Team Idea
Solo build for the Options Alpha Agents challenge. I built an options agent whose defining behaviour is refusal: in the judged week it evaluated 204 candidates and took none, because the credit on offer never paid for the width of the spread. Every refusal is written to a hash-chained decision log together with the gate that rejected it and the numbers that failed, so the reasoning can be audited rather than taken on trust, and one command verifies the whole chain.
Public MIT repository, hosted browser demo, 170 offline checks that need no network or API keys.
Not recruiting, but glad to compare notes with anyone measuring what an agent refuses rather than only what it produces.