OpenAI DevDay 2026: Sam Altman Unveils 'Dots' Autonomous Persistent Agent Workspaces, GPT-6.1 Sol, and Cloud-Hosted Computer Use

SAN FRANCISCO, CA — September 30, 2026 — At a packed Fort Mason Center for OpenAI's annual DevDay developer conference, CEO Sam Altman took the stage to announce what industry observers describe as the most consequential paradigm shift in computing since the graphical user interface: the formal death of the stateless conversational chatbot and the dawn of persistent, autonomous agentic workcells named "Dots".

Rather than waiting idle for user prompts in a browser tab, a Dot is an always-on, persistent software entity that lives in ChatGPT and enterprise workspaces. Backed by dedicated cloud virtual machines, persistent Linux filesystems, and headless browser sessions, Dots run continuous multi-step tasks across productivity suites, enterprise repositories, and third-party SaaS platforms around the clock.

1. Anatomy of a Dot: Dedicated Cloud Computers and Computer Use

Each Dot is assigned its own secure micro-containerized cloud workspace. When a user requests an outcome—such as auditing an enterprise procurement pipeline, drafting compliant RFPs, or resolving a GitHub issue backlog—the Dot boots its environment, uses the newly launched Agents API with Computer Use, navigates interfaces, runs terminal commands, inspects diffs, and verifies output against internal company rubrics.

"The paradigm of modern software has been that humans do the thinking and computers do the calculating," said Altman during the opening keynote. "Dots invert this. You delegate an objective, and your personal swarm of Dots works through the night on isolated cloud machines, coordinating across Slack, Jira, Salesforce, and codebases, delivering completed deliverables when you wake up."

"We are moving from AI that talks with you to AI that works for you. A Dot isn't a conversation—it's an autonomous colleague with an operating system, a browser, and a mandate."

2. GPT-6.1 Sol: High Intelligence at One-Fifth the Cost

To make multi-agent loops economically feasible for production deployments running millions of daily inference steps, OpenAI unveiled GPT-6.1 Sol. Designed specifically as the workhorse engine for background agent swarms, Sol delivers reasoning capabilities closely matching the flagship GPT-6 Astra on standard software engineering benchmarks (SWE-bench Verified 2026) while reducing per-token pricing by over 80%.

Complementing Sol is the GPT-6.1 Luna edge variant for real-time local verification, and the introduction of Codex Ultrafast, an inference tier optimized for IDEs and terminal agents that streams up to 450 tokens per second.

3. Enterprise Collaboration: Spaces, Pages, and Identity

DevDay 2026 also introduced ChatGPT Space & Pages, enabling cross-functional teams to share active agent sessions, live slides, and dynamic dashboards. OpenAI additionally rolled out Sign in with ChatGPT, allowing external enterprise platforms to authenticate users and securely hand off authorized agent credentials across enterprise borders.

4. Enterprise Orchestration with SyncFlo AI

As organizations deploy swarms of autonomous Dots, the need for unified governance, lead synchronization, and telemetry management becomes paramount.

SyncFlo AI has launched day-zero support for the OpenAI Agents API & Dots Protocol. With SyncFlo's native orchestration connectors, enterprise teams can manage Dots directly from their existing CRM, marketing pipelines, and sales automation engines—monitoring execution logs, human-in-the-loop approval thresholds, and multi-tenant security boundaries in real time.