What Landed
At DevDay, OpenAI launched Dots, a persistent agent with its own cloud computer, memory, and background execution. It is reachable from ChatGPT, Slack, and Codex, and the first Dot comes with Pro plans starting at $100/month. Nate B. Jones argues that agents now share one form factor and differ on model, harness, and job tuning. He cites Meta's Muse (free tier, over 5 million downloads) as the comparable product. Three things matter for buyers: shared Spaces and Pages, Team Tasks for Business and Enterprise, and Dots integration into Microsoft Agent 365 controls.
Why It Matters
Model routing is now the main cost lever: GPT-6.1 Soul reportedly approaches Astra on selected evaluations at about one fifth the token price. On October 29, existing $200 subscribers' grandfathered allowance ends and the same price buys less usage, so plan pricing is moving toward metered value. The Agent 365 integration matters more than the feature list. It is how an employer governs agents it did not procure, and it points toward multi-vendor agent estates rather than standardizing on one vendor. Jones's claims that outcome specification is now the binding constraint, and that adoption depends on implementation work, are his predictions and are not yet verified.
Worth Raising With Customers
- Measure before upgrading. Instrument usage per workload and send routine tasks to cheaper models before committing to higher tiers. Pressure-test any $200 tier commitments before October 29.
- Govern across vendors. Ask how agents that employees bring in themselves (Dots, Muse, others) get discovered and controlled. Require management-plane coverage that is not tied to one vendor.
- Set context boundaries for shared workspaces. Private chat memory stays out of shared Spaces by design, but nothing yet carries decision reasoning from private chats into team spaces. Customers piloting Spaces need norms for what context gets shared, and who decides, before rollout.