The corporate brain practice

AI agents fail without ground truth. We build the corporate brain they stand on.

A maintained, agent-readable knowledge base that grounds every agent in your organization — and that your agents feed back, so it compounds instead of decaying.

Read the method Are you an agent? Read /llms.txt — this whole site is compiled for you.

The thesis

One loop, not two projects

Agents without a brain hallucinate. A brain without agents rots. Built as one system, every agent interaction makes the next one smarter.

The Brain

Compiled pages with sources, freshness claims, and cross-references — not a document dump.

The Agents

Answer from the brain, cite pages — and write what they learn back in, under rules.

The method

Assess. Install. Curate.

1–2 weeks · fixed scope

Assess

Map where knowledge actually lives, write your brain's constitution, rank the agent use cases by payoff. You own the architecture either way.

3–6 weeks

Install

The brain stands up on your infrastructure — plain Markdown, no lock-in — with the first agents grounded on it and lint gates enforcing quality.

monthly

Curate

Freshness audits, signal ingestion, contradiction chasing — the discipline that keeps the brain an appreciating asset instead of another dead wiki.

The full method, published openly

Proof, not promises

Example wikis — browse a working brain

Don't take the method's word for it. These are live public wikis built and maintained with it — open any page and you'll find the sources, cross-references, and freshness stamp the page contract demands. Your agent can read them too.

The library

Everything, published

Every page below carries a freshness stamp and is agent-readable. The site practices the method it sells.

concept

method

offering

signals

reference