“Ari Morkos argues that frontier model labs have a strong incentive to build proprietary Reinforcement Learning (RL) environments in-house to create a competitive differentiator, rather than buying them from third-party vendors.”