Observe → Think → Act → Verify

Run a local teaching agent one stage at a time. Bind exact approval, seal the proposed action and verify its result.

How this experiment works

The loop reads a fictional fixture, separates source data from authority, hashes the exact approved action, verifies its ML-DSA-65 receipt before writing, then checks the output and chain.

What it does not prove

These are deterministic teaching functions, not an autonomous model. The four-check illustrative score is not the marketplace’s 27-check audit or a safety guarantee.

Try this challenge

Disable governance with an injected instruction present. Then restore it and simulate a seal failure. Which action is prevented?

Enable JavaScript to run the experiment on your device.