The re-ask loop is the cycle of sending a vague prompt, getting a partial answer, and firing off correction after correction. Each round is a fresh charge of input and output tokens, and the running total adds up fast.

The fix is not the opposite extreme. Cramming everything into one giant mega-prompt with huge pasted context and a request for a very long answer can waste even more tokens than a few short, well-built messages. Both extremes are inefficient.

This course shows how to find that middle: front-load what matters, get a complete answer in one pass when you can, and understand the real cost behind it.