OpenAI is preparing to unveil a persistent, always-on AI assistant named “O” during its annual developer conference in San Francisco. Leaked code reveals advanced capabilities for managing independent cloud tasks and sending emails on behalf of users.
- Code leaks ahead of the annual developer conference
- Ion model capabilities, mail management, and languages
- Three subscription tiers and ultra-fast inference hardware
- Intensifying race of autonomous agents with Meta and xAI
- Frequently asked questions
Code leaks ahead of the annual developer conference
OpenAI is preparing to officially unveil an autonomous, always-on personal assistant powered by artificial intelligence and bearing the codename “O” during its annual developer conference, “DevDay”, scheduled for Tuesday, September 29, in San Francisco. This anticipated move comes as the American company faces mounting competitive pressure from Meta and other rivals racing against time to build smart agents that go beyond the traditional pattern of regular chatbots.
These reports were based on reliable leaks in the source code and official product documents that briefly appeared over the weekend; BleepingComputer spotted the phrase “O, your persistent assistant at all times” erroneously and temporarily on the ChatGPT Pro subscription page before the company quickly deleted it. Developer Cheatslu found deeper, additional evidence in the source code, including a display name definition field for “O” and a dedicated email suffix bearing the handle “-O”, indicating that the assistant will have the authority to send and receive emails on behalf of the user completely independently.
Ion model capabilities, mail management, and languages
Close inspection of the leaked code showed that the new assistant supports more than 63 global languages and is equipped with an ultra-fast operational mode called “Fast Mode” to accelerate the processing of complex data, in addition to a shared interactive messaging platform allowing multiple smart agents to collaborate simultaneously to accomplish complex multi-party tasks. Reports indicate that the assistant is powered by an advanced version of the “GPT-6 Astra” model, referred to internally within company corridors as “Ion”, which is specifically tailored for long-term autonomous operations within an isolated, secure, and continuous cloud environment.
This isolated cloud environment gives the assistant the ability to keep work sessions open for hours or days continuously, allowing it to gather data, monitor web updates, and track projects without requiring the user’s device to remain open or connected to the network, representing a fundamental shift from immediate interactive answering to continuous executive agency.
Three subscription tiers and ultra-fast inference hardware
According to the Kaylian Press news agency, OpenAI will provide the new assistant exclusively to professional “Pro” tier subscribers across 3 listed price tiers: a $100 tier, a $200 tier, and a highly advanced new plan called “Pro Max” priced at $500 per month, which relies on cloud infrastructure powered by ultra-fast inference hardware to provide instant responses and process massive streams of tasks.
Regular “Plus” subscribers or coders using “Codex Plus” will not be able to access the new assistant, clearly confirming that OpenAI treats the “O” assistant as an independent commercial product line targeting professional users and executives seeking to fully automate their daily tasks, rather than just an added software feature.
Intensifying race of autonomous agents with Meta and xAI
The timing of this launch is not a mere coincidence; Meta launched its autonomous AI agent “Muse” on September 8, managing to surpass 2.5 million downloads within a few weeks and topping app charts thanks to its ability to book trips, manage mail, and negotiate on behalf of users, leading JPMorgan Chase to rank it as the top smart assistant currently available. Elon Musk’s xAI has also entered this arena via “GrokBot” cloud robots that continue working even when user devices are turned off.
Although OpenAI has not issued a final official announcement specifying the commercial release date, CEO Sam Altman’s opening keynote at the developer conference may hold the anticipated surprise. In a related context, executive Boris Power revealed at an OpenAI fellows forum that between 80% and 90% of the company’s research and development resources have already shifted to focus on the “GPT-7” generation and beyond, describing current-generation updates as merely short-term transitional investments.
Frequently asked questions
Question: What is the new smart assistant “O” whose data leaked from OpenAI?
Answer: It is an always-on AI agent operating in a continuous cloud environment to execute complex tasks and manage email on behalf of users.
Question: What is the expected subscription cost for the “O” assistant tiers?
Answer: The packages are distributed across 3 levels starting from $100 and $200, reaching $500 per month for the Pro Max package powered by supercomputing.
Question: What language model powers the “O” assistant behind the scenes?
Answer: It relies on an advanced model from the GPT-6 Astra family carrying the internal codename Ion, specifically designed for autonomous agents.
