AI & Development
mini-swe-agent: 100 Lines That Beat Claude Code on SWE-bench
mini-swe-agent scores 74%+ on SWE-bench Verified with 100 lines of Python — and outperforms Claude ...











