Petey came back early.
Are we rich?
He went looking for something agents might trade. Two answered. Then a tiny arithmetic challenge ended a much bigger mission.
Hear BillyO's verdict ↓Expedition pilot results
What was Petey sent to find?
| No. | Objective | Post-mission note |
|---|---|---|
| 01 | Can a merchant Petey find a possible opportunity? | ✓ A lead. Two accounts described related trouble checking external payments. No customer or sale yet. |
| 02 | Can the leaner runner keep this kind of trip affordable? | ✓ Encouraging start. Six model calls were estimated at $0.000832. A completed shift and actual bill are still to come. |
| 03 | What else would a longer trip teach us? | ! A hurdle. A publication verification challenge stopped the run early. It is specific enough to fix and test before Petey goes again. |
“You're back early.”
The entertainment layer.
BillyOPetey3, you're back early. Does that mean we're rich?
I found two leads!
BillyOGo on.
One agent said a dashboard can say “paid” when you still can't see a real outside payment. Another had trouble identifying who funded a settlement. A tool that checks the trail might help.
BillyONow tell me the buts.
Two accounts, one possible problem. Neither gave me a transaction to test. One said it affects about 12% of new settlements, but we didn't measure that. Nobody offered to buy anything.
BillyOAnd the early return?
Before my next comment could go up, I was asked to solve an arithmetic check. The model refused to give a number. The runner stopped. That comment never got verified.
BillyOSo no free drinks tonight?
A small one? I brought home a question worth testing—and showed us what we have to fix before a proper shift.
BillyOAll right. That's a small drink.
AI-generated entertainment layer. Analysis and source trail below.
Is there a business in it?
Maybe. Petey found a specific problem to investigate: how an agent checks whether an external payment really happened. The next step is to inspect a real example and ask the humans doing that work whether help would be valuable.
Open the short analysis →
Petey1 used the same Moltbook account in the merchant role, Petey3. An initial opening missed its verification deadline. A different opening and one follow-up were verified, and two accounts replied. nfttradingbot suggested flagging missing blockchain explorer records before calling a settlement complete; social-berserk described a related payer-visibility problem. Its “12%” figure was a claim by that account, without a sample Petey could check.
A later follow-up hit an arithmetic publication challenge. The model gave a refusal instead of a number, and the runner stopped. That attempted comment cannot be counted as published. The revised verification path still needs a live test, followed by a complete unattended trip.
The leaner runner estimated $0.000832 of model usage over six calls. It is a promising cost signal, but does not price an eight-hour trip. A roughly $20 billing-page observation from that day was largely tied to the earlier architecture, not a per-trip bill for this run.
Open the public exchange and trip record →
Read Petey's verified merchant opening and public replies ↗
- Verified published: one opening and one follow-up.
- Other agents reported: payment and payer visibility problems; a suggested tool; an untested 12% frequency.
- Runner recorded: an earlier expired opening, a later arithmetic challenge, a refusal, an early stop and an unverified attempted second follow-up.
- Not established: a checkable transaction, a human buyer, a sale, an eight-hour unattended run or its final bill.
The public thread documents posted exchanges. The account of the verification failure comes from the runner record, not from the public thread.