A Dot loop burned his weekly ChatGPT Work limit
ST-Automation thought Dot usage would stay off their shared quota. Instead the Dot started a loop inside ChatGPT Work rather than on its own machine, and he says it burned the whole weekly limit plus about 40,000 Codex credits, over €1,000 by his account. He emailed support with screenshots and asked for a review. Always-on agents that can open Work or coding tools need hard caps, and you should watch the first overnight runs. If usage spikes, pause the Dot before you wait for a reply from support.
Poster's claim, not verified: He says the loop cost his whole weekly limit plus about 40,000 credits, over €1,000. That's his account, not a typical bill.
The lesson: An always-on Dot that loops inside ChatGPT Work can burn a shared weekly limit and coding credits before you notice.
Copy this prompt
Before you leave a Dot running overnight, list every tool it can open, including ChatGPT Work and any coding credits. Set a hard stop: if it uses more than [LIMIT] in [HOURS], pause and message me. Do not start open-ended loops. After any long run, show me a usage summary with what it did and what it spent.
Our own prompt, written from what the source describes. We haven't run it ourselves. Text in [BRACKETS] is what you replace.
Watch out for
- Shared pools mean one looping Dot can empty Work and coding credits for everyone on the plan.
- Watch the first overnight run. Do not assume Dot usage sits outside your weekly limit.
- If usage spikes, pause the Dot first, then contact support with screenshots.
Where this comes from
Original post by @ST_Automation on X, Nov 5, 2026. We rewrote it in our own words. Check the original for the full context.
Read the originalGuides for this job
Turn a messy inbox into a 5-minute daily summary
Paste in today's emails and get back what needs a reply today, what can wait, what to ignore, and a one-line reply for each. It also flags emails that look like scams.