Deep Dive
Introduction of Dots
OpenAI introduced a new AI assistant called 'dot' at Dev Day in San Francisco on September 29, 2026. Unlike previous versions of ChatGPT, dots operate continuously on a dedicated cloud machine. They can perform tasks like logging into Gmail, writing code, and alerting users to potential issues. These assistants are designed to work on user goals 24/7, maintaining context between interactions, which marks a significant departure from the stateless nature of prior ChatGPT iterations.
Capabilities and Limitations
Dots are equipped with their own virtual machines, allowing them to perform tasks autonomously, such as browsing the web and running code. They connect to over 4,000 apps, including Gmail and Slack, but are restricted from accessing sensitive data like passwords. OpenAI has implemented an auto-review system to ensure that any significant actions require user approval, maintaining a balance between autonomy and security.
Real-World Applications and Future Plans
OpenAI demonstrated several scenarios where dots could be useful, such as managing schedules, fixing software bugs, and conducting research analysis. However, these were curated demos, and the true capabilities of dots remain to be independently tested. OpenAI envisions future updates where teams of dots work together, but currently, each account is limited to one dot, available only through a paid ChatGPT Pro subscription.