The Emergence of Manus: China’s Next-Generation AI Agent in the Post-DeepSeek Era

Introduction

The recent unveiling of Manus, a general-purpose AI agent developed by Chinese startup Monica, has reignited global discussions about China’s accelerating advancements in artificial intelligence.

Positioned as the world’s first fully autonomous AI agent capable of independent task execution, Manus represents both a technical leap and a strategic escalation in the geopolitical AI race.

Building on the momentum generated by DeepSeek’s R1 model earlier this year, Manus distinguishes itself through its ability to plan, reason, and execute multi-step workflows—from website creation to financial analysis—with minimal human intervention.

Early demonstrations have drawn comparisons to OpenAI’s DeepResearch, though skepticism persists regarding its operational transparency, data privacy practices, and the geopolitical implications of its success.

FAF examines Manus’s technical foundations, market positioning, and the broader ramifications for global AI development.

Development and Corporate Origins

Monica’s Evolution From Browser Plugin to AI Pioneer

Monica, the Wuhan-based startup behind Manus, began in 2022 as a niche developer of AI-powered browser extensions.

Initially offering a “ChatGPT for Google” plugin, the company secured early funding from ZhenFund and later attracted Series A investments from Tencent and Sequoia Capital China, reaching a $100 million valuation by late 2024.

This trajectory mirrors China’s broader pivot toward applied AI solutions, emphasizing practical integrations over theoretical research. Monica’s decision to reject a $30 million acquisition offer from ByteDance in early 2024 underscored its ambition to carve an independent path in the AI agent space.

Foundational Leadership and Strategic Vision

CEO Xiao Hong, a serial entrepreneur with a history of building WeChat management tools, recognized the transformative potential of large language models (LLMs) early. His collaboration with co-founder Ji Yichao—a prodigious developer who dropped out of high school to create the Mammoth Browser—combined pragmatic business acumen with technical ingenuity.

Xiao’s public statements reveal a deliberate strategy to leverage user data collected through Monica’s browser tools as a competitive moat, akin to TikTok’s growth playbook.

However, this data-centric approach has raised questions about compliance with China’s evolving AI regulations, particularly given Manus’s unclear model architecture and potential reliance on Western technologies like Anthropic’s Claude models.

Technical Capabilities and Benchmark Performance

Autonomous Task Execution and Workflow Automation

Manus is a cloud-based agent that interprets high-level user commands, devises step-by-step plans, and interacts with digital tools to deliver finalized outputs. Demo videos showcase its ability to:

Generate Functional Websites

Manus writes HTML/CSS code from a single prompt, sources images, and deploys the site through integrated platforms.

Conduct Financial Analysis

The agent scrapes real-time market data, performs technical evaluations, and compiles reports with visualizations.

Plan Complex Itineraries

Manus autonomously books hotels compare transportation options and balances budgetary constraints for a cherry blossom tour in Japan.

These use cases align with Monica’s claim that Manus outperforms OpenAI’s DeepResearch on the GAIA benchmark, a third-party AI-assistance evaluation framework.

While GAIA’s methodology remains undisclosed, the emphasis on multi-modal reasoning and tool integration suggests Manus excels in dynamic, real-world problem-solving.

Architectural Uncertainties and Model Transparency

Despite its capabilities, Manus’s technical underpinnings remain opaque. Internal system files retrieved during testing hint at reliance on Anthropic’s Claude models—a potential regulatory violation given China’s restrictions on foreign AI technologies.

Monica employs a hybrid approach, using specialized models for distinct tasks (e.g., coding vs. natural language processing), but refuses to disclose training data sources or computational requirements. This lack of transparency complicates efforts to assess Manus’s originality versus its adaptation of existing open-source frameworks.

Comparative Analysis with DeepSeek’s Legacy

From Model Efficiency to Agentic Autonomy

DeepSeek’s R1 model, launched in January 2025, prioritized cost efficiency and reasoning speed, leveraging quantization techniques to run on consumer-grade hardware. Its success stemmed from democratizing access to high-performance AI, albeit with security and political neutrality compromises.

In contrast, Manus shifts focus from model optimization to agentic autonomy—the ability to chain discrete actions into coherent workflows.

DeepSeek provided answers to questions, while Manus focused on solving problems. This represented a fundamental shift in the way humans and AI collaborate.

Market Reception and the “DeepSeek Moment” Narrative

Media outlets have dubbed Manus “China’s second DeepSeek moment,” but analysts debate the comparison. DeepSeek’s impact lay in proving Chinese startups could rival Silicon Valley on pure technical benchmarks, while Manus emphasizes application-layer innovation. An AI policy analyst, Dean Ball, argues that Manus transcends imitation: “DeepSeek replicated existing capabilities; Manus pushes boundaries.

The most advanced AI agent now comes from China”. However, critics note that both companies share a tendency toward hyperbolic marketing, with Manus’s invite-only launch and influencer-driven hype mirroring DeepSeek’s controversial rise.

Reception, Challenges, and Controversies

Viral Adoption and Compute Scalability Issues

Within 20 hours of its March 6 launch, Manus’s preview video amassed over 21 million views, crashing Monica’s servers due to unanticipated demand.

The company restricted access to invitation codes, which subsequently appeared on secondary markets priced at $10,000–$14,000—a strategy criticized as “hunger marketing” to manufacture exclusivity. Early users reported severe latency spikes, suggesting Monica underestimated the computational resources required for real-time agentic operations.

Operational Shortcomings and Ethical Concerns

Independent tests by TechCrunch and Corpora.ai revealed inconsistencies in Manus’s outputs, including factual inaccuracies in stock analyses and endless loop errors during website generation.

Privacy advocates highlighted risks associated with Monica’s data collection practices, particularly given its ties to Tencent and servers in an undisclosed location. European AI researcher Luiza Jarovsky questioned, “Does user data route through China? If so, Beijing’s intelligence agencies could gain unrestricted access.” These concerns echo the backlash against DeepSeek, which faced South Korea and Italy bans over data sovereignty issues.

Geopolitical and Market Implications

Accelerating the Sino-American AI Divide

Manus’s debut coincided with a 3% Nasdaq decline, reflecting investor anxiety that Chinese AI innovations could undermine U.S. tech dominance. Oppenheimer analysts linked the sell-off to Manus’s potential to “underpin China’s AI narrative” amidst recession fears and soaring GPU costs in Western markets. The agent’s ability to automate tasks like coding and data analysis at scale threatens to reshape global labor markets, particularly in software development and financial services.

Regulatory Crossroads and the Compute Dilemma

Monica’s ascent highlights China’s progress in circumventing U.S. semiconductor restrictions through software optimizations and hybrid cloud architectures.

However, co-founder Ji Yichao acknowledges persistent hurdles: “Our bottleneck isn’t funding—it’s the chip embargo.”

This admission underscores the precarious balance between China’s AI ambitions and dependence on foreign hardware.

Meanwhile, Western policymakers face mounting pressure to counter Chinese AI advancements without stifling domestic innovation through overregulation.

Conclusion: Toward a New Era of Agentic AI

Manus exemplifies China’s strategic pivot from foundational model development to applied AI agents that integrate seamlessly into economic and social systems.

While its technical achievements are undeniable—particularly in autonomous planning and cross-platform tool usage—the agent’s long-term viability hinges on resolving transparency deficits, scaling the computational infrastructure, and navigating an increasingly fragmented global regulatory landscape.

As DeepSeek races to launch its R2 model, the Sino-American AI rivalry is entering a new phase. Success will be measured not by benchmarks independently but by the ability to embed intelligence into the fabric of daily life.

For researchers and policymakers, Manus serves as both a precursor of AI’s potential and a cautionary tale about the ethical complexities of autonomous systems.