OpenAI DevDay 2026: The Year AI Stopped Chatting and Started Working
At DevDay 2026, OpenAI made its most consequential product pivot since the launch of ChatGPT. The company is no longer just selling an AI you talk to. It is selling an AI that works for you—continuously, autonomously, and with its own computer.
The centerpiece announcement was dots, a new class of "always-on" AI agents that operate 24/7 on your behalf. Alongside it, OpenAI introduced ChatGPT Space, a shared workspace where humans and AI agents collaborate on documents, and a suite of developer tools that open the door to a new kind of AI-powered workforce.
The mood at the event was clear: OpenAI wants to cement the idea that AI is moving from a chatbot you consult to a worker you assign things to—one that keeps going while you sleep [citation:11].
What Are Dots?
Dots are OpenAI's answer to the persistent personal AI assistant. Unlike ChatGPT, which requires a prompt to produce an output, dots are designed to work continuously. You give a dot a goal, define what it is allowed to do autonomously, and it takes actions on your behalf across your connected apps [citation:19].
Each dot runs on GPT-6 Astra, OpenAI's most capable and aligned model, and gets its own dedicated cloud computer and web browser [citation:3][citation:11]. This is a critical architectural decision: the dot does not run on your machine. It operates in its own isolated environment, separated from your personal device unless you explicitly connect the two [citation:3].
Dots can plug into over 4,000 apps, use a browser to access the internet, and even use your saved credentials to log into services—without the underlying passwords ever being exposed to the model [citation:19]. You can interact with your dot through ChatGPT, via a voice call, or through Slack and Teams. Texting support is coming soon [citation:11].
OpenAI CEO Sam Altman described dots as an AI that will "always have your back," helping human employees focus on higher-level work. "It's the way I always wanted to work with AI," he said. "In the future, you can work with a whole team of dots" [citation:19].
Governance Built In, Not Bolted On
The challenge with autonomous agents is trust. A mistake made by a chatbot produces an inaccurate paragraph. A mistake made by an agent with access to your business software can alter records, send information, or trigger downstream workflows [citation:3].
OpenAI has built governance into dots from the ground up. Each dot operates under built-in rules that determine when it can act alone and when it must ask for approval. Some sensitive operations—like changing passwords—remain reserved for the human user [citation:3]. An Activity View lets users inspect background work and intervene if something is going wrong. An auto-review system checks potentially consequential actions against the user's instructions, Custom Rules, and OpenAI's safety requirements before determining whether work can proceed autonomously [citation:3].
For enterprise customers, OpenAI is piloting specialist dots—agents that receive their own organizational identity, credentials, and access to company systems, and are assigned a defined business responsibility. The company has experimented internally with agents working in procurement, invoice processing, email marketing, customer support, and commercial contracting. It is now working with Microsoft to integrate these specialist agents into Microsoft Agent 365, allowing businesses to govern them through existing security and management systems [citation:3].
This hints at a future where enterprises treat agents less like software licenses and more like new machine identities that require onboarding, credentials, access policies, monitoring, and offboarding [citation:3].
ChatGPT Space: Where Humans and Agents Share the Same Page
Dots do not work in isolation. OpenAI also announced ChatGPT Space, a shared workspace where teammates, ChatGPT, and each person's dots collaborate from the same pool of project knowledge [citation:11].
Space introduces Pages, a new document type built for people and agents to edit together. In a demo, an OpenAI employee tagged coworkers and her dot inside a Space, assigning tasks: write an FAQ, update the team, and comb through Slack DMs. The dot and ChatGPT worked alongside the humans, editing the same document in real time [citation:11].
The governance model for Space is nuanced. Workspace administrators' existing sharing rules apply, and teams can grant or revoke access to Pages. Files uploaded directly into a Page inherit that Page's permissions. OpenAI draws a boundary between a user's private ChatGPT context and shared Space content: sharing a Page does not expose private conversations or personal memory to coworkers. However, if ChatGPT uses something from a person's memory and writes it onto the Page, that information becomes visible to collaborators—making the Page an important checkpoint for employees and administrators [citation:3].
The Bigger Picture: A New Interface Battle
OpenAI is not entering an empty market. Over the past several weeks, a cluster of companies has converged on a remarkably similar idea: an AI assistant with a persistent identity, its own computing environment, access to a user's accounts, memory of what matters to that person, and enough autonomy to act without continuous prompting [citation:3].
Meta has been developing its own persistent agent, Muse. Other startups are pursuing the same vision. The competitive question is shifting from whose model gives the best answer to which agent a person or business is willing to trust with meaningful access [citation:3].
OpenAI's bet is that governance, safety, and enterprise readiness will be the deciding factors. By building dots with restricted environments, approval workflows, and organizational controls, it is positioning itself as the responsible choice for businesses that want autonomous capability without uncontrolled risk.
Whether that bet pays off will depend on how well the system works in practice—and whether enterprises are ready to hand over the keys to software that keeps working even when the laptop is closed.