Hands-on with ChatGPT's always-on agent Dot: it read Slack to flag a flight conflict
every · x · 2026-09-30
Every's Dan Shipper shares day-zero testing of ChatGPT's new always-on agent Dot. Each Dot gets its own cloud computer and works across connected accounts like email, calendar, and Slack, persisting between conversations even with the laptop closed.
Standout moment: while rebooking a flight out of San Francisco, Dan's Dot "Boo" warned it might overlap with a video recording—not on his calendar, but inferred from a Slack discussion he hadn't seen. After days of testing, Dan was reaching for Boo instead of starting new ChatGPT or Codex threads.
Rough edges remain: repeated permission prompts, awkward logins, and messages that sometimes get no reply. Verdict: worth a spin for heavy ChatGPT users; wait a week or two if you need reliability.
Related event: Hands-on with OpenAI's Always-On Agent Dot(4 posts)→
More from coding & agent
- Loop engineering: pushing the agent harness harder boosts long-horizon model capabilities — omarsar0 · 2026-09-30
- Faster AI coding can mean harder engineering, not easier — rseroter · 2026-09-30
- OpenAI rewrote ChatGPT Web on the desktop codebase in 5 weeks — nicoalbanese10 · 2026-09-30
- From 'when to stop' to loop engineering: agent harnesses shape model training — omarsar0 · 2026-09-30
- Swarms revives swarms-rs, an enterprise-grade multi-agent orchestration framework in Rust — KyeGomezB · 2026-09-30
- Codex CLI update adds /usage analytics, managed worktrees, in-terminal Mermaid and LaTeX — OpenAIDevs · 2026-09-30