OpenAI Dots

What it is

Dots is an always-on AI agent (an AI that handles tasks on its own through multiple steps) released by OpenAI. It researches information and creates documents, presentations and web apps. When you connect apps such as email and calendar, it refers to the information in them, and it also remembers what you told it before and uses that in later tasks. This means you need to repeat the same explanations less often each time you hand off work. Around the same time, Anthropic released a preview of Claude Duet, and services such as Meta Muse are also compared with it.

How it works

Dots works on a separate cloud computer rather than the user's laptop, so tasks continue even after the laptop is closed. Users can talk to Dots not only in ChatGPT but also in the workplace messaging apps Slack and Microsoft Teams. They can also choose where progress updates go. For example, they can ask for routine reports to stay in ChatGPT and only the items that need a decision to be sent to Slack. When a program needs to be built, Dots delegates the work to tools such as Codex, OpenAI's AI coding tool. Users can also open the cloud browser it is working in to check progress, or take control directly and then hand the task back.

Why it matters

It draws attention as a sign that competition among "proactive" AI assistants, which take care of needed tasks without waiting to be asked, has begun in earnest. OpenAI explained that when users edit the wording and structure of a result, Dots applies those edits to later tasks and gradually learns how the user works. Kim Kyung-hoon, head of OpenAI Korea, said Korea ranks third in the world by weekly active users of Dots. There is also a feature for customizing the assistant by choosing its color, eye shape and accessories such as glasses.

Things to know (limitations)

Users can grant permission in advance for Dots to send messages when conditions they set are met, but some tasks require additional approval. It is a good idea to make clear whether you want only a draft or want it actually sent. If you connect a personal computer and let Dots work on it, that computer must be on and online, and the ChatGPT app must be open. In addition, many websites, including Amazon, block AI agents for security and business reasons, and there have been reports of Dots users having tasks blocked. Because it accesses sensitive information such as email and calendars, security concerns also remain an ongoing issue.

Example

In an example OpenAI presented, an interview recording is the starting point. When the user hands over the recording, Dots picks out passages worth highlighting and prepares drafts of a show description and social media posts. If the user edits the style and structure of the drafts, Dots remembers the changes, so the next time the user hands over an interview, they do not have to explain the desired format at length again. Meanwhile, the user can ask Dots to do other tasks or continue the conversation.

Related terms

AI agent Claude Duet Always-on agent Agent memory Computer use Human-in-the-loop

References

• OpenAI's 'Dots' vs. Anthropic's 'Duet': which AI assistant should you hand your work to? (ZDNet Korea)
• Meta Muse vs. OpenAI Dots: Which Autonomous AI Agent is Right for You? (Innovation Village)
• OpenAI's Dots head says AI means workers now have a choice between doing less or more (Business Insider Africa)
• OpenAI releases voice model 'GPT-Live-1.1'... improves speech in six languages including Korean (AI Times)
• Meta Muse vs OpenAI Dots: Should You Trust AI to Handle Everyday Tasks? (Analytics Insight)
• Dots: Proactive AI Tool for Continuous Workflow (Dynamic Business)