The seed companies

The last MVP in the family is not a publication at all — it is the business idea itself, given away. Seven fully-formed, Creative Commons-licensed company descriptions, published so that somebody else can pick one up and build it. One of the seven is an evidence-research company aimed squarely at journalists and investigators, which is why this page is on a news site. Designed 17 May 2026; none of the seven was ever published.

Not a real company, and not a fake one

The source brief spends its first section on getting the category right, and the distinction it lands on is load-bearing:

The brief records that it tried “fake company” first and rejected it, on the grounds that the phrase implies deception. What it settles on is a genuinely unusual artefact defined by three properties held at once: fully-formed (market analysis, product, pricing, operating model, technology stack, investment ask — not a paragraph), Creative Commons-licensed (so it can be taken and built without asking), and platform-aware (the economics assume a specific substrate, which is what makes it viable at small scale).

Each of those exists on its own — business plans get written, CC licensing is ordinary, platform-dependent businesses are everywhere. The combination is the thing. And the brief insists on a fourth property that is really an editorial standard: open about its own status, so a reader is never in any doubt they are looking at a proposal rather than a going concern.

The seven

Seed 4: the evidence-research company

This is the seed that belongs to this site, and it is the commercial form of the argument the rest of the site makes abstractly. The company sells the evidence work rather than the article: the vault as an evidence locker, agents doing source-checking, citation management and document analysis, cross-document linking as the deliverable.

The economic claim is the interesting part, and it is a claim about who gets served rather than about margin:

That is the same argument as paying the fact creator approached from the buying side. This site argues that near-zero-cost settlement makes it possible to pay for a single fact; the evidence-research seed argues that near-zero-cost evidence assembly makes it possible to commission an investigation that no newsroom would fund and no law firm would quote for. Local accountability reporting, a single-issue campaign group, an expert witness working alone — these are the buyers that currently fall below the floor.

The brief also names, in one line, the relationship between this seed and the one property in the family that actually runs: myfeeds.ai is the live instance of the evidence-research seed. The seed is the generalised business; the property is one person's version of it, already operating. What that property charges and who it serves is deliberately not published here — see the note on the MVP index.

Why give the idea away

Six reasons are given. Five are ordinary strategy — it demonstrates capability without a sales pitch, it seeds an ecosystem, it is cheap to produce relative to building each one, it generates publishable content, it forces the authors to articulate what the platform actually enables. The sixth is the one with an edge on it:

For a news site there is a sharper version of that, and it is the reason this page is not merely adjacent to the argument. This site's whole thesis is that the value is in the graph rather than the paragraph — that the assembled, evidenced, structured thing is what is worth paying for, and the readable projection on top of it is nearly free. A publisher who believes that should be entirely comfortable publishing their business plan, because the plan is a paragraph and the operation is a graph. Giving away the seven seeds is the thesis being tested on its authors.

What a seed actually contains

Each seed is one vault with a consistent layout — the investor-pack template filled in, plus the working detail a builder would need:

seed-{name}/
├── pack.md                  the investor pack, against the shared template
├── operating-model.md       how the company runs day to day
├── product.md               what it builds
├── financials/              projections · unit economics · assumptions
├── go-to-market/            strategy · pricing · partnerships
├── technology/              architecture · platform leverage · build vs buy
├── examples/                sample customer flow, sample output
├── license.md               Creative Commons
└── readme.md                front matter; how to use this vault

Underneath the seven sits a reusable investor-pack template — the same thirteen sections any company would present, held as a vault rather than a slide deck. The brief argues the vault form buys four things a PDF cannot: the reader can change the financial assumptions and see the model move, every market claim links to the source behind it, each reader can annotate their own copy without affecting anyone else's, and when the pack is updated the reader sees what changed rather than receiving a new file. That last one is supersede-never-delete applied to fundraising.

The risks the brief names against itself

Risk 3 has a consequence the brief accepts explicitly — a numbered task requiring a documented decision per seed: published, internal-only, or deferred. Whether the evidence-research seed would have survived that filter, given that myfeeds.ai already occupies exactly that space, is not recorded. It is a fair question to ask of this page, and the honest answer is that the source does not answer it.