Claw Code Just Killed Claude Code...

TL;DR

Anthropic accidentally leaked 512,000 lines of Claude Code source, enabling a developer to rebuild it as the fastest-growing GitHub repo ever.

Key Points

• 1.Anthropic leaked 512,000 lines of Claude Code source by shipping debug files in their NPM package. On March 31, 2026, security researcher Shiaan Show found a 60MB file containing source maps — debugging tools that translate minified code into readable form — inside Anthropic's public Claude Code package, exposing their entire agent architecture.
• 2.Korean developer Secret Jin rebuilt Claude Code from scratch in under two hours, creating Claw Code. The repo hit 50,000 stars within hours and crossed 100,000 stars by Wednesday, becoming the fastest-growing GitHub repo in history — surpassing even OpenClaw — and now sits at over 164,000 stars.
• 3.Claw Code's key advantage is LLM flexibility, but it is not production-ready. Unlike Claude Code, it works with any LLM — GPT, Gemini, local models via Ollama — but crashes on complex refactoring, lacks rollback mechanisms, and has primitive memory management, making it a research tool rather than a daily driver.
• 4.Anthropic blocked Claude Pro and Max subscribers from powering third-party tools like OpenClaw starting April 4, 2026. Over 135,000 OpenClaw instances were running at the time, with ~60% using subscription credits; Anthropic cited infrastructure strain and a 5x cost gap between flat-subscription usage and actual API rates as justification.
• 5.The leaked code exposed Anthropic's secret roadmap, including an unreleased autonomous agent mode called 'Chyros.' Simultaneously, OpenClaw's creator Peter Steinberger left his project to join OpenAI — with Sam Altman publicly announcing he'd drive next-gen personal agents there — days before Anthropic enforced the subscription cutoff, raising competition concerns.