6Latent Space (swyx)·
Lila Sciences bets on scientific data as the last frontier for AI training
Latent Space interviews Andy Beam and Rafa Gómez-Bombarelli of Lila Sciences, a lab building robotic scientific infrastructure to generate novel training data for AI. The core thesis is that scientific experimentation—not internet text—is the next major untapped data source for frontier AI. The piece covers what this looks like operationally, including a room full of robots conducting experiments.
Related guides (3)
Related events (8)
Radical AI's Joseph Krause argues the lab infrastructure is the moat in AI-driven materials science
Latent Space interviews Joseph Krause of Radical AI about their 'self-driving lab' approach to materials discovery, where automated physical experimentation is the core differentiator rather than the underlying AI model. Krause argues that in materials science, the data generation pipeline and lab automation create defensible advantages that model capabilities alone cannot replicate. The piece highlights a deployment pattern where AI is tightly coupled with physical-world feedback loops in scientific research.
AINews: How to Land a Job at a Frontier Lab (on Pretraining)
A Latent Space AINews digest published on a quiet day before Google I/O highlights a notable blog post about landing jobs at frontier AI labs, with a focus on pretraining. The piece appears to surface career and technical insights relevant to the pretraining domain at major AI organizations. The timing suggests it is a low-activity news day filler ahead of a major industry event.
LabVLA: Vision-Language-Action model and RoboGenesis data engine for scientific laboratory robotics
Researchers introduce LabVLA, a Vision-Language-Action model designed to bridge written scientific protocols and physical robot execution in laboratory settings. To address the data scarcity problem, they build RoboGenesis, a simulation-based data engine that composes lab workflows from atomic skills and generates structured demonstrations across robot embodiments. LabVLA uses a two-stage training recipe combining FAST action token pretraining on a Qwen3-VL-4B-Instruct backbone with flow matching posttraining via a DiT action expert. On the LabUtopia benchmark, LabVLA achieves the highest average success rate among evaluated baselines in both in-distribution and out-of-distribution settings.
Doing Vibe Physics — Alex Lupsasca, OpenAI
A Latent Space podcast/essay featuring Alex Lupsasca of OpenAI recounts how GPT-5.x was used to derive new results in theoretical physics and quantum gravity. The piece documents a concrete case of frontier LLMs contributing to original scientific research rather than merely assisting with literature review or code. It represents an early data point on AI-driven discovery in hard sciences.
OpenAI and Los Alamos National Laboratory Announce Research Partnership on Biosafety Evaluations
OpenAI and Los Alamos National Laboratory (LANL) have announced a research partnership focused on developing safety evaluations for frontier AI models. The collaboration specifically targets assessing and measuring biological capabilities and risks. LANL brings national-lab-level biosecurity expertise to the effort, which aligns with OpenAI's broader preparedness framework for catastrophic risk domains.
Anthropic launches Claude Science, an AI workbench for scientific research
Anthropic released Claude Science in beta, an AI-powered research environment integrating over 60 curated scientific tools and databases for genomics, proteomics, structural biology, and cheminformatics. The platform features a coordinating agent with specialist sub-agents, a reviewer agent for citation and calculation checking, reproducible auditable artifacts, and flexible compute management across local machines, HPC clusters, and on-demand GPUs. It integrates with NVIDIA's BioNeMo Agent Toolkit, connecting to models like Evo 2, Boltz-2, and OpenFold3. Available to Claude Pro, Max, Team, and Enterprise users, this represents Anthropic's most significant expansion into scientific AI tooling.
AINews: Open Models, Model Labs vs Agent Labs, and What's Untrainable — Sarah Guo
A Latent Space AINews digest covers open model developments, the emerging distinction between model labs and agent labs, and a featured essay by Sarah Guo on what capabilities remain untrainable. The piece appears to be a reflective commentary day with a focus on strategic framing of the AI ecosystem. The 'model labs vs agent labs' framing and 'what's untrainable' angle suggest substantive industry analysis worth indexing.
Latent Space: Skill engineering and the case against one-shot AI design
Paul Bakaus discusses 'skill engineering' as a design philosophy for AI-assisted workflows, arguing against fully automated one-shot AI pipelines in favor of keeping humans in the loop. The conversation centers on Impeccable, a tool or approach Bakaus is developing, and the concept of 'loopmaxxing' — iterative human-agent collaboration cycles. The piece addresses why current agents still require human steering to produce high-quality outputs.