Is this new Chinese AI even better than DeepSeek?

Each month brings another groundbreaking development in AI, only for it to be swiftly overtaken by the next. Manus, launched by the Chinese tech firm Monica.im, claims it’s not just hype, though. Unlike the chatbot-style assistants we’ve grown accustomed to, Manus is an autonomous AI capable of independently performing complex tasks without requiring human prompts.

Manus’s capabilities, at least as demonstrated in its promotional video, are impressive. It swiftly processes dozens of job applications, identifying the ideal candidate; conducts sophisticated financial analysis, complete with an interactive, data-driven website; and rapidly generates detailed reports — all by scouring data on its own, generating its own commands, and executing them without human oversight.

Given its Chinese origins, comparisons to DeepSeek were inevitable. Readers may recall how, last month, the Chinese AI reasoning model briefly sent Silicon Valley into a frenzy. Likewise, Manus is already drawing similar praise: the head of product at the computing company Hugging Face called it “the most impressive AI tool I’ve ever tried.” AI policy researcher Dean Ball described it as “the most sophisticated computer using AI.” Demand has surged — its official Discord server amassed 138,000 members within days, and invite codes are already selling for thousands on the Chinese reseller app Xianyu.

Is Manus another DeepSeek moment? The market reaction has been notably less dramatic — Meta’s shares fell just over 4% compared to the trillion-dollar shake-up DeepSeek triggered. Yet, to understand why investors — and the rest of us — should still be paying close attention, we first need to grasp how AI actually progresses. Broadly speaking, it advances along two fundamental axes: unsupervised learning and reasoning.

Unsupervised learning is the brute-force approach — the firehose of data poured into a model until it begins to recognize patterns at a scale no human could match. This is how AI made its great leaps over the past decade: sheer scale. More parameters, more data, more training cycles. In AI, quantity has a quality of its own.

Then there’s scaling reasoning — not just feeding models more data but teaching them to think in structured steps. This approach makes AI methodical, capable of answering logic problems, conducting multi-step research, and reasoning like a human expert. OpenAI’s o1 and DeepSeek’s R1 were examples that pushed this boundary, making AI more deliberate and precise.

But a third paradigm is emerging: agency, or “agentic AI.” Until now, generative AI has primarily been about retrieving and generating information—you type a prompt, it responds. Useful, certainly, but passive. Agentic AI is different; it’s about active execution — AI that independently takes action rather than simply waiting for instructions. While reasoning makes AI better at thinking, agentic AI makes AI better at doing.

This is what Manus represents. Artificial intelligence is moving from automation to autonomy. No longer a reactive assistant, it’s becoming an autonomous operator, capable of independent decision-making and task execution.

Research, analysis, report generation — entire categories of work that once required teams of employees are being compressed into minutes. As these agents become more adept at reasoning, planning, and acting, the real question is no longer how they assist workers, but how much human involvement remains necessary at all.

Manus is part of a broader evolution unfolding across AI labs in both the West and East. Like OpenAI’s Operator, Cognition’s Devin, and Anthropic’s Computer Use, Manus belongs to a wave of AI agents shifting to active players.

As they transition from automation tools to decision-making agents, the question is not only about their capabilities but also about who controls them. Concerns over Manus’s Chinese origins serve as a reminder that AI is inherently political. It is not neutral. It never has been, and it never will be. Ideology is embedded in training data and reinforced through feedback loops. AI is a tool of power; the only question is who wields it.

In the cases of Manus and DeepSeek, the answer is clear. These are not just tools but strategic assets that ultimately answer to the Chinese Communist Party (CCP). Their deployment as instruments of the state is not a matter of if, but when and to what extent. A system capable of navigating networks, monitoring and controlling screens, and executing tasks with minimal oversight is also one that can be weaponized—not just in cyberwarfare or economic competition, but as an extension of state surveillance. In the past, Chinese self-driving cars sparked security fears—not just over data collection, but over what happens when control is in the wrong hands. The same concerns will follow AI — and rightly so.

The stage set by Manus promises something even more transformative, for both AI and the CCP: multi-agent systems. We are moving beyond solitary AI agents solving isolated problems into interconnected networks of specialized digital minds. Imagine an integrated team of AI systems, each an expert in its own domain, collaborating seamlessly to achieve complex, coordinated outcomes.

This is no distant prospect. AI has already broken free from the lab and embedded itself in daily life. Ready or not, Manus is proof that autonomy isn’t on the horizon — it has arrived. The question is no longer whether AI agents will shape the world, but in whose image — and to what end.

Take, for instance, the idea of not simply an AI lawyer, but an AI law firm — a network of specialized agents, each handling different aspects of a case. One sifts through contracts, another constructs legal arguments, a third methodically scans case law. Together, they form an integrated team operating at a speed and complexity beyond any human firm. This isn’t futuristic dreaming; it’s precisely what businesses are seeking to build today.

This revolution isn’t confined to the digital world. AI is now breaking out of the chatbox and into reality. Last month, Figure, an American robotics company developing humanoid workers — “cobots” or “collaborative robots” — unveiled Helix, a system designed not just to make robots move, but to think. Unlike Manus, which operates purely in software, Helix represents AI-driven autonomy in the physical world, allowing machines to perceive, plan, and manipulate their environment without human intervention.

All these developments are inevitably leading toward a future in which AI is an independent economic force. We’re approaching a world where AI agents — digital and physical — take on increasingly complex tasks, steadily reducing the need for human oversight.

This is a fundamental reordering of the economic landscape, and perhaps sooner than we’d like, we’ll need to ask ourselves just how comfortable we are with machines calling the shots.

Exactly how this economic reordering will reshape society is uncertain — but its implications are becoming starkly apparent. Could it finally deliver the long-promised four-day workweek? How comfortable are we with automated warfare? How will service-based economies cope when entire industries vanish overnight? The uncomfortable truth organizations are quietly acknowledging is simple: AI means fewer jobs, fewer workers, fewer people.

Why, then, are we allowing this to happen? Because alongside the unsettling realities, AI offers unprecedented opportunities — efficiencies, breakthroughs, transformative solutions — that are simply too promising to resist. Our institutions may creak under their own inertia, but AI holds the potential to reinvigorate society. A golden age of scientific discovery is inevitable, and with it, possibilities we’ve barely begun to imagine.

For now, such promises must be met with caution. Early users of Manus have reported absurdities, sluggish responses, and inconsistent outputs — hallucinations remain a feature, not a bug, of today’s AI systems. For the moment, at least, humans remain firmly in the loop, steering AI’s growing autonomy as we grapple with its imperfections.

We are witnessing advanced AI move decisively from the lab into our daily lives. Ready or not, beyond the hype, Manus is a clear sign that profound change is no longer coming — it’s already underway.