What does Proximal AI do?
Proximal is a San Francisco research lab and data company building high-fidelity, long-horizon RL environments and training data for frontier coding agents. It argues the bottleneck for frontier-model data is research systems rather than domain experts, and it publishes FrontierSWE, an open-source ultra-long-horizon software-engineering benchmark that current frontier models largely fail. The team was founded by Justus Mattern, Calvin Chen, and Navid Pour.