How AI Rewrote the PM Interview: Evals and 'Show What You've Built'
aakashgupta · x · 2026-09-26
A PM who got hired at Google, Epic Games and Affirm reflects that the interview questions that once landed him offers would no longer get him through today's loop.
- Behavioral hasn't changed — still the least-prepared round where offers die.
- Product sense now sounds like "the model is wrong 15% of the time, what ships" instead of "design a product for X."
- Analytical got tougher: picking a metric is just the start; the follow-up is what would make you drop it.
- Technical is completely different: three years ago API basics, today evals — how you built the golden set, what qualifies as a regression.
- Plus a new round: show me what you've built — the only one that rewards having done the thing, not preparing.
More from Companies & People
- Product is no longer the moat: AI winners will be decided by brand and attention — nikvassev · 2026-09-26
- Runway adopts an 'official clock' and puts all 7 offices on the same 'NOW' time zone — c_valenzuelab · 2026-09-26
- VC: new AI Systems of Record are greenfield plays redefining SoR, not replacing incumbents — matt_slotnick · 2026-09-26
- Build something at least once: creating teaches more about people than formal education — i_sudeepa · 2026-09-26
- Post-ROSCon hackathon kicks off, forcing ROS build farm to scale up — chrismatthieu · 2026-09-26
- Unverified: OpenAI halted training of its top upcoming model after Sept 20 incident — ccerrato147 · 2026-09-26