One sentence, 100 steps: YOYO turns the phone into a multi-step AI agent

A report from QbitAI on September 15 describes a simple test of the mobile assistant YOYO: the user spoke one sentence, and the phone carried out a full workflow on its own, running through roughly 100 steps in the process. The report frames the episode as "with one sentence, YOYO took over my life." The important detail is not the outcome of any one answer but the length of the chain — the assistant interpreted the request, broke it into steps, and then tapped, typed and confirmed its way through a sequence a person would otherwise perform by hand. That is a shift in what a phone assistant is for. Earlier voice assistants handled one action at a time, such as opening an app or setting a reminder. Multi-step execution instead requires task planning, a memory of how far a run has progressed, and coordination across apps — plus some way of recovering when an individual step fails. A hundred steps is the number that makes the point: the value of an on-device agent is not how clever any single action is, but how much repetitive tapping and app switching it removes from the user's day. For ordinary users the test is straightforward — fewer taps and fewer app switches. Execution at that scale also raises the practical bar. Longer chains demand broader permissions and carry more chances to go wrong, so being able to pause, confirm or roll back each stage becomes a baseline requirement rather than a bonus feature. What to watch next: which apps and scenarios YOYO opens up to, how the permission and privacy boundary is drawn, and whether success rates become reliable enough that users are willing to hand over an entire workflow rather than a single command.