An AI evaluator penalized me for breaking a rule that didn’t exist. It deducted leaderboard points because I used no hashtag. The campaign brief required none. Another time, I was penalized for not quoting a post when quoting wasn’t mandatory. Now imagine giving that verdict control of a bounty payout. A normal smart contract can enforce “pay only if approved” flawlessly. But it cannot open the brief to check whether the rejection makes sense. It depends on an outside verdict, delivered through an oracle. An Intelligent Contract on @GenLayer changes that. It can read an accessible public brief and submission, use a language model to interpret the requirements, and have validators independently assess the proposed result using different models. Disagreement can trigger another round. Eligible decisions can be appealed. That doesn’t guarantee the judgment is correct. Several models can still agree on a mistake. If I designed this evaluation, rejection would require evidence: show me the exact requirement the creator failed to meet. I don’t just want a contract that executes a verdict correctly. I want one that can examine the basis for that verdict. If a contract could decide whether you get paid, what would it need to show you before saying no?