GitHub COO: 14 billion commits expected in 2026 as agents flood platform

In an interview conducted by @hammer_mt for Every's AI & I, GitHub COO Kyle Daigle described the dramatic scale of agentic activity on GitHub's platform. Agents created 17 million pull requests in March alone, and while GitHub reached 1 billion commits last year, Daigle said the platform is on pace for 14 billion commits in 2026—a 14x increase he does not believe is plateauing. GitHub's stated philosophy in response to the surge is to keep open-source maintainers in control, avoiding any platform-level dictation of which PRs should be trusted or merged.

Daigle also discussed how AI is redrawing the boundaries of who counts as a developer: GitHub's own legal and finance teams are using Copilot to build apps, illustrating a broader collapse of the developer/non-developer distinction. On the business model side, he argued that per-seat pricing cannot survive a world where agents run autonomously around the clock, and that automatic model routing—swapping in cheaper models like Haiku for simpler tasks rather than always calling expensive frontier models—is the most viable path to sustainable economics. Separately, Daigle described a personal productivity practice: a daily AI loop he calls "Baxter," which reads seven days of his emails and Slack messages, flags communication patterns, and checks whether he followed the previous week's advice.

Key facts

1. 01Agents created 17 million pull requests on GitHub in March alone.
2. 02GitHub hit 1 billion commits last year and is on pace for 14 billion in 2026—a 14x increase.
3. 03GitHub's philosophy is to leave open-source maintainers in control of agent-generated PRs, not dictate trust or merge decisions.
4. 04GitHub's own legal and finance teams are using Copilot to build apps, blurring the developer/non-developer line.
5. 05Daigle argues per-seat pricing is unsustainable in an agentic world.
6. 06Automatic model routing—using lighter models like Haiku for simple tasks—is Daigle's preferred solution to agentic pricing economics.
7. 07Daigle runs a daily AI loop called Baxter that reads 7 days of emails and Slack messages, flags communication patterns, and checks adherence to prior advice.

Topics