Self-improving agent governance: 👍/👎 → Pre-Action Gates that block repeat AI mistakes. Stop paying for the same mistake twice.