Commit

🚀 Small models, smarter instructions. Can a well-designed AGENTS.md make small local coding models more reliable? This repo puts that idea to the test with a 50-task benchmark, finding that **short, concrete rules often outperform complex workflows** — with the best version passing 46 out of 50 tasks. A great reminder that better AI agents aren't always about bigger models. Sometimes, it's about better guidance. 🔗 https://github.com/alihanafi-me/small-model-agents-mdalihanafi-me/small-model-agents-md