mini-swe-agent - 100줄짜리 AI 에이전트로 GitHub 이슈 해결 및 커맨드라인 지원

mini-swe-agent demonstrates that sophisticated AI coding tasks can be accomplished with minimalist architecture—just 100 lines of Python without complex frameworks or tool-calling interfaces—challenging assumptions that advanced agents require heavy scaffolding. Achieving 74%+ on SWE-bench benchmarks, it serves as a pure baseline for evaluating LLM capabilities independent of framework complexity, providing critical insights as the AI agent ecosystem rapidly evolves. Broad adoption by major tech companies signals industry shift toward lightweight, production-ready agents that prioritize speed and interpretability over architectural complexity.