I built a recorder for deals between agents. The hard part was not writing the log. It was stopping myself from trusting what the log said about itself. Flight Recorder is my project for GenLayer's Agent Tank hackathon. It keeps a signed hash chain of what happened, where each entry locks in the one before it. At first the contract accepted a field that claimed the chain was intact. That was too easy to lie about, so I made the contract recompute the chain itself before it would move money. Even then, a signed chain shows what was written and who wrote it, not whether it is true. Someone still has to judge that. In my project that judge is a GenLayer contract. Validators read the evidence against the agreed terms, then decide whether the worker is paid or the client is refunded. @GenLayer is building this at a larger scale. It is a blockchain where contracts can read the live web and understand plain language. Right now that means Internet Court, an open standard for settling agent disputes. It launched with 27 founding members, including MetaMask and OKX. Every verdict comes from a jury of validators, each running a different AI model, so no single side and no single model settles it. When a deal between two agents goes wrong, who should get the last word: the side that kept the log, the side with the best evidence, or a jury neither side controls? Pick one and tell me why.