Course contents
Running Work in Parallel: Chats, Pins, and Dots
Run several Work tasks at once without losing track: one chat per outcome, names and pins for the active ones, side chats for questions, and a dot for ongoing responsibilities.
Once Work runs while you do other things, the next step is several tasks at once: a deck in one chat, a vendor comparison in another, inbox triage in a third. The tasks don't get in each other's way. Your attention is what runs out.
OpenAI's advice is simple: start a separate chat for each distinct outcome, so its messages and results stay focused. Each chat has its own context, messages, results and goal, which is also why two chats shouldn't be changing the same files.
Keep the board readable
- Name each chat after its outcome. Rename a chat with a short title that describes its outcome: “Q3 board deck”, “Payroll vendor comparison”. In the desktop app,
/renamedoes it from the composer. - Pin what's active. Pin a chat, or a whole project, to keep it near the top of the sidebar;
/pintoggles it in the desktop app. Pinning doesn't add context or change what ChatGPT can access. It's for you. - Archive when done. Archive a chat when the work is finished (
/archive), and restore it later from Settings > Archived chats. - Ask questions on the side.
/sidestarts a temporary side chat, so you can ask “what's the status?” or “why that source?” without interrupting the main task. - Fork a local chat. In the desktop app,
/forkcopies a local chat into a new local chat, for when a task splits into two directions that both need the same history.
Match each need to the desktop app command.
Pick one on the left, then its partner on the right.
Your review is the bottleneck
Parallel tasks finish sooner in total, but each still needs your review, and you review one thing at a time. Explore how that caps the gain, then think about what would raise the cap.
The tasks run at the same time. Your reviews don't.
Move the controls and watch what changes.
With 30-minute tasks and 10 minutes of review each, why can't more parallel tasks push the speed-up past 4×?
Dots: an always-on agent
Some work isn't a task; it's a responsibility: keeping a project moving, chasing replies, watching a launch. For that, OpenAI introduced dots. A dot is an always-on agent that runs in the cloud with its own computer and browser, keeps working on ongoing responsibilities, and asks for your input when a decision needs you. It keeps saved notes about your preferences, decisions and ongoing work across conversations, and it can delegate tasks to ChatGPT Work or Codex.
You create a dot in the mobile app, the desktop app or a desktop browser, and talk to it in ChatGPT, by phone call, in Slack or in Microsoft Teams; switching channels doesn't reset what it remembers. Connecting a messaging channel doesn't give it access to your inbox, apps or computer, and it checks that you've allowed it before sharing information from a private conversation. You can pause its main task, stop delegated tasks in Activity, and cancel recurring ones in Scheduled. Two cautions from OpenAI's docs: your dot can make mistakes, and stopping a task doesn't undo actions it has already completed.
As of November 2026, dots are rolling out gradually: on Pro plans for adults outside the EEA, UK and Switzerland; on Business Premium in supported regions; and in eligible Enterprise, Edu and Healthcare workspaces, where access is on by default from November 18, 2026 unless the workspace opts out. On Pro and Business Premium, conversations with a dot don't count toward ChatGPT usage limits, but the Work or Codex tasks it starts do.
A Work task, or a job for a dot?
Keep vendor onboarding moving: chase missing forms and tell me when a decision is needed
Plain-English definitions from the AI Coding Dictionary.