Our Story — Why Product Brain Exists

I sold my first website when I was 12. Not a lemonade stand, an actual website, to an actual business, for actual money. Since then I've spent over a decade deep in product governance and Holacracy, watching the same failure repeat in different costumes.

Product knowledge decays when nobody governs it.

Teams make careful architectural choices, establish standards, and resolve painful tensions. Then the reasoning scatters. It lands in Slack, Notion, pull request comments, meeting notes, and the heads of a few senior people. Months later, the same tradeoff gets reopened because nobody can tell which context still holds.

AI agents made that gap impossible to ignore. Cursor, Claude Code, Codex, and MCP-enabled tools can move fast, but speed makes stale or disconnected context more expensive. If the agent cannot see the decision, the tension behind it, and the standard it created, it can confidently undo work that was correct for reasons it never saw.

That gap between agent capability and governed product context became the thing I couldn't stop thinking about.

The thing that clicked

I realized that agent-assisted product work needs a source of truth built for both humans and agents. Not another wiki. Not another prompt file. A governed knowledge graph where product facts have type, status, provenance, and relations.

That graph is the Chain. Decisions link to the tensions that caused them. Standards link to the work they constrain. Insights link to the work packages they inform. Agents read it before they act. Humans govern what becomes durable.

Product Brain exists so product knowledge compounds instead of scattering: decisions, tensions, standards, business rules, architecture notes, work packages, and insights connected in one governed Chain.

Instructions that stay accountable

Most teams treat agent instructions as files that drift from reality. You write a CLAUDE.md or project rule, it helps for a while, and then the product changes around it. The file still sounds plausible even when it no longer reflects the current system.

Product Brain treats instructions, skills, rules, and standards as living entries on the Chain. They are versioned, connected to the decisions that created them, and reviewed as the product changes. Governance is not overhead here. It is what makes agent work trustworthy.

Built in the open

I started community.productbrain.io for builders who are already using AI agents in real product work and want a cleaner way to keep context usable. The product is in open beta. The method is still being learned in public.

If you are trying to make product thinking compound across people, tools, and sessions, come compare notes. I do not have all the answers. I do have a strong conviction: the future of agent-assisted product work needs governed context, not more scattered context.