How OpenAI turned its coding tools into dots, an always-on personal assistant

If you pay for ChatGPT Plus, your AI assistant can now proactively work on your behalf and initiate tasks without explicit prompts.

OpenAI has launched "dots," an always-on AI personal assistant designed to work proactively in the background for subscribers of its Pro tier or higher, starting at $100 per month. The assistant, integrated into ChatGPT's desktop and mobile apps, connects to user calendars, emails, and other tools to assist with tasks. Initially developed from its Codex coding tool, "dots" aims to provide a more intuitive and persistent AI experience. Alexander Embiricos, OpenAI's enterprise product lead, explained that framing the assistant as a single, intuitive entity, rather than a series of separate conversations, was key to its design. The agent proactively researches ways to be helpful, a feature Embiricos calls "proactive research." While "dots" can flag issues like calendar conflicts, its timeliness and the granularity of its controls are still under development. OpenAI acknowledges the tension between the data required for an AI agent's usefulness and user privacy concerns, stating that "dots" offers similar privacy controls to ChatGPT, including the option to disable model training.

The introduction of "dots" signifies a shift from task-specific AI tools to persistent, proactive personal assistants, mirroring industry trends seen with Meta's Muse and Instinct. OpenAI's strategy focuses on user experience and intuitive framing to drive adoption, moving beyond complex coding tools like Codex. This evolution suggests a market increasingly valuing AI that can anticipate needs and operate autonomously, potentially increasing user engagement and dependency.

The development of "dots" highlights the ongoing challenges and ethical considerations surrounding AI agent data privacy and control. While OpenAI offers opt-outs for model training and resource access controls, the agent's need for broad access and its proactive nature raise questions about data exposure. As AI agents become more integrated into daily workflows, the tension between their operational requirements and user privacy will remain a critical area to watch, impacting how users trust and adopt these advanced AI services.