Yeah pretty much everyone is scrambling for some sort of efficiency increase in the hardware or the models. It’s my opinion that the ability of the LLMs is going to hit the top of the S curve soon. Then they will start experimenting with different model strategies etc. that’s when things will really get scary.