Glad you liked it. Regarding power consumption I'm confident those issues will be solved once the incentives to do so are in place. Right now it's a gold rush, but eventually our attention will turn to lowering the cost of using these models. That's already started happening as evidenced by DeepSeek's extremely low cost of inference.