@JeffDean: Exciting results, @LiamFedus! Congrats to the whole team at Periodic Labs!
Summary
Jeff Dean congratulates Periodic Labs for exciting results in building high-throughput materials labs that integrate experiments with AI models using H200 GPUs to create a feedback loop for data-driven discovery.
View Cached Full Text
Cached at: 09/17/26, 06:20 AM
Exciting results, @LiamFedus! Congrats to the whole team at Periodic Labs!
Liam Fedus (@LiamFedus): We built high-throughput materials labs in Menlo Park to create a loop between experiments and models. The labs generate fresh data, the models learn from it, and then help us decide what to try next.
Using only 1,300 H200s, plus months of our experimental data, we mid-trained
Similar Articles
@eladgil: https://x.com/eladgil/status/2099916017510047788
Liam Fedus describes the setup of high-throughput materials labs in Menlo Park that create a loop between experiments and AI models, using 1,300 H200 GPUs and experimental data to mid-train models for next-step decisions.
@MTSlive: SITUATION DETECTED: Periodic Labs used 1,300 H200s, plus months of its own experimental lab data, to train and RL an op…
Periodic Labs trained an open-weight AI model using 1,300 NVIDIA H200 GPUs and internal lab data, surpassing GPT-6 Astra on their internal benchmarks.
@khoomeik: damn @periodiclabs just leaked that they're building an AI-accelerated semiconductor research lab is this what recursiv…
Periodic Labs has leaked plans to build an AI-accelerated semiconductor research lab, potentially enabling recursive self-improvement in AI hardware.
@pdhsu: Beautiful work - the Weissman lab at MIT strikes again!
The article highlights research from the Weissman lab at MIT, praising their recent contributions.
Nature Is Our Learning Environment (17 minute read)
Periodic Labs has trained the Periodic Neon model on their lab data to outperform frontier models like GPT-6 Astra and Claude Fable 5.1 in XRD analysis for scientific research, achieving a 20x improvement in success rate at lower cost.