Liam Fedus, co-creator of ChatGPT and former VP of post-training at OpenAI, discusses his new company Periodic Labs, which is building an AI foundation lab for atoms — applying AI to materials science, chemistry, and physical world engineering. He explains how the company uses a closed-loop system combining physics simulations, experimental data, and LLMs as an orchestration layer over specialized neural nets designed for atomic systems. Fedus covers the data challenges unique to materials science (vs. internet-scale language data), the role of robotics as an accelerator, commercialization strategy as an intelligence layer for enterprises bottlenecked by materials engineering, and his views on AGI's domain-specific nature and recursive self-improvement timelines.