Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard(github.com)
418 points by guregodevo | 202 comments
GitHub README for team-solve, a Memdoor multi-agent workflow where five Qwen3.8-27B roles solve programming problems, with benchmark results on LiveCodeBench Hard versus Fable 5.