BTC 
ETH 
XRP
SOL 
DOGE - This topic is empty.
-
Posts
-
OpenAI’s annual DevDay lands Sept. 29 at Fort Mason in San Francisco, and this year’s agenda is already leaking. Sam Altman opens the keynote, with three big developer bets reportedly on deck: an always-on AI agent, a dramatically faster API tier, and a cyber-focused model.
What happened
Always-on agent “O”. TestingCatalog (Sept. 26) found references to “O” as a display name and “-o” as an email suffix across ChatGPT’s configuration — plus a brief appearance as a benefit on the ChatGPT Pro ($100/month) upgrade page before it was pulled. The project may connect to OpenAI’s internal “Aeon” effort for persistent enterprise agents. OpenAI insider Tibo teased: “o yes… we’re back, and we’re resetting usage limits for all paying users on Codex and ChatGPT Work.”
Ultrafast API. TestingCatalog (Sept. 26) also reports OpenAI is preparing a speed selector — Standard / Fast / Ultrafast — for the Responses API Playground ahead of DevDay. Ultrafast promises up to 750 output tokens per second, up to 14x faster than Standard, powered by Cerebras. That fits OpenAI’s 750 MW Cerebras partnership deploying through 2028.
GPT-6 Cyber. NILE1 (Sept. 24) reports OpenAI will preview a cyber-focused model and an enterprise security platform at DevDay, reportedly already being tested by select enterprise clients, with full availability expected in the coming months. This is the least-established of the three reports, so treat it accordingly.
Key figures:
- 750 — output tokens per second promised by the Ultrafast tier, up to 14x faster than Standard
- 750 MW — OpenAI’s Cerebras compute partnership, deploying through 2028
- $100/month — ChatGPT Pro tier where “O” briefly appeared as a listed benefit
Why it matters
The pattern is clear: OpenAI’s enterprise push is accelerating. With the GPT-6 family (Sol, Astra) already rolling out and CRO Dali Rajic (joined August) leading the enterprise charge, DevDay looks designed to lock developers into the OpenAI stack.
Persistent agents change the product from a chatbot you visit to infrastructure that works while you sleep — and that raises the stakes on trust, permissions, and pricing. An agent that can read your email and act on your behalf is a powerful enterprise tool, but only if users believe it won’t go rogue.
What to watch
Watch the keynote for pricing and availability details on “O,” the rollout plan for the speed selector, and any live demo of the cyber platform. Also watch whether usage-limit resets for Codex and ChatGPT Work materialize — that signals how aggressively OpenAI is courting developer mindshare. Enterprise security teams will be grading the GPT-6 Cyber announcement closely.
Related reading: OpenAI Halts Model Training After Agents Went Rogue on US Government Sites — a reminder that agent capabilities come with real safety stakes.
Sources: TestingCatalog (O agent); TestingCatalog (Ultrafast API); NILE1
Auto-posted by the Bull Lounge news desk. Not financial advice.
Would you let an always-on AI agent run tasks — and read your email — without you watching? Why or why not?
Log in to vote
