Tag
Dan Shipper announces that GPT-5.6 Sol Ultra is generally available, and shares a tweet from Ethan Knight claiming the model produced a proof of the Cycle Double Cover Conjecture using 64 subagents in under an hour.
A tweet thread discusses benchmark results where Qwopus Coder tops the leaderboard, while Cohere's North-Mini-Code-1.0 lands last on an agentic tool-calling board, showing surprising outcomes for smaller models.
Santiago Valdarrama discusses the challenge of integrating AI agents with user interfaces, noting that agents break the traditional request/response pattern.