No Priors• Apr 3, 2026• 29:25Interview

AI for Atoms: How Periodic Labs is Revolutionizing Materials Engineering with Co-Founder Liam Fedus

From No Priors

Liam Fedus•Co-Founder, Periodic Labs

Executive Summary

Liam Fedus, former VP at OpenAI and co-creator of ChatGPT, has founded Periodic Labs, an AI foundation lab focused on applying AI to the physical world, specifically materials science and chemistry.
Periodic Labs employs a hybrid AI architecture, using large language models as an orchestration layer to direct experiments and coordinate specialized, symmetry-aware neural nets designed for atomic systems.
The company's strategy is to create a 'closed loop' system where AI directs physical experiments, generates novel data, and learns from the results, aiming to accelerate physical world development by one to two orders of magnitude.
The initial business model is to be a software provider offering an intelligence system for R&D, leveraging existing coding models and commoditized robotics to focus innovation on their core scientific AI.

12 quotes

Concerns Raised

The feedback loop for scientific discovery is inherently slower than for purely digital domains like software engineering.
Generating sufficient high-quality, diverse experimental data is a critical bottleneck for training performant models.
Current robotics are not yet general-purpose, requiring careful design and slowing down the automation of unstructured lab work.

Opportunities Identified

Accelerating development in the physical world by one to two orders of magnitude.
Creating a new paradigm for R&D and advanced manufacturing through an AI-driven intelligence system.
The increasing reliability and capability of general-purpose robotics will be a massive accelerator for this approach.
Applying AI to unlock breakthroughs in high-value sectors like semiconductors, aerospace, and energy.

Key Themes

AI for the Physical World

The core thesis is that the next frontier for AI is bridging the gap between the rapidly accelerating digital world and the slower-moving physical world. By creating an 'AI foundation lab for atoms,' the goal is to apply AI's power to materials, chemistry, and manufacturing.

This represents a shift from AI applications in language and code to solving fundamental bottlenecks in industries like energy, semiconductors, and aerospace, potentially unlocking transformative real-world advancements.

The Closed-Loop System for Scientific Discovery

Unlike purely digital domains, scientific progress requires interfacing with reality through experiments. The strategy involves creating a closed feedback loop where AI proposes experiments, automated systems execute them, and the resulting data is used to refine the models, creating a self-improving discovery engine.

This operational model is crucial for overcoming the data scarcity problem in specialized physical sciences and represents a new paradigm for R&D, moving from human-led discovery to AI-driven, high-throughput experimentation.

AI Systems Architecture for Science

The technical approach is not a single monolithic model but a system of systems. Large language models act as a general-purpose 'orchestration layer' that can reason, ingest literature, and call upon specialized, highly-efficient neural nets as tools for specific tasks like modeling atomic systems.

This modular, hierarchical architecture is an emerging best practice for applying AI to complex, multi-modal domains, combining the general reasoning of LLMs with the precision of domain-specific models.

Leveraging the AI Stack

Periodic Labs strategically chooses not to innovate in areas where powerful tools already exist. They leverage existing coding models (like Codex and Claude) and use off-the-shelf, commoditized robotics, allowing them to focus their resources on the core challenge of building AI for atomic-level discovery.

This demonstrates a capital-efficient strategy for vertical AI startups: build on top of the foundational AI stack to attack specific, high-value problems without reinventing the wheel.

The Physicist-to-AI Researcher Pipeline

The conversation highlights a trend of high-energy physicists moving into AI research. The hypothesis is that after the discovery of the Higgs boson, physics became bottlenecked by experimental apparatus, and the principled, quantitative skill set of physicists found a new, high-leverage frontier in AI.

This provides context on the talent and intellectual culture shaping the AI field, suggesting that the rigorous, first-principles thinking from physics is a key influence in the development of advanced AI systems.

Get started free

Topics

AI for Science Materials Science Chemistry Foundation Models Periodic Labs Liam Fedus OpenAI ChatGPT Google Brain LLM Architecture Orchestration Layer Closed-Loop Systems Scientific Discovery Lab Automation Robotics Data Generation Compute Costs

Processed Apr 10, 2026 yt-dlp + mlx-whisper + Gemini