Analysis: Claude Code currently authors 4% of all public GitHub commits and is on track to cross 20% of all daily commits by the end of 2026
SemiAnalysisDoug O'Laughlin
Context & Ripple Effects
Claude Code's reported commit share puts a concrete usage measure behind signs of widening developer interest, including Claude's more-than-doubled web audience in late 2025. It also arrives as GitHub has made Claude and Codex agents available inside its core developer surfaces through Copilot's agent integrations.
First-order effects
Claude Code becomes a material, measurable contributor to public GitHub activity, strengthening Anthropic's case that its coding agent is being used in production workflows rather than only evaluated.
Teams using the tool face a more immediate need to distinguish agent-authored changes from human-authored work in review and maintenance workflows.
Second-order effects
GitHub Copilot and other coding-agent vendors face pressure to compete on workflow integration, reliability, and controls—not just model quality—as agent-written code becomes more visible in repositories.
Higher agent contribution rates raise the value of tooling for code review, testing, provenance, and software-supply-chain defense; GitHub's embedded agent support is positioned to shape that workflow.
Third-order effects
If the projected share is reached, software development could shift from AI assistance toward supervision of agent-produced changes, making review capacity and accountability core constraints.
A growing share of machine-authored public code would make ecosystem-level security practices more consequential, particularly where downstream projects depend on fast-moving upstream commits.
The trend: Coding agents are moving from optional developer copilots toward a measurable layer of software-production infrastructure, with governance and verification becoming differentiators.
4% of GitHub public commits are being authored by Claude Code right now. At the current trajectory, we believe that Claude Code will be 20%+ of all daily commits by the end of 2026. While you blinked, AI consumed all of software development. Read more 👇 https://newsletter.semiana…
this chart feels like the those stats at the beginning of covid. “who cares about 400 cases in seattle? and why are all the epidemiologists buying toilet paper?”
@SemiAnalysis_ last half of this article is poor analysis: - makes assertion that 5.2 token inefficiency ruins long horizon planning? and yet it tops the METR chart for long horizon planning? half baked - counts NPM downloads as authoritative when Claude code numbers hugely infla…
I don't write any code manually. Everything is generated by Codex or Claude Code. At a certain point, you're taking a higher risk to allow humans to code than a machine.
for me the odds that AI is a bubble declined significantly in the last 3 weeks and the odds that we're actually quite under-built for the necessary levels of inference/usage went significantly up in that period basically I think AI is going to become the home screen of a
SemiAnalysis: “It [Claude Code] is set to drive exceptional revenue growth for Anthropic in 2026, enabling the lab to dramatically outgrow OpenAI.” [image]
Claude Code is the Inflection Point, What It Is, How We Use It, Industry Repercussions, Microsoft's Dilemma, Why Anthropic Is Winning. https://newsletter.semianalysis.com/ ...
Claude-using teams on @vercel: ▪️ generated 12.8% of deployments last week ▪️ ship 7.6x more often than non-Claude teams ▪️ growing deployments at 14% WoW
“4% of GitHub public commits are being authored by Claude Code right now. At the current trajectory, we believe that Claude Code will be 20%+ of all daily commits by the end of 2026.” “The cost of Claude Pro or ChatGPT is $20 dollars a month, while a Max subscription is $200
When we built the C++ app framework for building Windows Apps (called MFC, codename AFX) in 1990, one of the first things we did was make sure to stamp client executables with __AFXAPP so we could see. A proud moment a short time later was seeing that stamp on Mosaic. :-)
It seems obvious to a lot of people that something amazing is going to happen when this chart reaches 50% or 90% or whatever. But “authored by Claude Code” doesn't mean no human was involved. Claude code is an important new tool but it's not more than that.
This is exactly what happened with Deepseek last year. Fear of cost/token going down ≠ less demand for advanced compute Fear of AI agents being cheaper to run vs humans ≠ less demand for infrastructure layer software Long usage based bill companies $DDOG $NET $PANW $CRWD $RBRK [i…
this graph is actually breaking my brain with how much of a bubble i was in until literally last month lol i knew it was a bubble obviously but also, lol, lmao