Eventually I hope that models themselves become smarter and don't save the whole 54k tokens in their context window
Eventually I hope that models themselves become smarter and don't save the whole 54k tokens in their context window