In the breakneck commercial race around AI agents, few companies have completed a journey as dramatic as Manus: a nine-month roller coaster stretching across China and the United States.
In late 2025, it stunned Silicon Valley with more than $100 million in ARR and rattled almost every major tech figure. Among them was Meta founder Mark Zuckerberg, who, rumor has it, needed little more than ten days to push through a lightning $2 billion acquisition of Manus.
Control over this core AI agent technology then acted like an emergency brake, shoving the deal toward forced unwinding.
Today, the team that was born in China and later moved through Singapore has not fractured under the gravitational pull of a giant. It has instead staged a comeback against the wind.
Just as reports circulated that it was about to close a new $500 million financing round that would double its post-money valuation to $4 billion, the Manus team announced a clean break from Meta and a return to independence. Today it delivered its answer:
A fully re-architected Manus 2.0, plus a new standalone app for personal agents called Cue.
From tearing up the underlying execution framework to confronting human-AI co-creation workflows, the team that first ignited the “general-purpose agent” idea is trying, through a systematic upgrade across every dimension, to answer the fundamental question again: what should the next generation of agents actually look like?
