Wired tests six personal AI agents and ranks OpenAI Dots first

In short: A WIRED reporter tested six “always-on” AI agents for a day and found OpenAI’s Dots performed best overall, with Meta’s Muse close behind.

What's going on

AI agents are starting to work better than they did a year ago, according to a new hands-on test by WIRED. An AI agent is like a texting assistant that can also take actions for you, such as checking your email, watching your calendar, and using a web browser (like a helper who can click around websites for you).

The reporter set up six agents on an iPhone and ran them through basic, everyday tasks over 24 hours. The tools tested were Meta Muse, OpenAI Dots, Instinct, Anthropic Claude Duet, Google Gemini Spark, and xAI Grok Bot. Prices ranged from free (Muse and Instinct, though Instinct required an invite) to $100 per month for Dots.

In the tests, Dots did best overall. It scored highly on ordering lunch, spotting a scheduling conflict between an email reminder and a calendar event, and offering useful late-night “what should I do next” support. Muse was ranked the most accessible option because it was free and easy to set up, and it also did well on travel planning.

Other results were mixed. Claude Duet was praised for having the best app design, even though it did not top most task scores. Instinct was labeled the most annoying, partly because it runs through text messages only, which made it harder for the tester to see what it was doing.

What to watch

These tools often work best when you connect personal data like email and calendars, which raises privacy and security questions. As more companies push agents that can act on your behalf, it will matter how clearly they explain what data they use and how they avoid embarrassing mistakes.

Source: Wired