GitHub Hits 7.38 Billion September Commits as Former CEO Launches Rival

GitHub says developers and AI agents made 7.38 billion commits on its platform in September 2026, more than five times the number a year earlier. The extraordinary volume is forcing the Microsoft-owned company to rebuild the infrastructure beneath the world’s most prominent code-hosting service. It is also a pointed backdrop to the arrival of Entire, a rival platform launched this week by former GitHub chief executive Thomas Dohmke.

The figures are not simply a measure of how much software people have written. A commit is a saved snapshot of changes, and an agent can make many of them while working on a single task. Still, the jump shows why GitHub’s annual Universe event on Wednesday focused as much on the mechanics of storing, reviewing and shipping code as on new AI features.

In an engineering account, GitHub put September’s total at 7.38 billion commits and said monthly pushes rose from 690 million to 3.35 billion in a year. A push sends one or more commits to the service. GitHub Actions ran 3.26 billion jobs in September, creating another layer of work every time code changes set off tests, builds or security checks.

The broader Octoverse 2026 report counts 34.3 billion commits and 17.9 billion pushes in the 12 months ending 30 September. Those annual totals include public and private repositories. They are different from the 7.38 billion figure for September alone, a distinction that matters when comparing the two.

GitHub connects the surge to agent-assisted development, although higher activity does not by itself prove that every change is useful or production-ready. Its report says more than 40% of pull requests in September began with an agent’s first draft, up from 17% in March. That makes review, testing and traceability more important, not less.

At GitHub Universe, the company said it has added four million CPU cores, 160 petabytes of high-speed storage and 50 terabits per second of network capacity so far this year. It is also moving towards a multi-region Azure deployment, expected to be completed early in 2027. More hardware can absorb demand, but GitHub says its architecture must change because the busiest repositories are receiving concurrent work from both people and fleets of agents.

The planned design separates storage from the compute capacity used to serve Git requests and reduces the coordination needed for each write. In internal benchmarks, GitHub says the architecture has delivered up to 35 times higher write throughput. That is an engineering target, not a promise that every developer will see a 35-fold speed-up. Reliability, access controls and familiar review rules still have to survive the migration.

GitHub is also trying to move Copilot from a coding assistant into a wider software-delivery interface. The new Copilot app brings together agent tasks across projects. Copilot projects, available in the app from 28 October, can run managed sessions across multiple repositories, while session continuity lets a developer follow the same cloud agent work from the app, VS Code, the command line or the web.

Copilot Fusion chooses among models according to a task’s needs, rather than requiring the developer to select a single model for every step. GitHub also introduced a deeper code-review option called Max and expanded automations that can propose improvements for human approval. For enterprise customers, it is offering access to Copilot features without a separate subscription through a usage-based AI credits arrangement, initially via its sales team.

A former chief executive builds a rival

The timing is awkward for GitHub because Dohmke, who left the company in 2025, introduced Entire’s Git forge on 26 October, two days before Universe. Entire offers native private repository hosting, distributed infrastructure, a review system called Trails and a memory layer intended to preserve the context behind agent-written changes. It can also mirror GitHub repositories, so adopting it need not mean an immediate break with GitHub.

As TechBooky reported in its earlier look at Entire, the start-up argues that traditional pull requests struggle when agents generate revisions faster than people can assess them. Entire’s own performance comparisons, including its claim of much faster pushes, are company benchmarks rather than independently established results for every workload. GitHub’s response is to rebuild its core platform while extending the review and governance tools teams already know.

That is the larger contest behind both announcements. AI agents can increase the supply of code, but teams still need to know what changed, why it changed, whether it passed meaningful tests and who approved it. GitHub’s September figure makes the scale of that problem concrete. The question now is which platform can turn the flood of changes into dependable software without making developers surrender control.