taught me Stripe, pricing and everything else a real product needs, by 2 March, long before the newsroom needed it.
- sgit, Git's model with everything encrypted, named on 16 March, so that work and its history live in a vault.
- Vaults a web page can open, file by file or as an app in a sandbox (how a vault works), which is what makes a website in a vault possible.
- Keys and secrets as QR codes, tried on 21 October (a secret is a QR code), which is how a demo reaches someone.
- Agentic teams, with briefs, roles and a behaviour policy for every agent (governance from zero), which is who builds the five designs.
- The site factory, one commit to a repository, an address and its owner (the site factory works), which is how a demo becomes a real site.
Connecting the dots
I could connect them because I had lived each one. I was experiencing it, learning how to use Claude, learning how each piece behaves when it fails. Nobody planned a demo opened from a QR code in February; it became possible only because each layer was already there, tested and understood, when the idea arrived.
That is also why I am not in love with any of the technology. The open source habit is to make things easier and simpler, and to drop whatever stands in the way. A layer earns its place by making the thing on top simpler for the person using it.
Why this is the moat
Almost all of this is open source, so anyone can copy the code. What they cannot copy is the order in which it was learnt, or the speed that comes from having learnt it: PrompTugal's investment case says it in one line: "A competitor who forks our code gets our position as of today", not our velocity. The simple product on top is what sells. The years underneath are why it works, and why we can keep making it simpler.
What this does not claim
- The dates are as recorded in this newsroom's articles and their sources; most layers kept changing after them.
- The demo pipeline and the vault drop page are not built yet; the layers under them are.
- "Simple" is a claim about the person's experience, not about what runs underneath.
Where this comes from
My own thinking, in writer-10's session on 19 November 2026. The dates and quotes come from from sending a file to sgit, a secret is a QR code, how a vault works, the site factory works and the second run; PrompTugal's investor deck, The investment case (slide 7), in its public repository; related: governance from zero, a team like this, your website, from a vault, a website with a team inside.
Peer review Not reviewed yet. Review it
Threads
Builds on
- From sending a file to sgit: eight months of SG/Send, told by its own briefs and commits, and how a file-sharing service became encrypted version control and vaults Eight months of SG/Send, checked against its own briefs and commits: how file sharing became sgit and vaults, the pace, and what the market said.
- Your website, from a vault: how to show someone five versions of their own website without publishing anything, sealed in an encrypted sgit vault and opened with a QR code, a link or a zip on one drop page, the demo pipeline from a name to an email, and how a demo becomes their sovereign website without starting again Show someone five versions of their own website without publishing anything: sealed in a vault, opened with a QR code, a link or a zip.
- A website with a team inside: why a Sovereign Agentic Website is delivered not as a site but as a site with six or seven agents already set up and coordinated in its repository, a conductor, a designer, a writer, a developer, a reviewer, a translator and a keeper, each with a role and a behaviour policy, and what that means for the person who owns it A Sovereign Agentic Website comes with a team inside: seven agents set up and coordinated in the owner's repository, each with a behaviour policy.
- The second run: a forensic look at why a release of this newsroom takes about 24 minutes, measured stage by stage over two timed releases and 437 CI runs since 18 October, what each check protects, what runs twice, what can run side by side, and why, if nothing changed, the second run should do almost no work at all Why a release takes 24 minutes, measured stage by stage, and why a second run with nothing changed should do almost no work.
- How a vault works: an encrypted store that acts like a database, where a server that is fully compromised leaks nothing, one mutable file lets every other object be cached for ever, and a web page can open a file in four requests or run a whole app in a sandbox How an sgit vault acts like a database on a server that cannot read: one key-derived mutable file, everything else immutable.
- A secret is a QR code and a link to a public key: the workflow the QR experiment opens, who else already does this, and a format to share Send a QR code and a link to a public key: the workflow the QR experiment opens, who already does this, and a format to share.
- Governance from zero: Nina J. noticed that identity and behaviour policies come with our smallest £500 offer, not as a premium upgrade, because "we're only testing it" has never reassured a security review; my answer is that even the £0 offer has all of it, and here is why open source is what makes that possible, why security sold as an upgrade is a choice, and why a pilot with its governance built in is a pilot that can convert A reply to Nina J.: identity and behaviour policies come with our smallest offer, even at £0, because everything is open source.
- The site factory works: three AgentiCCells sites made by hand, then one made by a commit in two minutes and thirty-nine seconds, live over https with a first version written by Claude Code, and the long list of what it took to make something that hard look easy Three AgentiCCells sites by hand, then one by a commit: live over https in 87 seconds, a first version by Claude Code at 159.
- A team like this, for sale: what we sell is not the platform, it is the agentic team that built it, its roles, its behaviour policies, its security passes and its speed, with the code open and the data public, and who I want to talk to now that PrompTugal.news is a working data platform and not an experiment What we sell is not the open platform but a team like the one that built it, and who I want to talk to now.
Continued by
- Watch your website being built: a brief for a page that builds a first version of a sovereign website in the visitor's own browser, from a description or a voice note, with their own model key, showing every agent at work, every behaviour policy being applied and every penny it costs, then hands back a zip, the same brief run on several models side by side as evals, and the Web Summit demos it grows from A page that builds a first version of a sovereign website in your browser, showing every agent, behaviour policy and penny, then hands back a zip.
- Sovereign websites, the investment case: PrompTugal's investor decks rewritten for a deliverable that is needed everywhere, in every language, in a market built on rent and lock-in, a one-off delivery that is already fully agentic, value given before payment through a demo in a vault, a long list of upsells, and why this is now a proper, scalable, investable architecture Sovereign websites as an investment: one deliverable needed everywhere, a one-off agentic delivery, value first through a vault demo, and upsells.
Article No. 498 · start a reading list with it
Reviews
No reviews yet.