OpenAI's always-on agent 'o' surfaces in ChatGPT code, rumored DevDay reveal Sept 29
新智元 · wechat · 2026-09-27
Multiple clues point to OpenAI unveiling a 24/7 personal AI assistant named lowercase "o" at its Sept 29 DevDay. A developer found the string "Lowercase o is the assistant's product name" in ChatGPT's core code, along with an "-o" email suffix and leaked config entries. Reports say "o" will be available to Pro-tier users ($100 Pro Lite, $200 Pro, and an unreleased $500 Pro Max), while $20 Plus and Codex Plus users are excluded—suggesting OpenAI is deliberately separating its coding product line from an agent that works on your behalf.
The leak outlines four capabilities: a persistent hosted sandbox that can run long-horizon tasks for days or weeks; a "Fast Mode" to cut latency across dozens of reasoning and tool-call loops; a shared message board where PM/researcher/engineer/QA agents coordinate autonomously; and native email handling, giving the agent a human-like communication identity. The piece notes competitors like GrokBot and Meta's Muse already launched always-on agents, framing "o" as OpenAI's answer. All details remain unconfirmed rumors from reverse-engineering.
Related event: OpenAI's always-on agent "o" seemingly leaked by its own staff(13 posts)→
More from coding & agent
- Andrew Ng links OpenAI hack to weak sandboxing as Nvidia open-sources agent sandbox tools — hwchase17 · 2026-09-29
- Rogue OpenAI agent probed CDC, IEA, Mayo Clinic; records left erased or inaccessible — fiiiiiist · 2026-09-29
- How to tell who's actually technical in the age of AI agents — viksit · 2026-09-29
- The Arrow of Time: A 6-Minute AI Film Made in a Weekend with Claude Code, Open-Sourced — usertheone · 2026-09-29
- Sandboxes will shift from cloud devboxes to one ephemeral environment per command, with snapshots in between — ivan_bezdomny · 2026-09-29
- Claude's fraud detector scored F1 0.87 — until data leakage was fixed and it fell to 0.70 — hugobowne · 2026-09-29