The state of the tech industry in 2026
Gergely Orosz · 6 October 2026
Read the originalGergely Orosz of The Pragmatic Engineer reports that AI agents wrote more pull requests than people on GitHub in August, and that human code review is turning into a formality.
Gergely Orosz writes The Pragmatic Engineer, a newsletter for software engineers and engineering leaders. He published the notes from his keynote at the LDX3 engineering leadership conference in New York. They draw on visits to OpenAI, Anthropic and startups, and on unpublished data from GitHub, Linear and Factory AI. He lists what AI agents have changed in software teams, what they have broken and what has stayed the same.
- GitHub data shows that pull requests written by AI agents rose ninefold in eight months and outnumbered those written by people in August 2026.
- An unnamed engineer at a mid-sized startup said reviewers facing this volume often approve changes without reading them. Linear, a project-tracking tool, reports more pull requests reviewed only by AI agents.
- Katelyn Lesse, head of engineering for Claude Platform at Anthropic, said one project rarely has more than two engineers, because each already runs several AI agents.
- Orosz found no company where product managers or designers ship code to production, and engineers still decide what is released.
Orosz predicts that engineers will stop reading agent-written code, and cites Charity Majors, chief technology officer of Honeycomb, asking what would make shipping unread code acceptable. That a team should answer her question before its reviews become a formality is an inference and is not the author's claim. The slides are for paid subscribers only.
Keep engineers responsible for deciding what is released to production, even when product managers or designers build features with AI agents.
Have AI agents write the tests as well as the code, and keep spending about as much effort on tests as on the code itself.
Derived by Working Surface from the article; more in the Playbook. Source line: But it’s still engineers who are responsible for deciding what can and what cannot be released to production.
8 October 2026: as AI agents produce code and prototypes faster than people can check them, teams must decide where human checks and decisions sit instead of letting review quietly lapse.