OpenAI DevDay 2026: Always-On Personal Agent 'Dots' Launch

OpenAI launched Dots at DevDay 2026 on September 29, an always-on personal agent powered by GPT-6 Astra that proactively manages work and personal tasks. It also introduced the cheaper GPT-6.1 Sol model, the UltraFast speed tier, and ChatGPT plugins, with Dots open to Pro, Business Premium, and Enterprise users.
Key insights
Dots Positioned as Proactive Always-On Personal Agent
Sam Altman framed Dots as an always-on agent with its own cloud computer and browser that can write code and run tests, proactively pushing tasks forward and seeking user approval only at key decision points. Dots inherits user permissions across 4,000-plus connected apps, can later be reached by text message and phone call, and runs on GPT-6 Astra, which Altman called OpenAI's best-aligned model. Users can restrict which apps a Dot uses and write custom instructions for its operations. Altman said his own Dot filters overnight messages each morning and returned some of his attention, and he cited a Dot migrating an application off a deprecated API end to end, from dependency tracking through tests to a pull request for review.
ChatGPT Space Adds Shared Workspaces for Human-Agent Collaboration
ChatGPT Space is a page-based workspace where users write plans, do research, and generate images and data, with pages and files stored in one space like a cloud drive. Dots work directly on pages, take tasks via @-mentions in comments, and pages can carry standing instructions such as monitoring a Slack channel daily and posting findings. In the demo, slash commands inserted interactive charts, tables, and runnable prototypes; a Dot refreshed a feedback-filtered chart hourly and located onboarding data referenced only vaguely from Slack DMs to update an FAQ. Presentations in agent-readable formats and team editing alongside Dots are planned.
Dots Already Handle Daily Engineering Work Inside OpenAI
Holly described Dots operating inside company Slack with their own identities, where employees @-mention their Dots for small requests and add Dots to group chats from the start so everyone sees results. In a staged pre-launch scenario, a Dot caught a moved review meeting, read overnight tester feedback, rebuilt a redesigned homepage via Codex running on an iPhone simulator, and submitted a pull request. Engineers' Dots reportedly pick up reported bugs, investigate, and submit fix PRs, fixing dozens of bugs daily, and large parts of the Dots product itself are written by Dots.
GPT-6.1 Sol and UltraFast Target Cheaper, Faster Agent Workloads
GPT-6.1 Sol delivers capability close to Astra at one-fifth the price, with cached input 95% cheaper than standard input, positioned as the daily workhorse for developers since agent loops repeatedly re-read context. UltraFast is a new speed tier across API, ChatGPT, and Codex offering 8x speed at 6x price with 300 tokens per second, supporting Astra now and Sol later. A new $500 Pro 500 subscription carries the highest quota at 25x Plus and includes UltraFast in ChatGPT and Codex; Pro 200 reopens. A Decisions API preview constrains Luna models to predefined choices for sub-one-second responses, drawing robotics interest for near-real-time vision-driven action.
Developer Stack Gains Agents API, Security Cloud, and Open Harness
OpenAI open-sourced the Codex harness that also