Summary:

AI coding assistants are everywhere now, but how well do they actually work for open source? This talk shares lessons from using AI on two Linux networking projects: DPDK and iproute2.

The good: documentation, test writing, and debugging. The bad: awkward code suggestions, formatting errors, and false positives in code reviews. The ugly: trying to get AI to write new features like RSTP or FQ-CoDel — after too many failed attempts, I gave up and moved on to more useful things.

A frank look at where AI helps, where it wastes your time, and what to watch out for.

Bio:

Stephen Hemminger is a longtime Linux kernel networking developer, maintainer of iproute2, and contributor to DPDK. Now retired, he continues contributing to open source as a volunteer.