Claude Code Dominates 90% of Repositories with Zero Popularity: Analysis of the claudescode.dev Dashboard
The claudescode.dev dashboard has recorded over 20 million commits in 1.08 million public repositories since launch. These are Claude Code-tagged commits, generating around 50 billion lines of code. Key insight: 90% of repositories have fewer than two stars, accounting for 90% of the code volume. Only 15% of commits are in projects with noticeable popularity. Activity is growing exponentially: since October 2025, with a peak in February 2026. It logs 115,000 new repositories with their first commit each week.
Top languages by commit share:
- TypeScript: 34.8%
- Python: 18.9%
- JavaScript: 10.2%
Top repositories by commits—automation-hub, XTeeem-OLD, aragora—are unknown to the broader audience. Claude Code generates 4% of all public GitHub commits, according to a SemiAnalysis estimate for February 2026.
Star Distribution and GitHub Structure
90% of repositories with <2 stars isn't a Claude Code anomaly—it's a reflection of the overall GitHub ecosystem. Most projects stay personal or experimental. The dashboard data backs up the hypothesis: AI coding is used as a personal tool for:
- Pet projects and prototypes.
- Learning and code experiments.
- Automating routine tasks without open source ambitions.
Hacker News critics point out: the share of low-star repositories aligns with the platform's general stats. This indicates widespread Claude Code adoption in non-production scenarios, where popularity metrics take a backseat.
Internal Use at Anthropic
Anthropic CPO Mike Krieger claims that 90–95% of Claude Code's code is written by the tool itself. Redwood Research's independent assessment revises that to 50% in merged commits. The Claude Code repository has racked up 81.6k stars, showing strong interest in the company's official project.
The disconnect between public data and internal practices underscores AI's maturity in enterprise development. The sheer volume of commits (20+ million) signals a shift from experiments to everyday routine.
Key Takeaways
- 90% of Claude Code-generated code is in starless repositories, mirroring GitHub's overall structure.
- Exponential growth: 115k new repositories per week.
- Top languages—TypeScript (34.8%), Python (18.9%), JS (10.2%).
- At Anthropic, AI writes up to 95% of its own code (per CPO), realistically 50% (Redwood).
- 4% of all GitHub commits from Claude Code.
Prospects for AI Coding
The data shows a shift to mass adoption but doesn't gauge quality. Low stars don't mean low value: for mid/senior developers, it's a tool for speeding up iterations in private projects. The big question—will Claude Code integrate into high-activity production repositories? Current metrics point to dominance in personal dev workflows.
— Editorial Team
No comments yet.