An OpenAI always-on assistant codenamed ‘o’ has surfaced in ChatGPT’s own product listings before any official announcement, offering a rare glimpse at where the company’s consumer AI is heading. References to the unannounced feature appeared briefly on the ChatGPT Pro plan page, where ‘o, your always-on assistant’ was listed as one of the plan’s benefits, before being quietly pulled.
The listing placed ‘o’ alongside other Pro-tier perks: additional Work and Codex usage, maximum memory, 100GB of file storage, and early access to new features. That context matters. Bundling an always-on assistant with memory and generous storage suggests OpenAI is thinking of ‘o’ as something that persists and accumulates context over time, rather than simply being available when you open the app.
There is also a technical breadcrumb in a dynamic configuration file, where a reference to display_name: "o" sits alongside email_suffix: "-o". The email suffix detail is the most intriguing part: it implies ‘o’ could eventually be given some form of email identity, potentially allowing it to receive messages, handle correspondence, or act on your behalf when you are away from the interface. OpenAI has not confirmed or denied any of this, and what shape the feature takes in a finished product remains unknown.
What an OpenAI Always-On Assistant Actually Means
The ‘always-on’ framing is a meaningful shift in how AI assistants are positioned. Most current interactions with ChatGPT are session-based: you open a conversation, get a response, close the tab. An always-on model implies persistent background operation, handling tasks, monitoring inputs, or progressing work without requiring you to be present. Add an email identity to that and you have something closer to an autonomous agent than a chatbot.
OpenAI has been building in this direction for a while. The report also references an internal ‘Aeon’ system, which relates to custom agents for ChatGPT workspaces. OpenAI describes ‘o’ as a separate, consumer-facing experience from Aeon, though the underlying ambition (AI that operates independently on your behalf) is shared across both.
DevDay 2026 and What OpenAI Has Already Revealed
OpenAI has confirmed it is hosting DevDay 2026 on 29 September in San Francisco, at Fort Mason Center, according to GrackerAI. The event is billed as a developer showcase with technical sessions, demos and new tool announcements, making it a plausible venue for a formal ‘o’ reveal, though nothing is confirmed.
DevDay has already produced at least one substantial announcement. According to The Verge, OpenAI announced Dots, an AI agent product powered by GPT-6 Astra, which will initially be available to paid subscribers of ChatGPT Pro, Business Premium, and Enterprise. That launch signals OpenAI’s continued push to give paying customers access to persistent, task-capable AI, the same territory ‘o’ appears to occupy.
Not everything at DevDay went smoothly, however. CNBC reports that OpenAI announced it decided to pull its plans to launch GPT-6.1 Astra because the model did not meet its safety standards. It is a telling admission: OpenAI is willing to shelve a model publicly rather than ship something it cannot stand behind, at least when the spotlight is on. The safety bar, whatever it actually is, appears to be a live constraint on the release schedule.
Whether ‘o’ faces similar scrutiny before it reaches users is unclear. What the ChatGPT Pro listing confirms is that OpenAI has been planning this feature in enough detail to integrate it into public-facing copy, configuration files and pricing structure, suggesting it is considerably further along than a whiteboard idea. The 29 September DevDay date now looks like the most likely moment for OpenAI to put a proper name and a feature set to what its own systems have already started describing.
