Tibo admits Codex quota bug hit 40,000 accounts — credits going out, fix in place
Tibo's Oct 20 post admits a limits inconsistency hit ~40,000 Codex/Work accounts; OpenAI says it is fixed, is contacting users, and will add credits.
What happened
On October 20, 2026, Thibault Sottiaux (Tibo), OpenAI's Codex and ChatGPT lead, wrote on X that roughly 40,000 accounts had an inconsistency in usage limits: the usage being applied did not match the plan the user had actually selected. The admission came in reply to user reports of quota draining faster than expected over the weekend. OpenAI says it has fixed the problem, is contacting affected users one by one, will add credits to their accounts, and apologized.
What it doesn't say
Tibo's post does not name which plans were affected, does not say whether the error made limits tighter or looser, does not say how long the mismatch lasted, and does not mention GPT-6.1 Sol — so it cannot be read as explaining the separate Reddit reports of heavy quota drain this week.
Why it matters
Usage metering is the trust layer for subscription AI. This is the second October billing-integrity admission from OpenAI's Codex team in a week, after the October 12 round of quota-drain bug fixes — and it landed the same morning OpenAI reset everyone's limits with a one-line post. Verified 2026-10-21.
Source: pasqualepillitteri.it
Sources
Get updates like this every morning
- ② Card on Stripe
- ③ 7 days free
Then $2/month · cancel anytime in one click