@HelloSurgeAI
A good model can typically ace an academic benchmark. But even the best model out there will often hit a wall as soon as it’s handed a messy, real-world problem to solve. That’s why we build our own RL environments, to help frontier labs create models that can cope with the sort of rich, weird, unexpected world we all live in.