Particle.news

Leaked Traces Suggest OpenAI Is Testing 'o,' an Always‑On ChatGPT Assistant

If launched, the cloud agent would run persistently to manage email and multi-step tasks and would change ChatGPT from a turn‑based chat tool into an autonomous workflow assistant.

Overview

  • Multiple outlets this week reported leaked UI entries, localization files, and internal configuration that name a new assistant called "o" and label it as an always-on feature.
  • Code trackers found a display_name set to "o" and an email_suffix of "-o," which point to a built-in identity or email workflow capability rather than a simple chat feature.
  • Leaked descriptions portray 'o' running in a persistent cloud sandbox for hours or days to organize schedules, fix errors, and manage email and other multi-step workflows without repeated user prompts.
  • Internal references and feature flags link the project to OpenAI systems called Aeon and Astra and to a multi-agent controller design, suggesting it builds on the company’s agent and workspace infrastructure.
  • The traces briefly appeared on ChatGPT upgrade pages next to a $100 Pro tier and some leaked pricing places 'o' on higher Pro plans; OpenAI has not confirmed the product and more details may be revealed at DevDay on Tuesday, Sept. 29.