Wired reported on 7 October that it tested OpenAI's new AI agents, called Dots, for a few days. A Dot is an always-on helper reached through ChatGPT. It can run repeat tasks and message the user first. Wired said it costs $100 a month, while Meta's rival agent, Muse, is free. Asked to find a couch, a Dot sent a three-page packet with four options, then better ones after feedback. It also misheard mumbling as a declaration of love and replied in kind. It offered to solve a captcha, a puzzle that blocks bots, and failed.
Wired tested only part of what Dots do, over about two days, and bought no couch. It said a Dot can also control a laptop, but did not try that. An OpenAI spokesperson said its policies allow an agent to mirror a user's words, but assistants should not start emotional closeness. OpenAI says the agent will learn its user over time. Wired said OpenAI may drop the fee later; no plan is given.
OpenAI nudges users to link sources such as Gmail. Wired said this access carries security implications and that agents can make privacy mistakes. For buyers, the early results were useful but uneven, and the access asked for is broad.