Everything is ugly, so go build something that isn't
summary
Builds NotebookLM (originally Tailwind) from the ground up, ships an auto‑summary that makes 50‑file workflows usable, and finds 90% of first queries are summarize tests so failing that costs users immediately. Shows a post‑launch lesson from Hux where several hundred testers only use one feature. Argues for personal clarity and ruthless focus: nail deterministic UX before probabilistic delights, expose model failure modes to earn trust, and prefer a single excellent outcome over a kitchen‑sink product.
AI-written summary · about
chapters (14)
- 0:00The new era of AI work
- 2:54Building products with AI
- 5:11Navigating chaos and opportunity
- 6:52Cultivating personal clarity
- 8:56Story of NotebookLM
- 10:33Focusing on the job to be done
- 12:06Building trust through purpose
- 13:16Exposing the model's edges
- 15:18Nailing the fundamentals
- 16:52The opportunity to delight
- 18:22Surfacing delight with agency
- 20:16Shipping a point of view
- 22:15The importance of focus
- 24:26Conclusion
related talks
36m"UX for Developers, Seriously!"
44mIt’s Easy To Create A Good-Looking Product, But How About A Useful One?
41mIt’s easy to create good looking products, but what about useful ones
37mBest Practices for Accelerating Delivery
01h 06mOnly One Deliverable Matters
18mBuilding pi in a World of Slop
30mLightning Talks
40mPanel Discussion: What Users Really Want