OpenAI Launches Dots: Persistent AI Assistant Evolves from Coding Tools

OpenAI has unveiled dots, an always-on personal assistant that marks the company's latest step toward persistent, background AI agents. Announced at the firm's developer day in late September, dots is available to ChatGPT Pro subscribers starting at $100 per month and connects directly to calendars, email, local files, and other desktop tools to anticipate user needs without explicit prompting.

From Codex to Continuous Agent

The product traces its lineage to OpenAI's Codex coding assistant, released earlier this year. According to Alexander Embiricos, who leads enterprise product, dots repackages the same agentic capabilities into a single, continuous conversation rather than discrete coding sessions. The goal is to make proactive assistance feel intuitive — an entity that "thinks about you and how to help you" rather than a tool awaiting commands.

Early testing shows the assistant can flag calendar conflicts and upcoming deadlines, though timing remains a work in progress. Embiricos noted his own dot once warned him of a meeting three minutes before it started. The team expects latency and relevance to improve as the system learns user patterns across integrated data sources.

Why it matters for GPU / AI infrastructure: Persistent agents like dots imply sustained, low-latency inference workloads running 24/7 across user devices and cloud backends. This shifts demand from bursty, prompt-driven spikes to continuous token streams — a pattern that favors always-warm GPU capacity, efficient KV-cache management, and inference-optimized serving stacks. For infrastructure providers, it signals growing need for reserved, high-utilization GPU fleets rather than purely on-demand scaling.