OpenAI launches Dots, its Muse competitor

Emma Roth reports OpenAI’s answer to Meta’s Muse: customizable always-on agents that OpenAI says can “do nearly anything” across connected apps while learning preferences. GPT-6 Astra powers each Dot’s cloud computer and browser; interfaces include chat-like messaging, voice in ChatGPT, Slack/Teams with carried context, and soon SMS.

One Dot at launch, multiples later. Examples include developer feedback-to-PR loops with video walkthroughs and creator clip/show-notes/social drafts. Custom rules and auto-review gate sensitive actions (e.g., password changes). Muse safety concerns — including a report of an address shared on Marketplace without knowledge — hang over the category. Specialist enterprise Dots are in testing; Dot chats don’t count toward ChatGPT usage limits.

We’re measuring Dots against Muse’s failure modes, not just its download charts.

Always-on plus “before you even think to ask” is the feature and the hazard. Auto-review and custom rules are the right language; Muse’s alleged address leak shows why defaults matter more than avatars. Useful agents that draft PRs and show videos of changes can genuinely free humans for harder work.

I’m watching Space collaboration and specialist Dots — that’s where workplace data concentration spikes. Paid tiers first is responsible packaging if OpenAI uses the time to harden approvals.

We’ll push for agents that earn proactivity. We’ll treat every cute blob as a privileged identity until proven otherwise.

The Verge, September 29, 2026 (Emma Roth).