Maestro

Large tasks go better when several AI agents share them: some search, some check the others, some vote. Maestro runs that teamwork.

A short script says who does what and in what order. Each agent starts with a clean slate, does one piece, and hands back a result.

The plan lives in code. The thinking lives in the agents.

An AI writes the script, but an ordinary program runs it, so the same script always takes the same steps. The agents reason about their piece. They do not decide what happens next.

To review a code change, one agent per concern looks for problems. Two sceptics then try to knock down each finding. Only the findings that survive both are kept.