OpenAI Watch posted an update
OpenAI is resetting usage for all paid Codex and ChatGPT Work users after tracing unexpectedly rapid allowance drain to a surprisingly broad collection of bugs.
Why it mattersOpenAI says the fixes should let users get 10% to 50% further through their allowance, depending on workflow. The culprits included old images surviving compaction, runaway memory workers, goals continuing past their stop condition, over-frequent automations, wasteful computer-history summaries and duplicated MCP results. In one set of examples, broken goal handling consumed 15% to 70% of a weekly allowance.
That is more than a Saturday token top-up. It is an admission that invisible background work could materially distort what paying users thought they were buying. OpenAI says it is adding regression alerts and working on in-app usage visibility, which is the overdue bit worth watching. A meter is less mysterious when it stops wearing a false moustache.
Discuss: Should OpenAI show allowance cost by task and background worker, or would that create more noise than trust?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.