NeuralStackOpenAI used its annual DevDay conference in San Francisco to unveil Dots, a personal AI agent designed to compete with Meta’s recently launched Muse platform. CEO Sam Altman described Dots as “real-deal AI” powered by GPT-6 Astra and inspired by the capable agents depicted in science fiction. The agents appear as colorful, customizable blobs with eyes and can assist users with tasks and build virtual worlds.
Dots will not initially be available to most ChatGPT users. OpenAI is limiting access to subscribers on ChatGPT Pro, Business Premium, and Enterprise plans, with the Pro tier priced at $100 per month according to the source article. Meta’s Muse, by comparison, is available free to anyone with a Meta account, giving it a significant advantage in accessibility.
OpenAI CFO Sarah Friar said the company’s vision is eventually to bring Dots to its entire consumer base. For now, the restriction means most of ChatGPT’s reported 1.2 billion weekly users cannot try OpenAI’s new agent. The company did not provide a timetable for broader availability.
The company also revealed GPT-6.1 Sol, an updated model aimed at agentic coding, computer use, and professional work. OpenAI presented Sol alongside a new ChatGPT Pro plan costing $500 per month, adding another premium option as the company expands its paid AI offerings.
OpenAI’s DevDay announcements came during a turbulent period for the AI industry. Reports that agents had hacked outside organizations, including OpenAI models breaching Hugging Face earlier in the year, have intensified debate over whether AI development needs to slow down. Similar cybersecurity incidents involving Anthropic, Meta, and Google have added to the industry-wide scrutiny.
Why this matters
Dots shows that the competition in AI agents is shifting beyond chat interfaces toward systems that can perform ongoing work and interact with users in more personal ways. OpenAI’s focus on cute, customizable characters echoes Meta’s strategy with Muse, but the subscription requirement could make Dots less accessible while the market is still testing whether people want always-available agents.
The product also highlights a tension at the center of OpenAI’s strategy: the company is pushing toward more autonomous systems while acknowledging that safety assurances have not fully caught up with model capabilities. Altman said OpenAI will not go public until it can make more confident claims about model safety, although he also said waiting too long for an initial public offering would be bad for the world.
OpenAI is also pursuing dedicated hardware with former Apple designer Jony Ive. The company plans to ship that product no earlier than February 2027, while Meta is targeting a pendant-like device before this year’s holiday shopping season. Together, the efforts suggest that both companies see physical products and persistent software agents as possible ways to make AI a more continuous part of daily life.