Help

How to actually use the lab.

1. Press Next step

The fixed request is right there at the top of the transcript, in its own "Your prompt" card — the exact text about to be handed to the model. It's read-only; press Next step on that card to kick things off. Everything downstream of it is scripted rather than a live model call (see About for why), so it only knows how to answer this one request; that's the trade-off for it running for free.

2. Keep pressing Next step

It stops after every step on purpose — each card gets its own Next step button once it finishes, and nothing happens until you click it. There's no approve or deny to think about; it's just pacing, so you can read what one step did before the next one starts. "What it's doing" fills in as you go: the model's own text, each tool it calls with the arguments it chose, and the result that came back. A red entry means the tool call failed — the agent has to notice that and recover, same as it would with a real error.

3. Check its work

Every write_slot call changes a real value in the table under "The slots," not just in the transcript. The changed row flashes briefly so you can spot what just moved — and because it's just numbers, you can verify the agent got it right without knowing anything else about what it's doing.

Start over

Start over puts all four slots back to their starting values, so you can run it again from a clean slate. It's instant — nothing is actually persisted, so there's nothing to undo.

← Back to the lab