Where Autonomous Coding Tools Excel

New AI-powered programming tools like OpenAI’s Codex or Google’s Jules might not be able to code an entire app from scratch just yet. But when it comes to working with large amounts of previously written code, those products could soon be saving developers a ton of time.

For example, Codex excels at finding bugs in a large code base and suggesting potential fixes, according to Ryan Cox, who manages a team of roughly two dozen engineers at 250-person startup Temporal Technologies. Though the company hasn’t conducted a formal study, Cox estimates that the suggestions from Codex, which runs in the background without the need for human supervision, improved engineers’ productivity by at least “low single digits” in percentage terms by saving them from having to scour the code base themselves.