OpenAI may be preparing a product for long-running agent tasks, but the specific name, model and launch timing currently rest on one unconfirmed social-media claim. RuntimeWire reports that X user Leo expects an agent called “o” to appear at OpenAI’s DevDay in San Francisco on September 29.
The user describes “o” as a competitor to Grok Bot and Hermes Agent and says it runs on an Astra variant called “aeon.” The post offers no screenshots, product documentation, pricing, access plan or technical demonstration. OpenAI has not announced the name or model.
Confirmed event, unconfirmed product
OpenAI’s DevDay page confirms a September 29 event at Fort Mason, with a keynote, technical sessions, demonstrations and workshops. That schedule makes a product announcement plausible but does not establish what the company will reveal.
The same source claims OpenAI made an acquisition for the product without identifying the deal. OpenAI separately announced an agreement to acquire Ona, a company building cloud infrastructure for long-running developer agents. RuntimeWire cautions that Leo’s post does not connect that transaction to “o.”
Long-running work needs more than a model
A hosted long-running agent would have to preserve state across extended tasks, manage tool permissions and credentials, recover from failures and give users a way to audit what it did. Those operational details would determine whether a product is meaningfully different from an ordinary chatbot loop.
For now, the reliable development is that OpenAI is holding DevDay and has invested in agent infrastructure. The rumored “o” name, “aeon” model and launch remain claims to evaluate against whatever OpenAI actually announces.
