InfluxDB co-founder Paul Dix ran AI coding agents on real production tasks and documented what shipped, what failed, and why he temporarily returned to writing code by hand. This is a first-hand account from someone with deep engineering judgment, not a blogger running toy demos.
The update buried in the episode description is the real hook: Dix reversed course and let the AIs write code again, but with significantly more oversight. That pivot, and the reasoning behind it, is what makes this worth 60 minutes of your time. His LinkedIn post 'Build the machine that builds the machine' and his prediction of '2026: the great engineering divergence' frame a thesis about where senior engineers either compound their leverage or become irrelevant.
The episode is on Changelog. The conversation is hosted by Adam Stacoviak and Jerod Santo. If you work in engineering leadership or are deciding how much autonomy to hand your AI toolchain, Dix's trial-and-error findings are more useful than any benchmark.
[READ ORIGINAL →]