dots vs ChatGPT Tasks: how they actually differ

ChatGPT scheduled tasks carry out a scoped action at a defined moment. A dot pursues an objective continuously, has its own cloud computer, and keeps working when you are away. The difference is not one of degree: it is the step from run this to own this.

The comparison, point by point

Criterion	Scheduled task	dot
Unit of work	One action at a set time	An objective or responsibility
Continuity	No: it fires and ends	Yes: keeps context across conversations
Execution	Limited to what the feature allows	Own computer, browser, code
Channels	Inside ChatGPT	ChatGPT, SMS, Slack, Teams
Initiative	None beyond the trigger	Proactive research in the background
Control	Simple configuration	Per-action rules, auto-review, approvals
Cost	Uses ChatGPT limits	Conversation does not; Codex and ChatGPT Work do

Where a scheduled task wins

It would be dishonest to sell dots as a universal replacement. A scheduled task wins when:

• The work is repetitive and predictable.
• The output is scoped: a summary, a reminder, a periodic lookup.
• No accumulated context is needed: each run is independent.
• You want zero configuration: no permissions or rules to define.

A daily headline summary is a task. It needs neither a computer nor a set of approval rules.

Where a dot wins

A dot earns its place when any of these is true:

• The work is a project that changes shape over time.
• It requires crossing several tools, not just one.
• The value comes from something keeping moving while you are not looking.
• You need executed output, not a suggestion: tested code, a finished draft, an updated proposal.

OpenAI’s example — a dot that reads customer feedback, fixes bugs and hands over pull requests with a video attached — is not something a scheduled task can do. It requires browsing, writing code, testing it and iterating.

The practical question

Before deciding, answer one thing: does the work end when it fires, or does it change shape as it goes?

• If it ends when it fires → scheduled task.
• If it changes shape → dot.

What a dot adds in scheduling practice

If your case is “remind me of this every morning”, a task is enough. But a dot handles the full cycle in ways a scheduled task cannot:

• Tasks live in the dot’s profile, organised under “In progress”, “Scheduled” and “Completed”. You can open one and change its repeat schedule, time or completion notifications.
• It can be paused entirely when you do not want it working, and resumed afterwards.
• It detects changed context. A task fires regardless of whether the world changed; a dot can reassess whether the assignment still makes sense.
• It executes, it does not just notify. If the morning review concludes something needs moving, the dot can do it within the permissions you granted.

That is the difference: the task informs you, the dot can close the loop.

What is still unclear

OpenAI has not published a feature-by-feature comparison between dots and ChatGPT’s earlier agent and task features, nor a definitive list of which functions are superseded. Exact consumption numbers are not published either. If your decision hinges on a very specific use case, test it with the dot included in your plan before reorganizing your workflow.

Verified against the 29 September 2026 announcement. ChatGPT features evolve alongside dots, so this table can go stale: check the article’s update date.