ScaffoldBench: An Opinionated Local LLM Benchmark
I got tired of hot takes about local models from people who don't run them. So I built a benchmark that tests what I actually do - surgical edits in real codebases, on my own hardware, graded deterministically.