How AI coding tools became the real programmers
The author describes how AI coding tools, especially Cursor, now generate all their code while they direct the work. Cursor reportedly found and fixed more than 20 potential race conditions in the author's multithreaded code during pull request review. The piece argues that tools like Cursor, Claude Code, and Codex have become stronger coders than humans, echoing calls to stop writing code by hand.
The author recounts a gym conversation with a newer CrossFitter, John, where he described his job as writing software but then recognized that Cursor now produces his code while he gives direction. At Rails World 2026, Ruby creator David Heinemeier Hansson reportedly encouraged developers to stop coding by hand.
At the author's workplace, Cursor is connected to pull request review. It identified over twenty possible race conditions in his multithreaded code and repaired them within minutes. The article also compares this shift to Hidden Figures, where human "computers" moved from manual calculations to programming, and notes 37signals' hope for native apps across platforms via coding agents.
This shift could affect software developers most directly, as their daily work may move from writing code toward specifying, reviewing, and validating AI-generated changes. Teams and customers might benefit from faster detection of subtle bugs, yet reliance on these tools may also alter skill development and create new review burdens. Businesses may gain speed and broader platform reach, while the profession could gradually redefine what programming expertise means.