OpenAI just reset usage limits for all paid Codex and Work subscribers after fixing eight background bugs that were silently draining user quotas.
Users had been complaining about burning through their weekly limits surprisingly fast. OpenAI investigated and discovered multiple issues: looping background tasks that ate up to 70% of a weekly quota, subagents launching higher-cost models unprompted, and images miscalculating during dialogue compression. Computer History alone could swallow up to a fifth of a user's limit, while errors in MCP triggered repeated data transfers.
With these fixes rolled out, OpenAI estimates subscribers will get 10% to 50% more actual work out of the exact same limit, depending on their workflow. To make up for the glitchy behavior, OpenAI wiped the slate clean and fully restored used limits for all paid users, with detailed token usage breakdown stats coming soon.
Source
#AITools #GenAI
September 1, 2026 907 1
