A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
Browse cluster: Cluster 61 →This repository provides a single `CLAUDE.md` file containing behavioral guidelines designed to improve how Claude Code (an AI coding assistant) approaches software development tasks. The guidelines, derived from observations by Andrej Karpathy about common LLM coding mistakes, address four key principles: thinking before coding to surface assumptions and tradeoffs, prioritizing simplicity over overengineering, making only surgical changes to existing code, and defining verifiable success criteria to enable independent looping. The project is maintained by an individual contributor and can be installed either as a Claude Code plugin or merged into project-specific instruction files, with additional support for use in Cursor.