Manus’ emergence marks a shift in AI development from a technology race to practical application, standing out for its efficient integration and autonomous execution capabilities

By Tian Feng

When Chinese AI startup Butterfly Effect launched its autonomous agent Manus on March 6, it set off a frenzy in China’s tech circles. Social media buzzed with demos of Manus autonomously drafting reports, analyzing stocks, and even coding games while users slept. Invitation codes for its beta version sold for over $14,000 on secondary markets, and AI-related stocks surged. Yet outside China, the global reaction was muted, reflecting a growing East-West divide in AI development priorities.

A super product, not a super model

Unlike OpenAI’s trillion-parameter ambition or Google’s deep neural architectures, Manus follows a more pragmatic engineering approach. I categorize AI evolution into three tiers: Super Technology (foundational AI models like GPT-4), Super Systems (integrated platforms), and Super Products (user-facing applications). Manus falls into the last category — an applied solution that orchestrates existing AI tools into a highly functional service.

Instead of creating an entirely new model, Butterfly Effect built Manus by integrating multiple open-source and proprietary AI engines, including Claude 3.5 and DeepSeek. A three-agent framework (planning, execution, and verification) allows Manus to break down complex tasks, execute them within a virtual machine, and verify accuracy before presenting results. This approach prioritizes task completion over conversation, making Manus particularly suited for businesses where efficiency matters more than human-like interactions.

Manus claims 86.5% accuracy in the GAIA benchmark, outperforming OpenAI’s reference model at 74.3%. However, GAIA primarily tests structured tasks, such as resume screening, rather than open-ended reasoning. While its stock analysis took 40 minutes compared to the Bloomberg Terminal’s near-instant speed, the trade-off is autonomy: Manus users can offload multi-step workflows and wake up to a completed task, making it a digital assistant rather than an instant-response tool.

China’s regional innovation hubs

Manus’ rapid development showcases China’s strategy of leveraging regional innovation clusters. The Butterfly Effect team hails from Wuhan’s Optics Valley, a government-backed tech hub known for its university-driven R&D. Similar to Hefei’s success in incubating manufacturing giants like BOE and Nio, Wuhan benefits from state investment, subsidies, and industrial partnerships.

Meanwhile, Hangzhou has emerged as China’s private-sector AI and robotics hotspot, home to DeepSeek and Unitree. The inter-city competition between state-backed hubs like Wuhan and private-driven ecosystems like Hangzhou fosters an environment where rapid iteration thrives. The rise of Manus proves that China’s AI breakthroughs are no longer confined to Beijing and Shenzhen.

Global reactions: interest without alarm

Internationally, Manus has generated curiosity but little concern. On X, formerly Twitter, a demo video gained over 200,000 views in 24 hours, but U.S. tech media and regulators have largely ignored the launch so far. Unlike the intense scrutiny that met DeepSeek’s open-source release, Manus has avoided geopolitical backlash because it runs on China’s own cloud infrastructure and does not directly challenge U.S. AI model dominance.

This muted response perhaps reflects differing AI priorities. While U.S. labs chase general AI breakthroughs, China is commercializing task-specific AI tools for businesses and professionals. Western observers see Manus as a clever integration rather than a fundamental innovation, reinforcing the perception that China’s AI edge lies in practical applications rather than theoretical advancements.

Challenges and business strategy

Manus faces skepticism over technical originality, with critics dismissing it as a “glue factory” for third-party AI models. Yet, I argue that this is a feature, not a flaw. Just as Linux and Android grew from open-source foundations, Manus demonstrates how AI ecosystems evolve through modularity and integration rather than monolithic, proprietary models.

On the business side, Manus is challenging assumptions about AI monetization. Not long ago, consumer AI apps were considered unprofitable, yet Manus’ invite-only beta has created demand so intense that access codes resell for thousands of dollars. Early adopters in finance and HR report 15% to 25% efficiency gains, showing that there is a paying market for practical AI automation.

Manus’s biggest strategic decision now is whether to focus on broad consumer markets or enterprise solutions. While major financial institutions might be interested in custom versions, Manus risks getting bogged down in such development. Instead, Butterfly Effect seems intent on a scalable SaaS model, continuously refining Manus through mass-market feedback before engaging with enterprise clients.

AI as a tool, not a threat

Manus is not aiming to replace human workers but to augment them. I predict that AI assistants like Manus will become digital extensions of human expertise, handling routine tasks while professionals focus on higher-level strategy. For example, Manus can automate financial data collection and preliminary analysis, allowing researchers and investors to interpret insights rather than spend hours on manual data-gathering.

Manus also exemplifies China’s shift toward commercial AI solutions that generate immediate economic value. Unlike the Western focus on speculative artificial general intelligence (AGI) risks, China’s AI strategy prioritizes practicality, efficiency and monetization. By embedding observable workflows, Manus ensures users trust its decision-making process, making it a useful, scalable tool rather than a black-box algorithm.

Can manus stay independent?

As Manus grows, its relationship with China’s tech giants will be crucial. Historically, independent AI startups face a choice: align with a big platform like Tencent or Alibaba, or risk direct competition. Tencent’s Yuanbao platform, which integrates AI tools like DeepSeek, hints at how established players might absorb or compete with Manus.

Butterfly Effect reportedly has taken strategic investment from Tencent and other technology-focused VCs. Given the intensive attention it is attracting in China now, the company’s future independence and autonomy now hang in the balance. Until it surpasses 100 million users, Manus remains vulnerable to being acquired and either absorbed or sidelined. The alternative path is to scale independently, following the WeChat or TikTok model of becoming an ecosystem in its own right. If successful, Manus could evolve into a universal AI portal that competes with traditional app-based platforms.

The quiet revolution

Manus represents a broader paradigm shift in AI: from chasing intelligence to delivering usability. Instead of striving for AGI, it focuses on practical automation, a strategy that resonates with businesses and professionals. As Manus expands into global markets, it could pioneer a new era of task-specific AI assistants, integrated across industries.

For now, the AI race is not about who builds the biggest model, but who delivers the most useful tools. China’s AI sector is proving that incremental innovation can be just as transformative as foundational breakthroughs. The future of AI isn’t just about intelligence — it’s about impact, and Manus is setting a compelling precedent for how that impact can be realized.

Tian Feng is the founder of the Fast-Slow Think Research Institute, an independent AI think tank in China. You can reach him at:iamtianfeng@aliyun.com