GitHub's 2026 Octoverse data shows agentic coding reshaping activity

GitHub published its annual Octoverse report at GitHub Universe on October 28, and the headline numbers are steep. Merged pull requests reached 220 million a month, roughly triple a year earlier. Commits hit 7.38 billion a month, also more than triple, with most of the jump in the second half of the year. New repositories doubled to 30.3 million a month. The charts run from 2023 to 2026 and show a flat, low line that bends sharply upward in 2026. Simon Willison, who was at the event, calls it "a pretty stark representation of the impact of agentic engineering on GitHub."

The methodology is worth reading alongside the numbers, because it shows how much the measurement itself now leans on AI. Country rankings count pushes to public repositories but adjust for VPN bias by using a language model to pull country codes from user profiles. Industry breakdowns sort repositories into economic sectors with a fine-tuned embedding model. A category for "interoperability or reverse engineering" counts repositories whose names, descriptions, or READMEs mention related terms. The raw counts are blunt, since a coding agent can open many small pull requests where a person would open one, so tripling does not map cleanly to three times the work. Willison's link post points to the charts rather than adding deeper analysis.

Why it matters

If you manage engineers or review code, these curves are the clearest public signal yet that agent-generated commits and pull requests are flooding in, which changes what review and CI have to absorb. Treat the totals as activity, not finished output, when you plan capacity.