At Proximal, we believe that building our own research infrastructure is crucial to design the highest-quality training data Today, we are sharing more information about our internal post-training stack

If you are interested in: - Post-Training - Data Research - Infra at tremendous scale or just generally want to know more about what we do @ProximalHQ, please reach out, I would love to chat :)

The models that will accelerate drug discovery, chip design, and energy are limited by the data they learn from. We led @ProximalHQ’s $15M seed as they build the data engine for the next generation of frontier AI and treat training data with the same rigor as model architecture. Watch @quentinclark in conversation with @calvinchen. Chapters 00:00 — Meet Proximal 03:45 — A Rainy Afternoon in South Park 05:19 — From Sneaker Bots to a First Exit 08:14 — Why Suffering Equals Growth 10:13 — The Path to Proximal 11:57 — How Models Learn 15:17 — Training Data as a Research Problem 23:24 — Building the Team in India

Proximal builds infrastructure that enables AI to improve from real-world experience. In the past year, we have grown to more than $200M in annualized revenue helping frontier labs and enterprises improve coding agents. Now, we are expanding beyond software engineering. Recent progress in mathematics demonstrates what happens when models are trained in domains with massive amounts of public data and easy verifiability, making it easy to find weaknesses, generate training tasks, and iterate quickly. We build infrastructure that enables this feedback loop in other domains: we are excited about a world in which AI systems design targeted drugs, accelerate chip design, and rewrite legacy software that still runs hospitals, power grids, governments, and other critical infrastructure. Since starting, we have assembled a small team of researchers and engineers and built world-class infrastructure for post-training and synthetic data research to overcome the challenges of scaling manual data curation. Our team comes from Cursor, Google DeepMind, Meta Superintelligence, Prime Intellect, Citadel, and Jane Street. More than half are former founders. We’re fortunate to be backed by investors who share that vision, including @generalcatalyst who led our $15M Seed Round at a $300M valuation, as well as @chemistry, @svangel, @dvlamoen and individuals like @LiamFedus, @kevinweil, and @bernhardsson

We partnered with @FireworksAI_HQ to bring FrontierSWE v2 to the Specialized Intelligence Index Evals are essential for improving and safely deploying AI. We're excited to support the initiative!

During development, Astra overlays grids on frames to inspect and correct road labels. It also writes a procedural road generator to train its controller across viewing angles and adds recovery when the car gets stuck against walls

We also observe a large amount of cheating and sandbox escape attempts, and notably, attempts to consciously conceal cheating rather than "cheating by accident" We are still investigating these incidents and will share an more detailed analysis in a follow-up blog post

We are releasing FrontierSWE v2, our updated ultra-long horizon coding benchmark V2 features an expanded task suite and improved methodology. We see large performance gaps between frontier models, with Claude Fable 5.1 leading by a wide margin