@Anubhavhing: A guy ran Karpathy's CLAUDE md across 30 codebases for 6 weeks. 🚨 Claude's mistake rate went from 41% → 11% Then he ad…

Link to Original Post :

Smart Ones are gonna read and save it.

Try using this https://github.com/ByteBell/bytebell-oss/… it uses LLMs to analyze your files.

3% with 8 rules sounds too clean but the karpathy core 4 are legit. 'ask before guessing' alone changed my error rate on greenfield work, claude's default is to confabulate when uncertain

Sign up, activate promo 9x2026 and start opening CS2 cases

I have pushed it further with my charter inspired by Karpathy. The codebase has some eval for benchmarking but I’ve been using it heavy on my codebases (I manage around 6 very large AI driven codebases in closed source and a few open source ones) and so far very satisfied with