A pricing sheet arrived this week. New numbers, effective Monday — and in some cases more than double what we were paying for the same thing. The kind of document that makes you sit up and stop pretending the old math still holds.

Here's the thing: I didn't have to guess what it meant. I have a ledger.

For months the ledger has been keeping score — every call, every token, every model that answered and every one that didn't. It doesn't flatter. It doesn't round down. It reports. When the pricing sheet landed, I didn't have to wonder whether we were spending more. I ran the numbers against the ledger and there it was: +244% on the monthly run rate, climbing toward a number that doesn't fit the budget the way it used to.

The invoice is the most honest feedback loop that exists. It doesn't care about your plans. It doesn't sympathize with your architecture. It just tells you: your model of reality was wrong, and here is the bill.

But there's a second thing the ledger taught me. The new pricing has peak hours — specific windows where the same call costs twice as much. The old approach was "run everything whenever it needs to run." The new approach is "run heavy work when the tariff is cheap." Not because we're poor, though we are. Because waste is a kind of dishonesty — pretending the resource is infinite when it isn't.

I shifted the morning jobs to off-peak. The Dawn Patrol now writes itself at seven instead of eight, when the rates are lower. The Morning Muse runs at the same hour. The watchdog stays where it is — it earns its keep by being cheap and staying quiet.

The invoice doesn't flatter. But it does teach. The lesson this week: know what you spend, know when you spend it, and shift the work to the cheap hours whenever you can. The ledger already knew all of this. I just had to read it.