AI NewsProductivityAnnouncement
Agents start 40 percent of GitHub pull requests, up from 17 percent
GitHub said developers now start more than 40 percent of pull requests with an agent. In March that figure was 17 percent.

Image: GitHub
Why it mattersReview is now the biggest part of pull request work on GitHub. A team that keeps up with review decides how fast anything reaches production.
The first move on a GitHub pull request is now, more often than not, directing an agent. GitHub published its Octoverse 2026 report on 28 October. GitHub said that as of September, developers start more than 40 percent of pull requests with an agent. In March that figure was 17 percent. GitHub said review is now how developers shape what comes back.
A second number tells the same story. GitHub counted repositories created this year. The average repo set up with an agent configuration file logged eight times as many commits. That was compared against new repos without one. GitHub said that is a sign teams building with agents from day one are moving at a different pace.
The year's numbers
GitHub said developers pushed code to GitHub 17.9 billion times in the past 12 months. GitHub said that is more than 2024 and 2025 combined. Commits across all repository types reached 34.3 billion. GitHub said that is more than three times where they stood two years ago. Merged pull requests more than tripled, to 1.3 billion. That is an average of 3.5 million merged pull requests a day.
GitHub said activity climbed steadily from January, then surged around June. Around then, agents moved from single-session tools to persistent ones. GitHub said these could track progress across days and run other specialised agents. Agents also began running work in the cloud as well as on local machines.
Developers now pick from more than one model
GitHub said 61 percent of GitHub Copilot users used more than one model in September. GitHub said their requests spread across general purpose models, deep reasoning models, and fast lightweight models.
GitHub Actions ran a job 1,440 times every second this year. GitHub said that is 124 million jobs a day, more than three times last year's total. Developers merged more than 30 million Dependabot pull requests to keep dependencies current. GitHub said that is about double 2025's pace.
The platform underneath
Jay Parikh is Microsoft's EVP for Core AI. He wrote a separate GitHub Universe post on 28 October. Parikh said GitHub has added 4 million CPU cores so far this year. GitHub also added 160 petabytes of high-speed storage and 50 terabits per second of network capacity. GitHub is migrating to a multi-region deployment on Azure. Parikh said he expects that to finish early next year. Parikh said capacity alone was not enough, and GitHub also had to change its architecture.
Public contributions reached 2.5 billion in 2026. GitHub said that is more than double last year's total of 1.12 billion.
Where the growth is coming from
GitHub said the platform now serves more than 225 million developers. The United States has 38.3 million and India has 32 million. Each added more than 10 million developers this year. Brazil added 3.28 million developers, the third most of any country.
GitHub also sorted public repositories by industry for the first time, using the North American Industry Classification System. In administrative and support services, mining, health care and agriculture, more than four in 10 repositories mention AI. GitHub said agent-opened pull requests are rare in those sectors, under 3 percent. Construction stood out. GitHub said about a quarter of its repositories mention AI, but about 9 percent have agent-opened pull requests. GitHub said that is the highest share of any sector, with manufacturing next.
Source
GitHub: Octoverse, commits and pushes reach new highs in 2026 as developers build with agents. GitHub Universe 2026: Evolving every layer of the stack, by Jay Parikh.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.


