Claude Code just hit 15M commits — 90% go to repos nobody uses

A community-built dashboard tracking every public commit signed by Anthropic's Claude Code has sparked a fierce debate on Hacker News: 90% of Claude-generated code lands in GitHub repositories with fewer than 2 stars. That sounds alarming — until you realize 98% of all GitHub repos have fewer than 2 stars anyway.

The real story isn't about quality. It's about a fundamental shift in why people write code in the first place.

The Numbers Are Staggering

The claudescode.dev dashboard tracks Claude Code's public footprint across GitHub in a rolling 90-day window. Here's what it shows right now:

To put this in context: according to SemiAnalysis research, Claude Code already accounts for 4% of all public GitHub commits as of early 2026. At current growth rates, it could reach 20% of daily commits by year's end.

Why "Nobody Uses It" Misses the Point

The Hacker News headline — "90% of Claude-linked output going to GitHub repos with <2 stars" — went viral with 237 upvotes. But the top-voted comments quickly dismantled the sensationalism.

The most upvoted reply pointed out the base rate fallacy: roughly 98% of all GitHub repositories have fewer than 2 stars. By that measure, Claude Code's output actually performs better than human-created repos on average.

But the deeper insight came from developers sharing how they actually use AI coding agents:

In other words, GitHub is no longer just a portfolio for open-source projects. For AI-assisted developers, it's becoming a personal scratchpad — a place to store one-off tools, quick automations, and personal utilities that were never meant to have an audience.

The "Audience of One" Revolution

This is the part that matters for everyone — not just developers. When AI makes building software nearly free, people start creating tools just for themselves.

Think about it: before Claude Code, building a personal expense tracker or a custom file organizer meant hours of work. Now it takes minutes. The result? Thousands of tiny, personal projects that solve one person's specific problem — and never need a single GitHub star to justify their existence.

The language breakdown tells a similar story. TypeScript leads with 34.8% of all Claude Code commits, followed by Python at 18.9% and JavaScript at 10.2%. These are web and automation languages — the tools of people building practical, everyday solutions.

Stars Are Dead — Usage Is What Counts

Multiple HN commenters pointed out that GitHub stars have been "useless as signals for project quality for a while." Stars measure popularity and discoverability, not whether software actually works or solves a problem.

The hidden iceberg is even bigger: enterprise and private repository usage likely dwarfs what's visible on public GitHub. The 15.8 million commits we can see are just the tip.

SemiAnalysis's analysis suggests the shift goes beyond coding. As AI handles longer autonomous tasks — from minutes to hours to days — its role expands from "helpful assistant" to "primary process handler" across research, analysis, and report writing.

If You Use Claude Code

You can explore the live dashboard yourself:

# Visit the live dashboard
https://claudescode.dev

# Check your own Claude Code commits
git log --author="Claude" --oneline | wc -l

The dashboard updates in real-time, showing a live feed of Claude Code commits rolling in from repositories worldwide. Whether you see this as an exciting democratization of software or a flood of low-quality code probably says more about your perspective on who should be allowed to build things.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News