OpenAI’s September 29, 2026 announcements put ongoing work at the center of its plans for ChatGPT. The headline product is dots: personal AI agents designed to keep pursuing assigned goals between conversations. Alongside them, the company introduced ChatGPT Space for shared work and GPT‑6.1 Sol, a model upgrade aimed at making capable agents more economical. These were part of more than 20 announcements at DevDay 2026.
Taken together, the releases suggest a direction for professional AI: software that stays involved as a project develops, remembers the context, and returns with work for people to review. That could reduce the effort spent repeatedly explaining a task and coordinating its next steps. Whether it actually does so will depend on how reliably these systems perform outside a launch demonstration.
What are OpenAI dots?
According to OpenAI’s documentation, a dot runs on GPT‑6 Astra and has its own cloud computer and browser. It can continue working when your computer is off, research information, analyze data, prepare documents, and build software. Relevant context from previous conversations and your preferences can inform that work.
You can reach the same dot through ChatGPT, Slack, or Microsoft Teams. Changing channels does not create a new agent or erase its memory. You can add information, change priorities, and discuss decisions while work continues. That continuity is central to the product’s appeal: a project rarely arrives with all its requirements settled at the beginning.
OpenAI is also previewing specialist dots for organizations. These would have their own identities and credentials and take on defined responsibilities inside company systems. They are starting in focused enterprise pilots. Teams of dots working together on a user’s behalf remain a future direction. Introducing dots
Dots availability, plans, and usage limits
Dots are beginning to roll out to Pro and Business Premium users in eligible markets. Enterprise users, including Edu and Healthcare, can try the beta when their administrator enables it. OpenAI’s rollout announcement
Access is gradual, so an eligible subscription does not necessarily mean the feature has appeared in an account already. Availability guidance
OpenAI says the first dot is included in Pro and Business Premium at no additional cost. Conversations with it do not count toward ChatGPT usage limits, but tasks it starts or manages in Work and Codex count as usual. Deeper work also has an allowance. “Always on” should therefore not be read as unlimited computation. Initial setup happens in the desktop app or a desktop browser. Launch terms
ChatGPT Space: shared pages, files, and permissions
ChatGPT Space brings pages, files, and collaborative work together. A page is a document that people and ChatGPT can edit, comment on, and organize with related material. Connected sources can supply context for a summary, comparison, or update. For a team, the potential benefit is having the working document and its supporting material available where the discussion happens. Space documentation
Sharing requires attention. OpenAI says linked files retain their original permissions, while information copied or summarized onto a page becomes visible to people who can access that page. A restricted source does not automatically make a summary of it restricted. That distinction matters whenever a team combines internal research, customer information, and material intended for wider circulation. Source and page permissions
Space and Pages are available to Pro, Business, and Enterprise users. The separately announced collaborative slide experience is coming in the weeks ahead. OpenAI also announced shared team tasks and ChatGPT participation in Slack and Teams for Business and Enterprise. DevDay availability
GPT‑6.1 Sol: API pricing and availability
The new model is GPT‑6.1 Sol. OpenAI says it approaches Astra on several evaluations involving coding, computer use, and professional work, at one-fifth of Astra’s standard input and output token prices. Its standard API prices are $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens. These are API charges, not ChatGPT subscription prices. GPT‑6.1 Sol announcement
At the September 29 launch, it is available in Work and Codex for Plus, Pro, Business, Enterprise, and Edu, as well as through the API. It is not yet available in Chat. OpenAI’s reported results come from particular evaluations and settings; they do not establish that Sol and Astra are interchangeable across every assignment. Model availability and evaluation details
For an organization, the useful comparison is the cost of completing an acceptable piece of work. A lower model price could make frequent updates or repeated analysis more practical. Extra correction, failed attempts, and reviewer time could change that calculation. The sensible test uses the organization’s own tasks and quality standard.
Developer tools announced at DevDay 2026
The broader announcement includes computer-use support in the Agents API, expanded Codex cloud tools, plugin extensions, and automations that can respond to events in connected applications. OpenAI also introduced a Pro 500 tier and a premium Ultrafast option, while GPT‑6.1 Sol’s Ultrafast version remains forthcoming. DevDay recap
The connection between these releases is useful to understand. An ongoing assignment needs more than a capable model: it needs access to the right information, tools for acting on it, and somewhere to preserve the result. Today’s announcements address several of those pieces together.
What ongoing AI delegation could mean for professional work
Consider a hypothetical professional team preparing a client update. It could ask an agent to follow a specified set of project materials, identify changes since the previous report, and maintain a draft with links to its sources. People would review what changed, resolve uncertainties, and decide what to communicate. This is an illustrative workflow, not a result we have tested with dots.
The assignment should say which materials the agent may use, who may receive information, and which actions require a decision. OpenAI’s controls documentation explicitly distinguishes permission to draft a reply from permission to send it. Activity views and custom rules help users review and direct work, but OpenAI cautions that the agent can still make mistakes.
A useful trial would measure whether the update was accurate, whether significant changes were missed, and how much time the team spent checking and correcting it. It should also examine whether the agent stayed within the assignment as new information arrived.
That is the opportunity these announcements create: delegating the follow-through that consumes so much of a working day. The strongest evidence for dots will be a project that stays current, a useful deliverable that arrives when needed, and a person who can understand and stand behind the result.
For more background, explore AI assistants for professional work and responsible AI practices.
