GitHub's Octoverse 2026 report says agent-started pull requests hit 40% in September as Copilot Fusion launched multi-model orchestration.
Octoverse by the numbers
GitHub's Octoverse 2026 report, published as GitHub Universe opened, said developers logged 34.3 billion commits and 1.3 billion merged pull requests in the year through September 30, 2026, with pushes reaching 17.9 billion—more than 2024 and 2025 combined on recalculated definitions.
The headline workflow shift: more than 40% of pull requests started with an agent's first draft in September, up from 17% in March, totaling about 110 million agent-started PRs that month alone.
How agents change pull requests
Agent-involved PRs more often touched multiple files—about 87% versus 79% without detected agent involvement—and merge rates caught up. In February, 83% of agent PRs merged within two weeks versus roughly 86% for others; by September agent PRs reached 90% while the comparison group stayed near 86%.
Among Copilot monthly active users, 55% used agentic features, up from 35% a year earlier. Copilot's cloud agent mode, where tasks run independently, grew 230% year over year versus 85% for agent chat.
Copilot Fusion goes GA
GitHub announced Copilot Fusion general availability on October 28, 2026, combining Project HydraFusion orchestration with automatic model selection. Developers pick Fusion Standard, High, or Max effort tiers instead of manually swapping models mid-task.
GitHub cited Artificial Analysis Coding Agent Index results showing Fusion Max scoring 60, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task with roughly 40% less time and half the tokens. Workflows include single-model answers, cascade drafts with quality checks, and critique loops where one model reviews another's output.
Industry breadth and mining repos
Octoverse highlighted repositories tied to mining, construction, healthcare, and agriculture growing several times since 2022, suggesting agents spread beyond traditional software shops. Repos with agent configuration files averaged eight times as many commits as new repos without them.
GitHub noted 61% of Copilot users tapped more than one model in September, spreading requests across general, reasoning, and lightweight models as tasks demanded.
What engineering leads should do
The data implies review—not generation—is the bottleneck skill. Teams should invest in CI policies, ownership maps, and security scanning tuned for larger diffs.
Fusion's pitch reduces cost for high-reasoning tasks, but finance teams will still ask for per-team attribution as agent minutes replace keystrokes.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Language communities with slower agent adoption may see widening contribution gaps unless tooling democratizes access.
Teams that skip agent configuration may still see higher commit velocity from informal Copilot use without structured review gates.
Security reviewers worry multi-file agent diffs obscure authorship, complicating blame assignment for vulnerabilities.
Comments
Loading comments…