The talk examines why greater use of AI coding tools has not automatically made software teams ship faster. Its central distinction is between activity - such as producing code - and outcomes: working software delivered to users. It frames this as an activity-outcome gap, arguing that software engineering involves more than writing code, so faster code generation alone does not remove the work and constraints that stand between a change and a release.
The research behind the discussion includes 30 interviews about AI and software engineering, a study comparing productivity effects across generations of coding tools, and investigations into what AI enables developers to do, where they limit its autonomy, and how well agents detect bugs in code review. Together, these topics focus attention on the conditions that shape whether AI-generated work becomes usable, reviewed, and shipped - not simply on how much code developers produce. The talk is a 21-minute UC Irvine software engineering seminar from October 2026, accompanied by a concise handout and a full slide deck.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.