GitHub’s Octoverse 2026 report, published on 28 October, says that in September 90% of agent pull requests merged within two weeks, while the comparison group of pull requests without detected agent involvement held at about 86%. In February, 83% of agent-involved pull requests merged within two weeks, a touch behind the roughly 86% for the rest, according to the report.
By our arithmetic, that is a 7-point rise for agent pull requests in seven months, from 3 points behind the comparison group to 4 points ahead. The same report says more than 40% of pull requests on GitHub started with an agent’s first draft in September, up from 17% in March, or 110 million in September alone. On Microsoft’s earnings call the same day, Satya Nadella said credit consumption from agentic features on GitHub “has more than doubled since usage-based billing went into effect”.
Short explainer · GitHub agent pull requests
A vertical news Short narrated from the published headlines and opening lines of this piece and a related FourWeekMBA piece on GitHub Copilot Fusion.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on how AI models and the harnesses around them fit together. It teaches the general idea only and does not discuss GitHub, Microsoft or this report.
The key insight: As we read it, GitHub’s report shows agent pull requests merging within two weeks more often than the rest, 90% against about 86% in September, in the same month that agents started more than 40% of pull requests.
GitHub’s report measures how often a pull request merged within two weeks. It says: “In February, 83% of agent-involved pull requests merged within two weeks, a touch behind the roughly 86% for pull requests without detected agent involvement”. It then says: “By September, agent pull requests were at 90%, while the comparison group held at about 86%”.
The report also says agent-started pull requests more often change multiple files: “about 87%, compared with 79% of pull requests without detected agent involvement”. GitHub describes the agent signal as a floor, “since not all agent use may be detected”.
Share of GitHub pull requests merged within two weeks, February and September 2026: agent pull requests 83% to 90%, pull requests without detected agent involvement about 86% in both months. Source: GitHub, Octoverse 2026.Pull requests started by an agent (March to September 2026) and GitHub Copilot monthly active users on agentic features (a year ago to now). The growth bound in the note is our arithmetic. Source: GitHub, Octoverse 2026.
What Was Posted on X
Microsoft-owned GitHub says AI agents now start over 40% of pull requests, up from 17% in March.
That’s 110M agent-started PRs in September alone.
Developers pushed code 17.9B times over the past year, more than in 2024 and 2025 combined, per its Octoverse report.
Wall St Engine posted GitHub’s headline Octoverse figures on X at 11:08 UTC on 29 October 2026: agents now start over 40% of pull requests, up from 17% in March, 110 million agent-started pull requests in September, and 17.9 billion pushes over the past year. This piece checks those figures against GitHub’s own report and adds the report’s merge rates, which the post does not mention.
How Much of the Work Starts With an Agent
GitHub says that “By September 2026, more than 40% of pull requests on GitHub started with an agent’s first draft, up from 17% in March”, and that “That is 110 million in September alone, about 6x March’s volume”.
By our arithmetic, about 6x March’s volume puts March near 18 million agent-started pull requests. Because the agent share rose about 2.4 times while agent volume rose about 6 times, all pull request starts grew at most about 2.5 times between March and September; GitHub gives the share as more than 40%, so this is an upper bound.
Among GitHub Copilot’s monthly active users, 55% now use agentic features, up from 35% a year ago, the report says. It adds that Copilot cloud agent, where a developer assigns a task and an agent works on it independently, “grew 230% year over year, against 85% for agent chat”.
Where Microsoft Bills for the Agent Work
Nadella said on the 28 October call that “The number of Copilot customers actively consuming credits through usage-based billing grew over 75 percent over the quarter”, and that “MAI Code Flash had 3 million users in GitHub Copilot last month alone”. Our earlier piece on Microsoft taking Copilot past per-seat pricing covers those billing changes.
Chief financial officer Amy Hood said M365 Commercial cloud results “were ahead of expectations driven by stronger GitHub Copilot adoption and usage”. She said ARPU growth came from premium seat-based offerings, “with additional contribution from usage-based offerings, primarily GitHub Copilot”. GitHub launched Copilot Fusion, which picks the model for each step, on the same day.
Which Languages Agents Touch Most
GitHub charted the share of merged pull requests with detected AI involvement by language. It says TypeScript, Python and Rust, “the fastest-growing communities, have the highest shares, about 40% to 44% of merged pull requests”, while Java, C# and C++ “sit lower, between about 21% and 34%”.
Among repositories created in 2026, the report says TypeScript was the primary language in 35% of those with an observed agent signal, compared with 24% without one.
The Structural Read
As we read the report, the share of agent work and its acceptance moved together: agent-started pull requests went from 17% to more than 40% of starts between March and September, while the two-week merge rate for agent pull requests went from 83% in February to 90% in September.
By our arithmetic, at about 6 times March’s volume, September’s 110 million agent-started pull requests compare with roughly 18 million in March.
As we read Microsoft’s remarks, the billing side tracks the same usage: Nadella tied GitHub’s agentic credit consumption to usage-based billing, and Hood named GitHub Copilot as the main usage-based contributor to ARPU growth.
GitHub, Octoverse 2026
“For a growing share of work, the first move is now directing an agent, and review is how developers shape what comes back”
Three Implications
FOR MICROSOFT INVESTORS Hood said usage-based offerings, primarily GitHub Copilot, added to ARPU growth alongside premium seat-based offerings; the prepared remarks we read give no GitHub Copilot revenue figure in dollars.
FOR ENGINEERING TEAMS GitHub reports that agent-started pull requests change multiple files about 87% of the time, against 79% for pull requests without detected agent involvement.
FOR MODEL MAKERS GitHub reports that Copilot cloud agent, where an agent works on an assigned task independently, grew 230% year over year, against 85% for agent chat.
The framework says: “This is the market’s central misread: adoption looks early by seats and nearly complete by output.”
As we read it, GitHub’s figures are an output measure of that kind: the share of pull requests started by agents and how often they merge, rather than a count of seats. The framework does not discuss GitHub’s Octoverse data.
What Is Not Established
GitHub’s merge figures measure whether a pull request merged within two weeks; the report does not say how the merged code performed afterwards. The February figure covers agent-involved pull requests and the September figure agent pull requests, in GitHub’s wording, and the comparison group is given as roughly or about 86%.
The report counts only agents GitHub can detect. Neither the report nor the prepared remarks we read give GitHub Copilot revenue in dollars or the dollar value of the credits consumed. The 2.5 times and 18 million figures are our arithmetic from rounded numbers in the report.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
GitHub’s Octoverse 2026 report says agent pull requests merged within two weeks 90% of the time in September, up from 83% in February, while the comparison group held at about 86%. More than 40% of pull requests started with an agent in September, and Nadella said credit consumption from agentic features on GitHub has more than doubled since usage-based billing went into effect.
A note on sourcing. The merge rates, pull request shares, Copilot usage figures and language shares come from GitHub’s Octoverse 2026 report, published on the GitHub blog on 28 October 2026 and read in full on 29 October 2026. Satya Nadella’s and Amy Hood’s remarks come from Microsoft’s official prepared remarks for its FY27 first-quarter call on 28 October 2026. The 7-point rise, the March volume and the growth bound are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
