Dots, Grok Bot, Muse push always-on AI agents mainstream

This wave did not appear out of nowhere. The author frames it as the payoff of a shift that began late last year and early this year, when OpenClaw showed that an agent could connect to mail and messengers, accumulate notes about its user, and run continuously. The current services are, in that sense, doing what OpenClaw already did — the difference is that setup and management are now minimal, with the processing moved onto cloud infrastructure rather than the user's own PC. That is why what once demanded a Mac mini or Mac Studio and real technical knowledge can now be used from a phone.

The competitive field is also wider than the three headline names. Manus is rolling out Cue as part of its Manus 2.0 update, giving it its own phone number and email address so it can contact others on its own. TANKA, which recently launched its full service in Japan, targets teams and small and mid-sized companies by remembering internal communication and prompting decisions. Microsoft is packaging the same idea for large companies as Autopilot, and the author expects other enterprise software vendors to follow the same route.

What the outcome hinges on is partly cost and partly trust. Always-on agents need enough compute to maintain reasoning and analysis, which pushed early users toward local AI to escape cloud bills. Handing over mail, calendars and files also raises the risk of an AI mishandling data rather than spying — Muse has already drawn complaints in the US about information users did not intend to share. Meta's Secure VM and its plan to hide contents even from Meta, along with Apple's stricter full disk access controls, suggest the market is moving toward clearer limits, though whether that is enough to win over cautious users is likely to be the test.