Users Report Widespread ChatGPT Degradation After DevDay: Slow Responses, Multi-Platform Bugs, Wasted Paid Quotas
Within 24-48 hours after DevDay, multiple Reddit users reported a sharp, broad degradation of ChatGPT spanning response speed, instruction following, multi-device sync, and tool integrations, with some users losing large amounts of paid quota. OpenAI has not publicly responded.
Confirmed
- Per @B3ttor, one user reported ChatGPT became noticeably slower and dumber in the past 24 hours: tasks that once took 20 seconds to 2 minutes now take 5-20 minutes, the model fails to follow instructions, forgets context, repeats mistakes, and can no longer execute bootstrap base instructions that had worked for two months.
- Per @Airaneon, one user spent over 14 hours on a single task where the model "refused to understand" basic instructions, wasting substantial Codex usage; new sessions, externalized context, and past successful task examples all failed to help.
- Per @affluendo, a ChatGPT Work user hit "Error in Message Stream" after each prompt diligently ran for about an hour, with all subsequent prompts failing; one incident ran 82 minutes and wasted 20% of the weekly quota, leaving Work nearly unusable.
- Per @Kayters, a heavy user reported macOS flooding "Turn complete" notifications after DevDay, broken sync across iOS/macOS/web (deleted conversations persisting, recent chats not updating), and "Error loading conversation" errors.
- Per @the-human-user, a user maintaining a repo with the ChatGPT agent for weeks found it looping—apologizing and explaining instead of editing a markdown file to remove outdated "draft" wording—despite verified GitHub tool permissions; the user, on GPT-6 Astra Pro, suspects another "dumbing-down" update.
- Per @NoCherry606, a user said ChatGPT Work was in terrible shape after DevDay: broken Notion integration, denied MCP access, PDF reading failures, and doubled task times; switching from Sol 6.1 to Astra 6 briefly helped before Astra 6 also failed.
Unconfirmed
- All reports are one-sided user accounts; the scope, whether the issues stem from DevDay changes, and whether this is a "dumbing-down" update lack official confirmation.
Why it matters
- The reports cluster within 24-48 hours after DevDay, span free and paid (Pro, Work) users across platforms and model versions, and some failures directly waste paid quota. If confirmed, they may point to a systemic issue from recent server-side or model configuration changes—a concern for anyone relying on ChatGPT for production work.
2026-10-02 ~ 2026-10-03 · 6 related posts
- Episode 1: Paying ChatGPT users report old Projects chats failing to load(2026-10-01, 2 posts)
- Episode 2: Users Report Widespread ChatGPT Degradation After DevDay: Slow Responses, Multi-Platform Bugs, Wasted Paid Quotas(2026-10-02, 6 posts)
Primary sources
- GPT-6 Astra Pro user hits agent stuck in edit loop, suspects another OpenAI 'lobotomy' update — the-human-user · 2026-10-02
- [source] User burns 14+ hours of Codex usage after GPT suddenly stops following basic instructions — Airaneon · 2026-10-02
- ChatGPT reliability falls off a cliff after DevDay: notification spam, sync bugs, dumber models — Kayters · 2026-10-02
- ChatGPT Work riddled with bugs since DevDay: broken PDFs, MCP denials, 10x slowdowns — NoCherry606 · 2026-10-02
- [source] ChatGPT works for 82 minutes then errors, wasting 20% of user's weekly usage — affluendo · 2026-10-02
- [source] Users report ChatGPT suddenly slow and lobotomized, responses taking 5-20 minutes — B3ttor · 2026-10-03