multica-ai/andrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
218K stars22K forks+2K today
What it does
This repository provides guidelines to improve the coding behavior of large language models, inspired by Andrej Karpathy's insights on common pitfalls. It aims to enhance model output quality through four key principles: thinking before coding, simplicity first, surgical changes, and goal-driven execution.
Star history
Tracking
- Last trending
- 2026-05-19
Creator kit
Hook
Learn how to make large language models code better with these principles derived from Andrej Karpathy's insights!
Content angles
- Create a video explaining the four principles in CLAUDE.md and their importance for AI developers.
- Write an article comparing Claude Code guidelines with other coding standards used in machine learning projects.
- Host a Twitter thread discussing real-world examples of how these principles can be applied to improve LLM-generated code.
Who should care
AI researchers, software engineers working on large language models, and developers interested in improving the quality of AI-generated code.