This interview is a GOLDMINE with Lenny Rachitsky and Chip Huyen. Popped it on and found myself actually really really engaged. We’re seeing more and more companies lean into developing their product for and with AI, so thought it was something we could all learn from.
https://www.youtube.com/watch?v=qbvY0dQgSJ4
Big Ideas:
- Most “AI problems” aren’t AI problems. They’re usually UX, communication, or data quality issues.
- Top performers gain the most from AI. Skilled people use AI to go faster—low performers rely on it without understanding.
- Data prep beats database choice. Structuring and cleaning your data drives bigger gains than choosing new infrastructure.
- User feedback > new models. The best improvements come from understanding users, not chasing the latest tech.
- Fine-tuning last. Fix prompts, scripts, and data first, fine-tuning adds costly maintenance.
- Good enough often wins. Aim for value, not perfection; 80% working now can beat 100% later.
- AI impact is hard to measure. Productivity gains are fuzzy, many still prefer hiring people over AI tools.
- Tool power ≠ clarity. Even with advanced AI, most people struggle to know what to build, start by fixing real daily frustrations.