The most useful system we run is the one that's not allowed to build anything

Description and use

Brain is a read‑only planning layer that never writes, ensuring trustworthy, auditable AI decisions.

Problem

Agent systems that can act autonomously risk making incorrect changes based on stale or contaminated data, leading to silent failures and loss of trust.

Outcome

By separating the smart, read-only reasoning layer from the deterministic, auditable execution layer, the system achieved high trust, reduced silent failures, and maintained clear accountability.

Workflow

• 1. Brain scans all repositories, docs, and deploy states to build a knowledge base. 2. When queried, Brain retrieves evidence from the relevant project only. 3. Brain presents answers with source references. 4. A separate deterministic execution layer receives Brain’s proposals and performs actions. 5. All actions are auditable line-by-line.
• 6. Memory is stored in Alchemyst AI and can be traced back to its source.
• 7. Refresh loops keep Brain’s view up-to-date.
• 8. Filters prevent docs and chat history from leaking across project boundaries.

Evidence

Brain knows the state of every project we run. Ask it "where did we land on the pricing page copy" and it answers from the actual state of the repo, the docs, and the decisions around them - with one boundary: it will not mix two projects into one answer.

Source

We split the layers. A judgment layer that reads, remembers, routes, and answers - and separate, boring, deterministic systems that write.

Source