Running agents all day turns model choice into a budget line. These guides cover routing work to the cheapest model that can do it, prompt caching, context size, and how to find out which task is burning the tokens.
The first guides in this track are being written and checked against the tools' own documentation.