OpenAI launches dots, always-on agents
OpenAI launched dots, GPT-6 Astra-powered agents with dedicated cloud computers that work continuously across connected apps, Codex, Slack, and Teams. They learn user preferences, execute multi-step workflows, and request approval for consequential actions.
dots moves AI from answering prompts toward delegated, persistent work—but its success will depend more on reliability and permission controls than agent personality.
- –Dedicated cloud computers let dots continue research, coding, analysis, and content tasks without constant supervision.
- –Developers can delegate bug triage, implementation, testing, and pull-request preparation while retaining review control.
- –Connections to more than 4,000 apps make dots broadly useful, but also expand the blast radius of permission mistakes.
- –OpenAI’s approval gates, custom rules, activity monitoring, and read-only proactive research are essential safeguards, not optional polish.
- –The initial Pro, Business Premium, and Enterprise rollout positions dots as a premium power-user agent competing in an increasingly crowded market.
DISCOVERED
1h ago
2026-09-29
PUBLISHED
1h ago
2026-09-29
RELEVANCE
AUTHOR
OpenAI