OpenAI has transformed its coding tools into something far more autonomous and user-centric. The company’s new product “Dots,” officially unveiled at DevDay 2026, reimagines what personal assistants could be: not just chat-bots, but always-on agents that work in the background—sometimes even when you’re not directly interacting with them.
Initially built from the framework of OpenAI’s earlier tools like Codex and ChatGPT Work, dots live inside the ChatGPT ecosystem and aim to make the transition from manual prompting to proactive assistance. They are designed to anticipate user needs, detect issues like calendar conflicts or software outages, and suggest or take steps to help—all while you focus on more important tasks. The name “dots” comes from OpenAI’s logo negative space, chosen for both its simplicity and aspirational ubiquity.
Designed for users on the Pro tier and above—with pricing starting around $100/month—dots are available on desktop and mobile. Once users connect their dots to apps like email, calendar, and communication tools, the agent begins to observe, analyze, and act. OpenAI emphasizes privacy and control: users can elect to disable model training, determine what resources the dot accesses, and define which tasks it handles autonomously versus those that require approval. However, critics have flagged that some permission controls aren’t yet granular, and that the product currently pushes for connections beyond read-only access in some cases.
The Roots of Dots: From Chatbots to Agents
The lineage of dots traces back to OpenAI’s coding tools, particularly Codex and agentic features inside ChatGPT. Initially, coding agents required explicit instruction loops and hands-on oversight. With dots, OpenAI repackages many of those capabilities into a continuous assistant that learns user preferences over time.
In practice, a dot can monitor feedback channels (e.g., Slack), detect bugs or overdue tasks, scope improvements, generate fixes or features, and bring ready-to-review work. It can also handle non-coding responsibilities: finding unread invoices, preparing proposals, or updating launch materials. Over time, the agent develops its own sense of priorities and how you prefer to work—both in style and substance.
Control, Safety, and Scaling Challenges
Privacy and safety emerge as central concerns in dots’ design. Each dot runs on its own cloud computer; your local machine remains separate unless you explicitly grant access. The agent’s background actions are governed by what’s called “proactive research,” which means the dot can only use read-only tools while operating without direct prompting. Built-in safeguards include auto-review systems, custom rules, and default limitations on tasks deemed sensitive. For example, actions like changing passwords won’t happen without explicit user input.
Dots are rolling out first to eligible users in Pro and Business Premium plans—excluding markets like the UK, Switzerland, and the European Economic Area for now. Enterprise, Healthcare, and Education users can access dots via beta, typically requiring admin enablement. While the base agent comes without additional cost for qualifying plans, deeper workloads consume usage limits when interacting via Codex or similar environments.
The concept of always-on agents isn’t new, and OpenAI’s dots arrive as part of a broader surge in the use of persistent AI helpers. Meta’s Muse, Anthropic’s Claude, and others are also delivering similar always-aware tools. The critical question is whether these assistants can scale: serving many users without crippling infrastructure, maintaining clear boundaries between useful aid and intrusive surveillance, and integrating with payment systems, email, calendars, and more—all without compromising safety or privacy.
OpenAI argues dots are the natural evolution of our interaction with AI: from reactive tools to collaborators that carry forward work even when we’re not present. In demonstrations, dots have flagged calendar overlaps too late, built proof-of-concepts for enterprise deals, and even noticed outages before their users did. While functionality is still maturing, the premise is compelling: less time prompting, more time acting.
Dots mark a meaningful step shift for OpenAI—not just upgrading models, but rethinking how humans and AI partner day to day. For early adopters, the promise is freeing up mental overhead; for critics, it raises critical questions about data access, autonomy, and oversight. What to watch next: how well dots balance proactivity with precision, how inclusive access becomes globally, and whether OpenAI’s control mechanisms hold up as dots take on more responsibility.
