OpenAI Bets on Persistence With Dots, Its Always-On Agent for ChatGPT
OpenAI announced Dots, a new personal agentic assistant for ChatGPT, at its annual DevDay conference in San Francisco, California. CEO Sam Altman introduced the system onstage. Dots is powered by GPT-6 Astra, according to reporting on the launch The Verge.
A Dot is an always-on agent in ChatGPT that takes on ongoing responsibility and keeps making progress between conversations, as described in OpenAI's support documentation OpenAI Help. The pitch is persistence. The agent does not reset at the end of a chat session. It retains task state and continues execution in the background.
Each Dots agent has its own cloud computer and learns from feedback over time OpenAI. In practical terms, that means isolated compute per agent coupled with a feedback loop for behavior tuning. For developers and IT operators, the model will look familiar: long-lived delegation with separate execution context, rather than a stateless prompt-response interaction.
Altman described the new agent as "remarkably capable, always-on" AP. The agents take the form of colorful, personalizable blobs with eyes. OpenAI has restricted Dots to users on its $20/month Pro plan and up. ChatGPT has 1.2 billion users, which gives the company a very large base into which to seed a paid agentic tier The Verge. Altman said Dots are starting out as a premium product because it uses a lot of compute. He also said the agent has been safety tested Reuters.
The enterprise path is explicit. OpenAI is working with Microsoft to integrate Dots into its Agent 365 platform. Specialist Dots trained for accounting and legal analysis are planned to launch soon. That positions Dots both as a consumer assistant and as a domain-tuned worker for back-office workflows where continuity across documents, deadlines and follow-ups matters more than single-turn answers.
DevDay 2026 featured more than 20 announcements, including GPT-6 Astra OpenAI. OpenAI also introduced GPT-6.1 Sol as a major upgrade to GPT-6 Sol. Dots was presented as the user-facing packaging for that model work, in the words of one launch report, a bubbly agentic avatar TechCrunch.
The other system to watch is Meta's Muse. Muse offers a free dedicated virtual machine that houses both the agent and the user's associated data. That is a direct contrast on packaging and price: OpenAI ties persistent compute to a paid ChatGPT tier, while Meta ties it to a free isolated VM.
The broader context here is cost and control around always-on execution. Persistent agents consume inference and infrastructure even when the user is absent. Someone absorbs that cost, and someone defines the trust boundary for data the agent can touch between sessions. OpenAI's answer for now is Pro gating plus per-agent cloud computers. Meta's answer is a free VM under its own stack.
Looking at what this means for builders, the near-term questions are operational. How durable is task memory across weeks of intermittent feedback. How cleanly Dots hand off to enterprise identity, policy and audit controls inside Agent 365. How specialist training for accounting and legal work handles domain accuracy without narrowing the general assistant too far. In this author's view, those integration details will decide adoption more than avatar design.
If that plumbing holds, the upside is straightforward. A billion-scale chat base already knows how to prompt. The next step is learning how to delegate, inspect, and correct agents that keep working after the window is closed. That shift, from conversation to ongoing responsibility, is what Dots is testing.
