InfluxDB co-founder Paul Dix ran AI coding agents on real production tasks, pulled back to manual coding, then returned to AI-assisted development with significantly more oversight. That full cycle, including what shipped to prod and what did not, is the substance of this episode.
The conversation goes beyond surface-level AI enthusiasm. Dix details the specific failure modes that sent him back to writing code by hand, the oversight mechanisms he now requires before trusting agent output, and his broader thesis on what he calls 'the great engineering divergence' coming in 2026. His LinkedIn piece 'Build the machine that builds the machine' is required reading alongside this.
This is worth reading in full because Dix is not a commentator, he is a practitioner with a production codebase and a documented opinion change inside a single news cycle. The how and why of that reversal is the story.
[READ ORIGINAL →]