AI and the future of CAD

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

Topic 12 of 23

Published May 17, 2026 · Clip 55:03–1:00:29 (5:27)

Caitlyn explains that most hardware work happens in 3D CAD — designing parts and assemblies, verifying vendors can produce them at tolerance, and integrating. AI doing real CAD is at the very beginning: Claude can produce surfaces or point clouds, but real CAD is dense, equation-backed solid geometry. AI is already routing PCBs and doing basic component selection, though day-to-day mechanical and electrical engineering isn't yet AI-driven; today she uses AI for strategy, planning, dependency analysis, vendor research databases, and Excel, which meaningfully speeds up design. For AI to do real CAD, she argues LLMs and video models lack understanding of friction, weight, contact, pressure, and texture — she suspects world models or new model types are required, framing her 'healthy frustration' as wanting 'codecs for engineering,' and notes World Labs and Gemini will be part of the solution but not all of it.