A 200-line ReAct harness lets open models race to fix 3 real bugs
MaziyarPanahi · x · 2026-07-25
A real bug-fixing setup shows how multiple open models can be raced against the same repository issues.
- The author used 3 hidden bugs in a real repo, with pytest as the judge.
- Local models ran on a Mac Studio: Bonsai-27B, Inkling, and gemma-4-26B, served via llama.cpp + MLX.
- A separate stronger model, Kimi K3, was called through OpenRouter until the weights become public.
- The agent loop is a roughly 200-line ReAct harness that can read / edit / runtests and talk to any OpenAI-compatible endpoint.
The post frames this as a model-agnostic, practical way to compare coding agents on the same debugging task.
Related event: Open-Source Models Fix Real-World Bugs in ReAct Framework Test(2 posts)→
More from coding & agent
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Astra storyboards plus Minimax H3 per-shot generation boost video success rates — Hailuo_AI · 2026-09-11
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- First-ever Three.js Conference lands in Paris, with a panel on AI-shortened design workflows — OdinLovis · 2026-09-11