The way AI use is being charged for currently I feel is currently broken. - I pay for input tokens. OK, that makes sense. the more tokens I send, the more I should pay since it's more work on the compute side. - I pay for output tokens but 80-95% of those tokens are thinking budget. I don't care about the thinking your model does. I just care about the answer. Can we not be charged just for tokens that are useful to me? But here's the part that really doesn't sit right: the meter is unaudit...