OpenAI just turned ChatGPT from a chatbot you summon into a worker you employ.
The Summary
- OpenAI launched Dots at DevDay, always-on AI assistants powered by GPT-6 Astra that work continuously in the background across 4,000+ apps
- Direct response to Meta's Muse, but with a key difference: Dots operate independent of specific hardware, pursuing goals with minimal oversight
- Each Dot gets its own cloud computer with web browser access, learning your preferences while you sleep
The Signal
This is OpenAI planting a flag in the agent economy before Meta can claim the territory. Muse launched with cute avatars and persistent context. Dots copied the aesthetic but went harder on autonomy. While Muse tethers to Meta's hardware ecosystem, Dots are hardware-agnostic, a design choice that matters more than the bubble animations.
The GPT-6 Astra engine is the real story here. OpenAI skipped GPT-5 entirely in public messaging and went straight to a model purpose-built for agentic work. Access to 4,000+ apps means Dots can touch your calendar, email, project management tools, and CRM without you building custom integrations. That's the difference between a demo and a deployment.
"Dots operate independent of any specific hardware or interface, pursuing user-defined goals continuously in the background."
The interface spans text and voice, accessible through ChatGPT on web, desktop, or mobile. You message your Dot like a coworker or call it when you need a faster loop. But the critical phrase is "minimal oversight." OpenAI is betting you'll trust a Dot enough to let it work while you're offline. That trust threshold is where adoption lives or dies.
Here's what makes this different from previous "AI assistant" launches:
- Persistent presence, not session-based interaction
- Native app integration at launch, not "coming soon"
- Dedicated cloud compute per Dot, not shared inference
- Hardware independence, making it platform-portable
The timing matters. Meta announced Muse six weeks ago and created the market category. OpenAI is playing fast follower but with deeper enterprise hooks. If Dots can legitimately manage workflows across Salesforce, Notion, Gmail, and Slack without constant supervision, knowledge workers just got their first real AI employee, not another chatbot.
The Implication
Watch which apps get integrated first and how OpenAI handles permissions. The security model for an agent that touches your email, calendar, and financial tools will determine whether this stays in early-adopter land or crosses into mainstream work. If Dots can genuinely reduce context-switching and handle multi-step workflows end-to-end, the market for standalone productivity tools shrinks fast.
The bigger question: what happens when everyone has three Dots handling their scheduling, research, and admin work? We're about to find out if the agent economy creates new work or just makes existing work feel less tedious. Either way, the shift from "ask AI a question" to "assign AI a job" just accelerated.