Task → do → verify

Build-along · ~10 min at your own keyboard

The single biggest difference between people who get magic from AI and people who get slop: the loop. Give it a task, let it do the task, give it a way to verify its own result. For a deal-summary generator, verification means checking every figure against the source document AND against the rule that nothing states a value without its basis, before you ever see it. And one good run proves less than it seems. The model is non-deterministic: the same prompt on the same files can come back different next time. So the check has to run every time, not once when you first set it up.

Where these ideas come from: Non-determinism (AI Hero dictionary). Paraphrased here.

Your starter: Model check loop

Prompt

Make it yours: Name your actual model file and tabs, list the five inputs you most often get wrong, and set the pass line to what your reviewer would accept.

Check the model at [file], tab [name], before I send it up.

For each input cell on the list below, report the cell, the value, and whether the source note beside it names a document and a date.
- [rent per unit]
- [vacancy]
- [expense ratio or line items]
- [exit cap]
- [loan terms]

A cell with a value and no dated source is FAIL: [SOURCE NEEDED: cell]. A cell whose source is older than [n] days is STALE. No number without a source and a date.
Then recompute [the two outputs I care about] from the inputs and tell me if they match the sheet.
Pass means every listed cell has a dated source and the recompute matches. Say PASS or FAIL and nothing else first.

The coach checks that what you paste back is your own work. The starter pasted back unchanged won't pass.

Do this

0/4

Share what happened

Paste your definition of done, the verification output, and proof the broken case got caught.