Build good habits with AI early on. Locally save useful outputs and start a fresh thread with only the relevant content to keep token cost from unneeded context at minimum.