Start with the first five minutes
Before a roadmap, I want to watch someone sign up. Where they hesitate, what they misread, the step where they give up. Most growth problems are first-mile problems in disguise.
Let experiments argue
At Bolt I owned the booking funnel. Instead of debating opinions, we ran large-scale experiments and kept what moved: conversion up 15%, bookings up 24%, trip completion up 12%, cancellations down 7%.
Know where authority ends
With AI, being right most of the time isn't enough. I design what the system may do alone, what needs a human, and how it fails safely. Evaluation is part of the product, not a side task.
Teach it to learn it
Teaching product at BrainStation since 2023 keeps me honest. If I can't explain why a decision was right to a room of new PMs, it probably wasn't.