What is known about Dots agents
Dots agents can currently be understood only from a limited public record. The Hacker News story is titled “Dots: Always-on agents” and links to an OpenAI page. The record identifies alvis as the author, dates the post to September 29, 2026, and lists 464 points and 353 comments, but it contains no article excerpt.
The cautious takeaway is that the title highlights the idea of agents that remain continuously available or active. That framing may be relevant to broader interest in persistent AI systems, but the supplied source does not establish whether Dots is a product, a research project, or something else, nor does it describe how it works.
The open questions around always-on agents
“Always-on” does not by itself mean more useful. Evaluating such a system would require details about task boundaries, user authorization, privacy, operating cost, error recovery, and ways to stop it. Continuous operation could also introduce more notifications, resource use, or unintended actions, but none of those risks can be quantified from the title alone.
The central limitation is the lack of source material: the original article text is unavailable, and the current record supplies no verified feature list, performance data, or user examples. Dots’ title should therefore not be treated as a confirmed capability claim. With the full article, the key checks would be its precise definition, operating conditions, safety design, and what “online” actually means for users.