Nothing is published automatically. Scheduled lists go live on the top of the hour at or after the chosen time.
Each generation costs roughly $0.022 of real model spend, from dedicated Creator accounts that are metered exactly like a customer. No real customer's allowance is touched, and no paying customer can be blocked by this run. Prepare first, then start.
| # | Account | Server state | HTTP | Code | TTFB ms | 1st token ms | Client total ms | Classify ms | Generate ms | Title ms | Retries | KV fail | Spent | Refunded | Net |
|---|
Net should be -100 where a prompt was delivered, which is the generation the account paid for, and exactly 0 on every failed attempt, which means the refund worked. Anything else is usage charged for nothing.