Memdoor workflow with five Qwen3.8-27B models matches Fable 5 on LiveCodeBench Hard

A multi-agent programming harness written as a Memdoor workflow matches or exceeds the performance of the authors' GVS5H method on LiveCodeBench Hard problems using Qwen3.8-27B models. Rebuilt using exact prompts and loops, the harness scored a mean of 9.8 out of 10 across five passes on the first ten problems, solving 49 of 50 runs. The workflow operates with fresh context per agent, file-based communication, and a single repair round without human intervention. Open article →