Writing

Autoresearch Service10 min readI ran many coding agents in parallel to see whether more compute produces faster code.Building an inference engine9 partsI built an inference engine from scratch to see what it actually takes to make a model fast.