We build
RL environments.
Production systems. Autonomous improvement loops.
Research
Environments for autonomous improvement of coding agents in complex enterprise systems.
When More Compute Stops Helping
Why inference-time scaling saturates on coding tasks governed by implicit semantic contracts.
COBOLBench: Evaluating Coding Agents on Enterprise COBOL
Public benchmark measuring frontier coding agents on 53 long-horizon enterprise COBOL maintenance tasks.
VideoBench: Evaluating Coding Agents on Source-Grounded Video Generation
A benchmark measuring AI agents on video, animation, and presentation generation from curated data rooms. 183 chapters, 41 courses.
FigmaBench: Throwing Frontier Models into Real-World Figma
A benchmark measuring AI agents on production Figma-to-code conversion through API interaction, design hierarchy extraction, and iterative deployment.
Metaphi Simhub
Interactive simulation environments that move agent evaluation beyond static, non-interactive benchmarks.
From Static Benchmarks to Dynamic Worlds
How principles from autonomous vehicle simulation can transform code generation agent training and evaluation.
Get in touch
Tell us about your use case.