Same week. Two very different answers to “what should AI get good at next?”
On Sept. 29, 2026 — OpenAI’s DevDay — the company rolled out Dots: always-on personal agents meant to stick with you between chats. Powered by GPT-6 Astra, a Dot gets its own cloud computer, plugs into 4,000-plus apps, and can show up in ChatGPT, Slack, and Microsoft Teams. You hand it a project; it keeps moving, brings work for review, and leans on Auto-review to check actions that could touch accounts or share information against your rules and safety limits. Specialist dots for orgs are in preview. Access is rolling out on ChatGPT Pro, Business Premium, and Enterprise in eligible markets.
(Same day, OpenAI also put GPT-6.1 Sol in the mix as a cheaper near-Astra workhorse for coding and computer use — useful cost context for why continuous agents might actually run all day, not the main plot.)
Also on Sept. 29, Microsoft Research introduced Quine: not a consumer avatar, but a multimodal biology world model wrapped in a harness that connects models, literature, tools, and the wet lab. In work with the Broad Institute of MIT and Harvard, Quine helped prioritize compounds predicted to shift pancreatic-cancer cell states; several top-ranked candidates were assay-validated. Microsoft says the triage — from huge search space to a shortlist worth testing — ran on a weekend scale versus months of conventional experimental grind. Access starts limited via Quine Fellows; as the research matures, the path points toward products like Microsoft Discovery. Quine is research tech, not clinical use.
Put them side by side and the contrast sharpens. Dots aim at everyone with a day to manage — a shipping product that stays on, learns preferences, and earns trust through Auto-review and human approval gates. Quine aims at specialist labs — a research system whose trust test is wet-lab validation, not a password change you kept for yourself. One is product iteration in the open; the other is a discovery loop that only counts when the dish agrees with the model.
Why it matters
Both are “future now,” but they open different doors. Personal agents that stick around could change how ordinary work gets done between meetings. AI that shortens model-to-medicine triage could change how fast scientists pick the next experiment worth running. Neither replaces judgment — Dots still hand you the decision, Quine still needs the assay — but both are bets that intelligence is most useful when it keeps working after you look away.
What’s next
Watch whether Dots expand beyond one primary agent and how specialist org dots behave under real IT controls. On the biology side, watch what Quine Fellows actually ship back as feedback, and whether validated cell-state findings widen into broader discovery products. Two doors; same week; plenty still to prove.
Sources: - Introducing dots (OpenAI) - Dots feature page (ChatGPT) - OpenAI launches Dots, its Muse competitor (The Verge) - OpenAI launches Dots, its bubbly agentic avatar (TechCrunch) - Introducing Quine (Microsoft Research) - OpenAI launches GPT-6.1 Sol (TechCrunch)