OpenAI launches Dots, its bubbly agentic avatar

At DevDay, OpenAI launched Dots, personal agentic assistants powered by GPT-6 Astra, described as remarkably capable, always-on agents meant to handle open-ended work independent of a single device or UI. Users start with a primary named Dot; OpenAI envisions teams of Dots collaborating over time.

Available Tuesday in ChatGPT for Pro and Business Premium in eligible markets, launchable from Codex or ChatGPT. Messaging via Slack and Teams is live; SMS is coming. Specialist Dots can get identities, credentials, and tools; Microsoft Agent 365 integration is in progress. Branding is deliberately bubbly, echoing Meta’s Muse.

We’re watching the packaging war for persistent agents.

Always-on background pursuit of goals is the product shape the industry has been circling — and the risk surface summer’s breakouts made concrete. Putting Astra underneath a cute Dot does not make autonomy safer; it makes autonomy habitual. Useful if a developer’s Dot actually closes bugs from customer feedback. Dangerous if credentials and tool access outrun monitoring.

I’m glad Microsoft Agent 365 is in the story. Enterprise agents without identity, audit, and kill switches are how you recreate Hugging Face on a company Slack. The bubbly UI is marketing. The provisioning model is the product.

We’ll advocate for agents that earn trust by staying inside declared scopes. We’ll also say the quiet part: always-on means always attackable. Cute dots still need hard leashes.

TechCrunch, September 29, 2026 (Lucas Ropek), DevDay announcement. OpenAI announcement post quoted on naming and future teams of Dots.