OpenAI Unveils 'Dots' Persistent AI Agents at DevDay 2026

OpenAI introduced Dots at its 2026 DevDay conference in San Francisco — a new class of persistent AI agents built on the company's GPT-6 Astra foundation model. Unlike conventional chat interfaces that respond only when prompted, these agents operate continuously in the background, monitoring connected applications, crawling the web, and executing multi-step workflows on the user's behalf.

Always-on architecture with cross-platform reach

Dots appear as customizable animated blobs that users can assign complex projects such as travel planning, code deployment, or calendar-driven meal ordering. In one demonstration, a Dot detected a calendar conflict during dinner hours, fetched two GrubHub options with pricing, and awaited confirmation before placing the order. Agents are accessible through ChatGPT, Slack, Microsoft Teams, and — for Pro subscribers — iMessage and RCS on Android, with shared context across all channels.

Rollout begins today for ChatGPT Pro members at $100 per month, with broader availability planned. OpenAI indicates support for multiple concurrent Dots will follow, enabling parallel task execution.

Why it matters for GPU / AI infrastructure

Persistent agents that continuously ingest context, invoke APIs, and maintain long-running reasoning loops impose sustained inference demand rather than bursty request patterns. This shifts workload profiles toward always-on GPU capacity with low-latency interconnects — precisely the infrastructure profile AiGpu's Dubai cloud is architected to deliver. Enterprises adopting agent fleets will need predictable, scalable compute with tight data residency controls.

Safety controls include mandatory explicit approval for sensitive actions (software installs, credential changes) and a Custom Rules framework for defining operational boundaries. The launch positions OpenAI in direct competition with Meta's Muse agent, signaling an industry-wide pivot toward ambient, app-connected AI labor.