Agents that ship: cost, caching, and picking the right model

Frontier models are expensive and often unnecessary. We're working out what to run where, what caching actually saves, and how to stop a runaway agent burning your budget.

Schuler Books on 28th Street, in the Chapbook Cafe. Coffee, real tables, no cover.

WHAT WE'RE DIGGING INTO
When a small fast model is genuinely enough and when it isn't. Prompt caching and what it really cuts. Routing easy work to cheap models and hard work to expensive ones. Why an agent costs multiples of a chat turn, and what that means for anything you run on a schedule. Setting a ceiling before you get the bill.

WHO THIS IS FOR
Anyone paying for API calls, or about to. Anyone who's watched a bill jump and not known which part did it. Beginners welcome — this is the practical end of the subject.

HOW THE EVENING GOES
6:30 — coffee, introductions, what everyone's running
6:50 — real numbers from Digi-Bot: what each part costs me
7:20 — open floor: bring your setup and we'll find the waste
8:30 — whoever's still talking keeps talking

Bring a laptop if you want to dig into your own usage. Schuler Books, 2660 28th St SE — big parking lot.