Recent Releases
View moreSimulation Complexity Leads Capabilities
Each jump in model capability has followed a jump in the richness of training environments: chess, then physics engines, then software and whole computers.
As scaling laws continue to hold, what limits the next generation of capabilities is how far we can scale realistic simulation complexity. Our work focuses on the technical barriers to pushing those limits.
- 1Infra
- 2Data
- 3Environments
- 4Rollouts
- 5Grading
Infrastructure for Scaling Simulations
We believe great simulation research is an infrastructure problem. Testing every variation, analyzing results, and iterating are all bounded by infrastructure. Our approach to advancing the frontier lies in the technical systems needed to scale domains like multi-agent simulations and robotics.
Get in touch
Frontier Labs and Researchers
Talk to our team about environments for your RL, research, and evaluations.
Contact usJoin Us
We are hiring across distributed systems, evaluation research, and post-training.
How to apply