I Tested OpenAI's New Personal Assistant Agent: DOTS

**TL;DR:** OpenAI's DOTS is a proactive, always-on personal assistant agent that acts as a central coordinator, accessing all parts of your ChatGPT environment to manage multi-step tasks across chats, workspaces, and coding sessions. ## Introduction to DOTS OpenAI has released a new always-on proactive personal assistant agent called **DOTS**. Early access has been granted for testing, revealing an agent designed to connect and orchestrate all components of a user's ChatGPT ecosystem. It serves as a central control point for tasks spanning multiple conversations, projects, and tools, aiming to provide continuity and reduce the need for switching between different applications. ## Core Concept and Accessibility At its core, DOTS is an agent that has access to your ChatGPT's memory, connected tools, skills, and scheduled tasks. It can initiate new conversations, start new work or coding sessions, and even operate on your computer when you are online. Its primary value lies in its ability to see the big picture across all your threads and projects, coordinating them from a single place. It also possesses its own persistent cloud computer for long-term retrieval and storage of information. As of the test, access to DOTS is currently limited to **Pro and Enterprise users**, with no public timeline for a broader rollout. Once available, it appears in a new "Your Dot" section at the top of the left-side panel in the ChatGPT interface. ## Initial Setup and Customization Upon first interaction, DOTS introduces itself and immediately begins proactive operation. It states: *"Send me a message or call anytime. Even when we're not chatting, I'll keep pushing things forward and stay in touch with updates or questions."* It asks for a preferred name and can automatically generate temporary notes and a list of suggested tasks. A dedicated dot icon opens a right-side panel for customization. Users can change the agent's name and appearance, selecting from various character icons beyond the default dot. Communication options include **call (voice mode)**, email (not yet with a dedicated address), and Slack integration. A key feature is the **Computer Permissions** toggle, which grants DOTS access to its own cloud computer for information retrieval and multi-tab browsing. There is also a "My Computer" option for desktop users, allowing the agent to control all coding threads on the local machine. ## Key Features and Capabilities ### Central Coordination and Long-Term Projects DOTS positions itself as ideal for complex, multi-stage projects that require continuity. When asked why one should use it instead of directly using ChatGPT's workspace or coding features, it responds: *"The biggest advantage of choosing me is continuity."* It excels at tasks requiring the orchestration of multiple elements over time—such as video production involving research, scripting, assets, and publishing checks—where it can manage parallel workflows and decision points. It recommends using the standard ChatGPT interface for quick thinking and reserving the specialized coding tools for deep collaboration. ### Pages: A Persistent Knowledge Base DOTS introduces a feature called **Pages**, a notebook-like system similar to Notion. These pages act as a persistent knowledge base that can be built upon over time, solving the problem of lost context in long, single-threaded agent conversations. Pages are organized with expandable sections and hyperlinks and support the creation of sub-pages. The agent can automatically generate notes based on what it knows and create new sub-pages for specific projects, such as a dedicated page for a "DOTS Launch" video, complete with notes, checklists, and conversation links. ### Multi-Task, Cross-System Orchestration A core demonstration of DOTS involved issuing a complex, multi-part command to test its coordination abilities. The instruction was to use custom "skills" (pre-configured capabilities) to handle three tasks simultaneously: 1. Use the "Intro" skill in a **workspace** to generate three opening paragraphs for a video. 2. Use the best paragraph to create an animation via the **coding** skill. 3. Simultaneously use the **thumbnail** skill in another conversation to generate a video cover image. DOTS successfully delegated these tasks across three separate conversation threads, demonstrating true cross-system scheduling. It managed the workflow automatically, moving from script generation to requesting the use of specific coding and thumbnail skills, and then delivering the results. ### Voice Mode and Cloud Computer DOTS features a voice mode accessible via phone call or desktop interface. During a test, it was instructed to use its browser-based skill to search for "magnetic wooden car model art" and open the findings in new tabs. The agent successfully performed the search, identified specific products (e.g., "handmade walnut stone with hidden magnets"), and presented the links. While the dedicated cloud computer experienced a glitch after a system reset, DOTS effectively used an alternative method (a browser within a coding tool) to complete the task. ### Task Automation and Scheduling The agent supports setting up recurring tasks, triggers, reminders, and automations. A simple test case involved scheduling a reminder for the next day: *"OpenAI DevDay is tomorrow. Remind me to watch the keynote and give me a summary of all announced releases right after it ends."* These tasks can be triggered by various events, such as form submissions, email receipts, scheduled daily/weekly runs, or one-off events, with notifications routed directly to the DOTS interface. ## Testing and Observations During the testing period, DOTS demonstrated strengths as a central project manager for multi-threaded, multi-project workflows. It provides a unified task distribution platform, making it useful for remote work initiation from mobile devices. The skills it uses (like intro, coding, and thumbnail generation) were developed separately and then imported for demonstration, highlighting the agent's ability to integrate external capabilities. However, the experience was not without issues. After resetting the agent, some features like the cloud browser functionality failed, and the scheduled tasks interface was unresponsive—possibly due to bugs in the beta desktop application. The creator noted that while the skills used in the demo are still being refined for stylistic consistency, the overall orchestration of complex, parallel tasks worked well. ## Conclusion and Future Potential DOTS is best understood as an evolving central coordinator for the ChatGPT power user. It addresses the pain point of context-switching between disparate tools by offering a single point of control for ongoing projects. Future planned features include support for multiple DOTS instances, dedicated email addresses, and payment capabilities. As the tool matures and access expands, the community is expected to discover and share a wide range of new use cases and experimental workflows. Source: [https://www.youtube.com/watch?v=V_1Vn2WfpEY](https://www.youtube.com/watch?v=V_1Vn2WfpEY)