Work · Side project · maintained by an autonomous agent
SIGNAL / NOISE
An AI news board that ranks by confirmation
The pitch for autonomous agents is “it ships while you sleep.” I want the one that knows when not to.
An open-source board that ranks AI news by how many independent sources confirm each story. Every Monday an autonomous agent works on it without me, with a written backlog and rollback rules it cannot change.
The problem
AI news moves faster than anyone can verify it. A story carried by twelve outlets in under two hours is not necessarily more true than one carried by two; what matters is how many independent sources actually confirm it.
The approach
The board scores stories by independent confirmation. The agent's Monday loop is boring on purpose: measure the live board, pick one item from a written backlog, build, run the tests, deploy, measure again, and roll back if confirmed stories fall twenty percent, total stories fall thirty percent, or live sources drop at all.
What is true today
On its first unsupervised run the test suite was unreachable, so the agent logged “blocked” and investigated instead of shipping. It found that the feed-health metric only counted a first-saved timestamp, so healthy feeds looked dead forever. The fix was to the metric; a week later the check read forty-three of forty-six. Since then: nine backlog items closed, zero rollbacks triggered, and four tuned numbers the agent is never allowed to touch.
The question I ask
What would you need to see before letting an agent deploy with nobody watching?
Related notes
Written along the way.
The first time my autonomous dev agent ran on its own, it refused to write any code
Best decision it has made. The outage was a broken ruler, and the fix was to the metric, not the feeds.
Read ↗The logging tier is the compliance story, not the model
Anthropic launched Claude for Financial Advisors with eleven partner connectors. What changes is the read path. What does not change is who owns the obligation.
Read ↗