Tech • AI • Robotics • Game
OpenAI DevDay 2026 Keynote (FULL)
10/10TL;DR
OpenAI unveiled a broad expansion of ChatGPT and its developer platform, centered on always-on AI agents called Dots, a collaborative workspace named Space, new model tiers, faster APIs, and enterprise tools aimed at turning AI into a persistent coworker.
KEY POINTS
Always-on AI agents
Dots were introduced as persistent agents inside ChatGPT that can monitor work, act on instructions, and keep operating without constant prompting. The product is positioned as an AI helper that learns a user’s preferences over time, handles routine work around the clock, and surfaces urgent issues such as overnight updates, scheduling conflicts, bug reports, or project dependencies.
From errands to complex workflows
The new agents are designed to go beyond simple consumer tasks such as booking travel or making reservations. Demonstrations showed a Dot updating a board presentation, reacting to a wedding vendor cancellation, tracking app migrations, monitoring bug reports, testing fixes, drafting pull requests, and helping move software off a legacy API before shutdown.
Cloud computer access and multichannel communication
Each Dot is described as having access to the user’s permissions, connected tools, and an “open computer in the cloud,” allowing it to write code and run tests. OpenAI said users will soon be able to contact their Dot outside ChatGPT through text messages and phone calls, with the agent also able to send updates or ask for decisions.
Space for human and agent collaboration
ChatGPT Space was presented as a collaborative workspace where people and agents can work on the same page. Users can tag agents in comments, ask them to check Slack, track dependencies, update documents, build charts, or maintain live project pages, with the system designed for ongoing teamwork rather than one-off chats.
Availability and enterprise rollout
Dots and Space are available immediately for ChatGPT Pro, Business Premium, and Enterprise users. One Dot is included in the plan, and conversations with it do not count against usage limits. For enterprise customers, OpenAI is also offering specialist Dots that can be configured as shared virtual teammates for broader organizational tasks.
New model and speed options
OpenAI launched ChatGPT 6.1 Sol, describing it as a lower-cost model with intelligence close to Astra, intended for longer tasks and repeated context use. It also introduced UltraFast, a new high-speed mode across the API, ChatGPT, and Codex, with claimed throughput of 300 tokens per second.
Subscription and API updates
A new Pro-500 plan was announced with access to UltraFast and up to 25 times the speed of lower tiers, while Pro-200 was reopened. The company also launched a Decisions API that constrains model outputs to predefined options, enabling subsecond responses for tasks that need fast deterministic choices.
Developer tooling expands into the cloud
Codex is now available in the cloud, letting users start tasks on one device and continue elsewhere. OpenAI also released the underlying Codex harness as open source and introduced a new agent API with components such as multi-agent controls and browser or desktop computer use for testing websites and automating workflows.
Security, privacy and reliability push
For businesses, OpenAI introduced Private Attention, a privacy feature designed to provide frontier-model capabilities without storing customer content on company servers. It said platform reliability remains above 99 percent, while time to first token has been cut by 45 percent.
Research automation and commercial distribution
The company said it has reached a milestone it once predicted: an internal AI “research intern.” It reported that current models can now complete more than one-third of day-long research tasks without intervention, and cited uses in math, antibiotics, ancient language studies, energy, and robotics. On distribution, OpenAI said roughly 1.2 billion people use ChatGPT, and it launched sign-in and plugin extension tools to help developers bring apps, dashboards, editors, and services directly into the ChatGPT ecosystem.
CONCLUSION
The announcements signal a shift from chat-based assistance toward persistent AI coworkers that can execute, collaborate, and operate across business software. The main test now is whether these agents can deliver reliable autonomy at scale while preserving trust, privacy, and user control.
Ask a question