← LibraryBuilder · B3
Cost caps
Build-along · ~8 min at your own keyboard
Long autonomous runs spend real money. A tidy-and-fit run across 200 exports, or a figure engine redrawing a month of panels, can add up fast. Before any big run: check /cost, estimate the ceiling, and write a stop-rule into the task itself ('if this exceeds ~X effort, stop and check in').
Your starter: Run budget card
TemplateMake it yours: Fill each bracket from your own session: the real /cost reading, the run you are about to start, and your own ceiling. Leave the price line as [NEEDS: ...] until you have read the provider's current page.
RUN BUDGET CARD Run: [what the run does, e.g. tidy and fit 200 plate-reader exports] Today so far (/cost): [paste the reading; do not estimate it] Size of this run: [count of exports, figures or files] Stop-rule, pasted into the prompt: "Before you start, count the [exports / files]. If the count is over [N], stop and show me the count. Do not continue until I say go." Daily ceiling: [amount] per day. Above this I decide it; I do not discover it. Price per unit of use: [NEEDS: copy from the provider's current pricing page, with the date you read it] After the run Did the stop-rule fire? [yes / no / not reached] /cost after: [paste] CANNOT ASSESS from this card: whether the run's output is correct. Cost is one check; the output still goes through the verify loop.
The coach checks that what you paste back is your own work. The starter pasted back unchanged won't pass.
Do this
0/4Share what happened
Paste your /cost reading, your in-prompt budget line, and your personal daily ceiling.