@axichuhai: Found an open-source tool that makes Claude even better — just a 65-line Markdown file, already reached 130k+ stars on GitHub. Inspired by Karpathy's observation that models like to guess when uncertain, write overly complex code, and quietly change things they shouldn't touch...
Summary
An open-source 65-line Markdown file that summarizes 4 programming rules for Claude to improve code quality and save tokens. It has garnered 130k+ stars on GitHub, inspired by Karpathy.
View Cached Full Text
Cached at: 05/20/26, 08:48 AM
Found an open-source tool that makes Claude even better to use
Just a 65-line Markdown file, already racked up 130k+ stars on GitHub
Inspired by Karpathy’s observation that models tend to guess when uncertain, write overly complex code, and quietly change things they shouldn’t.
This project distills AI programming into 4 golden rules:
- Think before you act, ask if unsure, never blindly guess
- Solve problems with the least code possible
- Make precise changes—don’t touch a single line beyond what’s needed
- Write test cases first, then get them passing
With these constraints, Claude writes more stable code and saves tokens in the process.
Axichuhai (@axichuhai): Holy crap, a programming guru distilled his engineering experience into an open-source project
Hit #1 on GitHub trending, star count shot past 90k+
The author is former Vercel engineer Mat, involved in early Next.js development, known for explaining complex technical topics in plain language.
He packaged his day-to-day engineering practices collaborating with Claude into 16 skills, installable with one command, injecting decades of software engineering experience into AI.
Similar Articles
@FakeMaidenMaker: Definitely bookmark this tool that cures Claude Code's 'messier as you write' problem — GSD. Its predecessor Get Shit Done racked up 64K Stars on GitHub, and it's now migrated to GSD Core for continued maintenance. Let Claude Code tackle a slightly larger task…
GSD Core is an open-source workflow framework that solves the problem of context pollution and code quality degradation in AI coding tools like Claude Code during long tasks through a five-step cycle of 'Discussion → Planning → Execution → Verification → Delivery' and sub-agents with independent contexts. It has gained 64K Stars on GitHub.
@axichuhai: Found a project on GitHub specifically teaching how to use Claude Code, now up to 4w stars. It's not an official document, but a complete learning path from entry to advanced agents. Slash commands, MCP configuration, hook scripts, sub-agent orchestration — each module has visual diagrams, teaching you "how" and also "why."
Introduces a GitHub project with 40K stars for learning Claude Code, offering a complete learning path from beginner to advanced agents, visual diagrams, and reusable templates.
@wanerfu: Over 136,000 stars on GitHub! Designers really should check out this Claude Code skill. Graphify can turn any folder directly into a knowledge graph. Give Claude a folder, then connect 6 repositories, and it’ll stop churning out those same old purple-gradient AI websites…
The article introduces a toolchain for design workflows centered on Claude Code, with a focus on Graphify—which can turn any folder into a knowledge graph. Combined with tools such as Figma MCP, Playwright MCP, and impeccable, it helps AI-generated web pages break free from cookie-cutter designs.
@Manorgw: Claude Code is really nice for small projects, but once you move on to complex ones, context, task breakdown, and debugging can easily turn into a complete mess. Recently, I came across a pretty well-organized open-source resource: claude-code-best-practice. Common Claude Code…
A post recommending the open-source resource repository claude-code-best-practice, which compiles Claude Code’s core concepts, advanced features, 10 workflows, cross-model collaboration approaches, and 83 practical tips—a useful guide to avoiding pitfalls when developing complex projects.
@Ryrenz: Discovered another open-source gem that truly lets Claude Code understand legacy projects—Understand Anything. GitHub stars have surged to over 80,000. Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more can all integrate.
Understand Anything is an open-source tool that uses a multi-agent pipeline to analyze codebases and build interactive knowledge graphs, helping developers understand complex projects. It integrates with AI tools like Claude Code, Codex, Cursor, and has gained over 80,000 stars on GitHub.