I recently saw a production AI pipeline that ingests multiple long-form conversations per user, runs analysis across dozens of dimensions, and generates fully personalized output. All on the most expensive models money can buy. The cost per user? Less than a quarter. Most of you are spending more than that asking Claude what to have for dinner.
That number should bother you — not because frontier AI is expensive, but because it isn’t. Frontier AI running on the most expensive models available costs pennies when you know what you’re doing. The problem is that most people don’t know what they’re doing. They’re burning 5x, 10x, sometimes 20x what they should on the exact same work, and they think that’s just what AI costs.
The models aren’t expensive. Your habits are expensive.
And with Claude usage limits dominating every AI community on the internet, those habits are finally catching up with people. So I want to be direct about what I’m actually seeing.
Here’s what’s inside:
The ChatGPT migration problem. Why habits from ChatGPT are catastrophically expensive on Claude, and the single fix that changes everything.
What I’m actually seeing. The four levels of token waste, from rookie to “advanced,” with real numbers.
The math that should scare you. Clean vs. sloppy sessions, Mythos pricing implications, and why the gap keeps widening.
A diagnostic. Six questions to find out if you’re the problem.
What we’re building. The Stupid Button, KISS Commandments, and the Heavy File Ingestion skill already in the OB1 repo.
The usage limit crisis started getting loud about a week ago. Some of it is real infrastructure strain — but some of it is fixable.
Listen to this episode with a 7-day free trial
Subscribe to Nate’s Substack to listen to this post and get 7 days of free access to the full post archives.
