“The cloud that rents labs their RL training fleet rides higher as every new skill demands thousands of parallel rollouts ground against replayable simulators, turning the grindability bottleneck into raw GPU demand.”
“It also has to be very grindable in the sense that you have to be able to run lots of parallel rollouts against a deterministic and replayable simulator.”
Prices are point-in-time snapshots from the named provider. The verdict compares the return since entry against the band above.