GitHub says as much as 30% of new code on its network is written with its AI tool Copilot, and 50% of developers who tried it since July have kept using it
The open-source software developer GitHub says as much as 30% of newly written code on its network is being done with the help of the company's AI programming tool Copilot. Source: GitHub .
Context & Ripple Effects
Four months after Microsoft and OpenAI announced Copilot in June 2021, GitHub is publishing its first usage proof points: up to 30% of newly written code across its network involves the assistant, and half of the developers who tried it since July have stuck with it. The numbers matter because GitHub hosts much of the world's open-source code, making its network one of the few places where AI-assisted programming can be measured at population scale rather than in a vendor's pilot.
The trajectory since then validates the claim: GitHub took Copilot out of preview at $10/month after 1.2M+ developers tried it, reported 46% of code files generated with Copilot by early 2023, and Stack Overflow counted 54.8% developer adoption in 2023. This 30%-of-new-code figure was the first signal that a suggestion tool could become a primary authorship channel.
First-order effects
- Developers who adopted Copilot are already offloading roughly three in ten lines they write to the model, and the 50% trial-to-retention rate gives GitHub evidence the habit survives past novelty — directly supporting the paid tier it would later price at $10/month.
- Microsoft and OpenAI gain their first hard dataset on model quality in real coding work, letting them tune suggestions against actual acceptance behavior across every language on GitHub.
Second-order effects
- A measurable 30% authorship share on the dominant code host pressures rival IDE vendors and code-completion tools to ship their own assistants or cede the editor surface where developers spend their day.
- If AI-written code becomes a large slice of new commits, code review, security scanning, and licensing workflows built around human-authored lines face retooling — the concern behind GitHub's later push to make Copilot's suggestions more secure before scaling further.
Third-order effects
- The pattern that follows — 30% of new code here, 46% of files two years later, majority developer adoption after that — points toward AI assistance as the default mode of software authorship, with whoever operates the code-hosting network positioned to set terms for how code gets written.
- As machine-generated share grows, provenance and accountability questions move upstream into platform policy: whether hosts like GitHub require disclosure of AI-authored contributions becomes an industry-structure question, not just a tooling one.
The trend: AI pair-programming is consolidating from experiment to default development workflow, with the code-hosting platforms themselves capturing both the usage data and the subscription economics.