Your local LLM isn’t dumb. Your harness is.
中文摘要
模型规模并非决定一切。9B模型成功修复了30B模型无法解决的Bug，证明推理框架比参数量更关键。
English Summary
Model size isn't everything. A 9B model fixed a bug a 30B model couldn't, proving the inference harness is more important than parameter count.
Original Excerpt
A 9B model just fixed a bug that a 30B model couldn’t finish. Same codebase, same prompt, same machine. The 30B had more parameters and… Continue reading on Medium »