[fsndz]

true. I think simulations will help a lot in that direction. Imagine if you can do RL a bit like DeepSeek for R1 but on corporate tasks. https://open.substack.com/pub/transitions/p/deepseek-is-comi...