Hardware principles and KPIs

Why the next AI boom is physical AI | Caitlin Kalinowski (ex-OpenAI, Meta, Apple)

Topic 8 of 23

Published May 17, 2026 · Clip 33:09–39:54 (6:45)

Caitlyn's core advice: define goals and KPIs early and change them as little as possible, because hardware can't absorb mid-course pivots — a $300 target becoming $150 burns early time. Each iteration costs three-to-five months, timing against competitors can be worth millions, and engineers never feel done, so written goals determine when you can ship. She urges designing the hardest, riskiest parts first — illustrated by an architect routing cables through a laptop hinge before finalizing it — and over-iterating on the parts users touch most, like trackpads and keyboards. Finally, she prescribes doing known work immediately — 'ruthless efficiency' learned from Shelley Goldberg and Kate Bergeron at Apple — then explains KPI buckets: pixels-per-degree in VR, clock speed, weight, price, and features on computers, citing the MacBook Air dropping the ambient light sensor and Elon Musk's practice of quantifying trade-offs like grams versus cost.