Charity Majors on AI, Determinism, Instrumentation, & Eating Your Broccoli

<p>James Governor talks with Charity Majors, Co-founder and CTO of Honeycomb, about what AI is doing to engineering practices. Majors argues that non-determinism takes more discipline, not less, and every CEO who wants their AI cookies has to eat their broccoli first. Agents in production are finally forcing teams to do the instrumentation work they should have done years ago. They get into Chad Fowler's Phoenix architectures and the idea that the system, not the code, holds the truth about how software behaves. Also: which problems AI has genuinely solved and which it hasn't, why bragging about token spend is like driving a Hummer, and the last year of AI hype has been aimed almost entirely at individual productivity when the hard useful work is always about software development as a team sport. They discuss Honeycomb's Canvas, the company's AI-guided Observabiolity workspace. And finally they discuss the difference between performative outrage and actually artists in the AI era - and make some recommendations accordingly!<br /><br />Honeycomb is a RedMonk client, but this video is free an independent content.<br /><br />Show notes: <a rel="noopener noreferrer nofollow" href="https://redmonk.com/videos/charity-majors/" target="_blank">https://redmonk.com/videos/charity-majors/</a><br /><br />Chapters<br />00:00 Intro<br />01:51 Nobody has any change budget left<br />03:15 AI is forcing the discipline conversation<br />05:37 Broccoli first, then ice cream<br />06:01 Phoenix architectures, and why code isn't the source of truth<br />08:47 The problems AI hasn't solved<br />11:30 Where determinism meets non-determinism<br />13:29 Token maxxing, Goodhart, and the Hummer problem<br />15:34 Superpowers for one vs engineering as a team sport<br />19:00 npmx and doing open source in public<br />20:41 Honeycomb Canvas, and social debugging at 2am<br />26:22 AI norms and values<br />27:42 Writing is thinking<br />32:25 Work you stand behind vs feedback you can't outsource<br />35:06 The other bitter lessons<br />37:33 If you support artists, support artists<br />43:05 Performative anger and Bluesky<br />44:49 Knowledge management is back<br />46:27 Canvas, timeline analysis, and what's next</p>