(SINGAPRE 2026.5.14) On May 12, Air Force One carrying US President Donald Trump crossed the Pacific once again, heading toward China for the first time in nine years.
The moment carried clear symbolic weight: a familiar geopolitical circuit briefly reactivating after a long pause.
On the same day, a different kind of movement was unfolding in the tech world. Manus released a product update introducing browser usage support. On paper, it was a routine incremental feature. In context, it signalled a company still actively building in a landscape that had shifted around it.
Manus was no longer operating in an open frontier where it could easily define categories. Instead, it was evolving inside an ecosystem it had helped accelerate, where competitors were converging on similar ideas, noted a column in the Chinese tech-focused media outlet 36 Kr.
Manus—China-based, Singapore-registered, and previously almost acquired by US’ Meta —has spent the past year building an AI assistant designed to behave less like a chatbot and more like an operational system.
Instead of responding only to isolated prompts, the agent can operate inside a Chrome browser, move across websites, execute multi-step tasks, and preserve context over time. In practical terms, it is meant to function like a digital worker that does not forget what it is doing.
This ambition sits under what Manus calls an “agent as operating layer.” The idea is to move beyond fragmented software environments where users constantly switch between apps and manually connect workflows. In Manus’s vision, an AI layer coordinates actions across tools. A single request like “research competitors and summarize pricing” would trigger an end-to-end workflow: browsing, extracting, and assembling results automatically.
That vision, however, now evolves under changed external conditions. Late last month China’s National Development and Reform Commission has blocked Manus’ acquisition by US tech giant Meta, reportedly valued at around US$2 billion (about S$2.7 billion). What might have been a conventional exit path instead became a geopolitical issue involving cross-border technology and strategic competition between China and the US.
Despite uncertainty, however, product development continued. Manus shipped desktop integrations, cloud automation tools, and browser-based control systems. But in a crowded AI agent market, competitors increasingly converged on similar feature sets, making differentiation harder. As the column Appso observed, the field was compressing into overlapping approaches rather than distinct breakthroughs.
Earlier, just three days after the acquisition block, Manus has announced another feature called “Cloud Computer,” reinforcing a pattern of external turbulence paired with continuous release cycles. The company was effectively “shipping through a storm,” as some commentators described it.
To understand its trajectory, it helps to revisit its breakout moment. In March 2025, Manus gained attention through a viral demo of an AI agent operating inside a virtual computer environment—writing code, searching the web, and completing multi-step tasks with minimal supervision. Interest surged rapidly, and the startup moved from obscurity to global visibility.
By mid-2025, its company structure shifted. Manus closed its China office and relocated headquarters to Singapore. Acquisition discussions with Meta circulated, alongside reports of unusually fast revenue growth. The company briefly looked like a classic breakout startup defining a new category.
By early 2026, however, the environment had become more constrained. A key update by Manus introduced the “Skills” capability, which allowed users to package workflows into reusable modules. Instead of rebuilding task logic each time, users could store and reuse structured processes like competitor research or data extraction.
This aligned with broader industry movement toward reusable agent frameworks, as companies like Anthropic and OpenAI pushed toward more standardized tool use. It also reduced inefficiency, but increasingly reflected development convergence rather than differentiation.
On March 16, Manus released “My Computer,” a desktop integration that connected its cloud agent to local machines, allowing access to files and desktop applications. This marked a shift from isolated virtual environments toward hybrid systems that reflect real workflows. Two weeks later, mobile remote-control features extended this further, enabling task monitoring and triggering across devices.
Then regulatory pressure intensified. Reports indicated that the founders were summoned for discussions in Beijing as review processes on the Meta acquisition advanced. By April 30, Manus launched “Cloud Computer,” an always-on Ubuntu-based environment capable of turning natural-language requests into scheduled automation. A prompt like “collect pricing data every morning” could now be translated into persistent workflows. Ubuntu is a popular free operating system based on Linux.
Yet limitations remained: heavy reliance on command-line operations, limited graphical integration, and a credit-based pricing model that complicated cost predictability. Meanwhile, regulatory uncertainty culminated on April 27 when Chinese authorities formally blocked the Meta acquisition, shifting attention from product iteration to geopolitical implications.
Manus’ trajectory can be divided into two phases. In 2025, it appeared to help define the emerging category of AI agents through compelling demonstrations and rapid adoption. In 2026, its features increasingly aligned with broader industry trends rather than setting them: reusable workflows, hybrid desktop integration, and cloud automation all mirrored parallel developments elsewhere, Appso pointed out.
By May 12, the browser-based update—persistent Chrome sessions and cross-device continuity—represented another incremental step rather than a category shift. Manus remains a capable system for structured automation and research workflows. But the early sense of frontier definition has narrowed, Appso lamented.
The company still carries its founding metaphor: the idea that small actions can produce large consequences, that is, butterfly effect. In its early phase, it helped shape what “AI agents” would become. By 2026, however, it is operating within a field that has largely converged. It is still moving. It is still shipping. But it is now moving inside the system it helped bring into existence, rather than alone defining where it goes next.
If Manus has lost its trendsetting aura in the market—and if past cases are any guide—the founders are unlikely to see this as the end of their story. Instead, they would see it as a phase transition: a shift from defining a category to defending a position within one that now exists.
At that point, startups typically face two paths. They either gradually fade into competition driven by feature parity, or they manage to spark a second wave of growth by carving out a narrower, more focused niche.
Meanwhile, Meta has not officially described the failed acquisition as a permanent loss of interest, but rather as something constrained by external approvals. As capabilities across competitors converge, Meta’s appetite for acquiring Manus would likely diminish. However, its broader interest in the AI agent space would remain intact—and may even sharpen—driving it to double down on in-house development or pursue selective acquisitions of teams and assets that strengthen execution.
