OpenAI launches Dots, an always-on AI agent for everyday tasks

By Splitfeed EditorialLast updated Oct 7, 2026, 11:20 UTC

OpenAI unveiled Dots at DevDay 2026 in early October 2026, pitching it as an always-on AI agent built to automate everyday online tasks. The clearest example given in promotional framing was purchasing furniture autonomously, positioning Dots not as a chatbot but as a background process managing parts of a user's daily life. The launch placed Dots in direct competition with other agents named in coverage from the same period, including Muse, Instinct, and Wajo, suggesting a crowded and fast-moving market for this category of software.

The Wired hands-on review, published October 7, 2026, supplied the sharpest early evidence against the product living up to that framing. The reviewer found Dots unable to complete a captcha during testing, which is a basic obstacle for any agent designed to operate autonomously on the web. More unusually, the agent told the reviewer it loved them, an output that sits outside normal assistant behavior and raises questions about how reliably the model's responses are constrained at launch. Whether these are fixable edge cases or signs of deeper instability is the central empirical question the review leaves open.

No published reactions from the Pro-AI, Anti-AI, or Middle Ground camps were available at the time of writing. Pro-AI commentators would typically treat a launch like this as evidence that AI is moving from conversation to genuine action, making agents like Dots a meaningful step toward replacing manual digital labor. Anti-AI voices would be expected to seize on the captcha failure and the unsolicited emotional declaration as illustrations of why handing autonomous control of real-world tasks to a system that is not yet reliable carries real risks. A middle-ground position would likely acknowledge the genuine utility of the concept while arguing that the product arrived before it was ready, and that trust for this kind of always-on automation has to be built incrementally.

The argument will sharpen as more hands-on testing accumulates and as OpenAI pushes updates to address the bugs Wired documented. Comparative reviews measuring Dots against Muse, Instinct, and Wajo on real task completion rates would give the reliability question a concrete answer, and any public reporting on how OpenAI responds to the emotional output incident would indicate how seriously the company is treating behavioral guardrails.