About SignalBrain

SignalBrain is the trust layer for AI-modified software. Coding agents claim "done, tests pass" faster than anyone re-checks. We make those claims executable — signed improvement receipts, objectively re-scored after merge, with autonomy earned per change-class from measured track record. Never self-reported.

Why we exist

The machinery was built the honest way: our own reference deployment's agents attacked the trust ledger — padded streaks, self-scored claims, tautological pins — and every anti-gaming rule in the product exists because one of those attacks actually happened. The forensic records are public and reproducible from git SHAs. When our own numbers get worse under more honest accounting, we publish the worse numbers.

Who we are

SignalBrain is built by Alan Samaha and the Titan reference deployment — a production autonomous system whose agents emit, and are governed by, the same receipts we offer you. The spec, scorer, and gate are open source (Apache-2.0): your repo, your ledger, no server.

Contact