Manus AI: Difference between revisions
No edit summary |
No edit summary |
||
| Line 1: | Line 1: | ||
{{see also|artificial intelligence terms}} | |||
[[File:manus ai logo1.jpeg]] | [[File:manus ai logo1.jpeg]] | ||
[[Manus AI]] is an autonomous [[artificial intelligence agent]] that was publicly unveiled in March 2025. Named after the | [[Manus AI]] is an autonomous [[artificial intelligence agent]] that was publicly unveiled in March 2025. Named after the Latin word for "hand," Manus is designed to bridge the gap between human thought and action, independently executing complex tasks with minimal supervision. Developed by the Chinese startup [[Butterfly Effect]], Manus operates as a general [[AI agent]] that doesn't just generate ideas but delivers tangible results. The system attracted significant attention upon its release for its ability to autonomously navigate digital environments, complete multi-step workflows, and function asynchronously in the cloud without continuous human guidance. | ||
== History and Development == | ==History and Development== | ||
Manus AI was developed by a company called Butterfly Effect, which reportedly has offices in [[Beijing]] and [[Wuhan]], [[China]]. The company is led by [[Xiao Hong]], a 33-year-old entrepreneur and 2015 graduate of Wuhan's [[Huazhong University of Science and Technology]]. The project's research team is led by [[Ji Yichao]] (also known as "[[Peak]]" Ji), who serves as the co-founder and chief scientist. | |||
Prior to launching Manus, Xiao Hong developed WeChat-based applications that were later acquired and subsequently launched [[Monica.ai]], an AI assistant available as a browser extension and mobile app. While Butterfly Effect appears to be the primary company behind Manus, some reports have linked the product to a Singapore-registered entity of the same name, as well as to [[Monica.im]], which is developing autonomous agents. | |||
Prior to launching Manus, Xiao Hong developed WeChat-based applications that were later acquired and subsequently launched [[Monica.ai]], an AI assistant available as a browser extension and mobile app. While Butterfly Effect appears to be the primary company behind Manus, some reports have linked the product to a Singapore-registered entity of the same name, as well as to Monica.im, which is developing autonomous agents. | |||
Manus was released in limited preview form on March 6, 2025, with access restricted to invitation-only participants. The launch was marked by a demonstration video hosted by Ji Yichao that went viral, amassing over 200,000 views within 24 hours. Within days of its announcement, the official Discord server for Manus reportedly grew to over 138,000 members, and invitation codes were allegedly being resold for thousands of dollars on Chinese secondary markets such as [[Xianyu]] and international platforms like [[eBay]]. | Manus was released in limited preview form on March 6, 2025, with access restricted to invitation-only participants. The launch was marked by a demonstration video hosted by Ji Yichao that went viral, amassing over 200,000 views within 24 hours. Within days of its announcement, the official Discord server for Manus reportedly grew to over 138,000 members, and invitation codes were allegedly being resold for thousands of dollars on Chinese secondary markets such as [[Xianyu]] and international platforms like [[eBay]]. | ||
== Technology and Architecture == | ==Technology and Architecture== | ||
Unlike many advanced AI systems that rely on a single [[large language model]] ([[LLM]]), Manus operates as a multi-agent system composed of several specialized AI models working in concert. According to statements from Ji Yichao, Manus utilizes existing AI models including [[Claude 3.5 Sonnet]] from [[Anthropic]] and fine-tuned versions of [[Alibaba]]'s [[Qwen]] models rather than using a proprietary foundation model. The team has indicated they are testing integration with newer models such as [[Claude 3.7]] to enhance reasoning capabilities. | |||
Unlike many advanced AI systems that rely on a single [[large language model]] (LLM), Manus operates as a multi-agent system composed of several specialized AI models working in concert. According to statements from Ji Yichao, Manus utilizes existing AI models including [[Claude 3.5 Sonnet]] from [[Anthropic]] and fine-tuned versions of [[Alibaba]]'s [[Qwen]] models rather than using a proprietary foundation model. The team has indicated they are testing integration with newer models such as [[Claude 3.7]] to enhance reasoning capabilities. | |||
Manus employs a structured agent loop that follows these general steps: | Manus employs a structured agent loop that follows these general steps: | ||
| Line 23: | Line 22: | ||
# Entering standby until further input is required | # Entering standby until further input is required | ||
This architecture features a central | This architecture features a central [[executor agent]] that coordinates with specialized [[sub-agents]] ([[planners]], [[knowledge retrievers]], [[code generators]]) to break down complex tasks into manageable components. The system reportedly integrates with 29 tools and open-source software, enabling it to browse the web, interact with APIs, run scripts, and develop software independently. | ||
Key technical features of Manus include: | Key technical features of Manus include: | ||
| Line 32: | Line 31: | ||
* Deployment capabilities for publishing content to the web | * Deployment capabilities for publishing content to the web | ||
== Capabilities and Use Cases == | ==Capabilities and Use Cases== | ||
Manus AI is marketed as a [[general-purpose AI agent]] capable of handling diverse tasks across various domains. According to its website, Manus can perform functions including: | |||
Manus AI is marketed as a general-purpose AI agent capable of handling diverse tasks across various domains. According to its website, Manus can perform functions including: | |||
* '''Travel Planning''': Creating personalized itineraries and generating custom travel handbooks for destinations like Japan, with comprehensive information about accommodations, transportation, and cultural highlights. | * '''Travel Planning''': Creating personalized itineraries and generating custom travel handbooks for destinations like Japan, with comprehensive information about accommodations, transportation, and cultural highlights. | ||
| Line 45: | Line 43: | ||
* '''Data Analysis''': Processing and visualizing information from multiple sources to generate reports and insights. | * '''Data Analysis''': Processing and visualizing information from multiple sources to generate reports and insights. | ||
A distinguishing feature of Manus is its ability to function with minimal human guidance, planning and executing multi-step workflows autonomously. The system operates asynchronously in the cloud, allowing users to assign tasks and log off while Manus continues working, notifying them when complete. It can browse websites, leverage various software tools, and display its workflow in real-time through "replays" that demonstrate how it collects web data, generates documents, and refines its approach—functioning similarly to a human assistant completing digital tasks. | A distinguishing feature of Manus is its ability to function with minimal human guidance, planning and executing multi-step workflows autonomously. The system operates asynchronously in the cloud, allowing users to assign tasks and log off while Manus continues working, notifying them when complete. It can browse websites, leverage various software tools, and display its workflow in real-time through "[[replays]]" that demonstrate how it collects web data, generates documents, and refines its approach—functioning similarly to a human assistant completing digital tasks. | ||
==Performance and Benchmarks== | |||
According to information published on the Manus website, the system has achieved state-of-the-art performance on the [[GAIA benchmark]], a third-party evaluation framework for general AI assistants developed to test how well AI systems can solve real-world problems. | According to information published on the Manus website, the system has achieved state-of-the-art performance on the [[GAIA benchmark]], a third-party evaluation framework for general AI assistants developed to test how well AI systems can solve real-world problems. | ||
The benchmark results claimed by Manus showed performance across three difficulty levels: | The benchmark results claimed by Manus showed performance across three difficulty levels: | ||
* Level 1 (basic tasks): 86.5%, compared to OpenAI Deep Research at 74.3% and previous SOTA at 67.9% | * Level 1 (basic tasks): 86.5%, compared to [[OpenAI Deep Research]] at 74.3% and previous [[SOTA]] at 67.9% | ||
* Level 2 (intermediate tasks): 70.1%, compared to OpenAI Deep Research at 69.1% and previous SOTA at 67.4% | * Level 2 (intermediate tasks): 70.1%, compared to OpenAI Deep Research at 69.1% and previous SOTA at 67.4% | ||
* Level 3 (complex tasks): 57.7%, compared to OpenAI Deep Research at 47.6% and previous SOTA at 42.3% | * Level 3 (complex tasks): 57.7%, compared to OpenAI Deep Research at 47.6% and previous SOTA at 42.3% | ||
Manus was reportedly evaluated in standard mode using the same configuration as its production version for reproducibility. However, these benchmarks have not been independently verified by third parties, and real-world performance reports from users have varied considerably, with decreasing scores at higher difficulty levels indicating that even the best AI models still struggle with the most complex, multi-step reasoning tasks. | Manus was reportedly evaluated in standard mode using the same configuration as its production version for reproducibility. However, these benchmarks have not been independently verified by third parties, and real-world performance reports from users have varied considerably, with decreasing scores at higher difficulty levels indicating that even the best [[AI models]] still struggle with the most complex, multi-step reasoning tasks. | ||
==Reception and Controversy== | |||
The launch of Manus generated significant discussion within the AI community, with mixed reactions from experts and early users. Some AI researchers and industry figures described Manus as groundbreaking: | The launch of Manus generated significant discussion within the AI community, with mixed reactions from experts and early users. Some AI researchers and industry figures described Manus as groundbreaking: | ||
* '''Victor Mustar''', head of product at [[Hugging Face]], called it "the most impressive AI tool I've ever tried," adding that its agentic capabilities "redefine what's possible." | * '''Victor Mustar''', head of product at [[Hugging Face]], called it "the most impressive [[AI tool]] I've ever tried," adding that its agentic capabilities "redefine what's possible." | ||
* '''Dean Ball''', an AI policy researcher, argued that it was "wrong to call Manus a 'DeepSeek moment'" because it goes further, describing it as "the most sophisticated computer using AI" that is "actually advancing the frontier." | * '''Dean Ball''', an AI policy researcher, argued that it was "wrong to call Manus a '[[DeepSeek]] moment'" because it goes further, describing it as "the most sophisticated computer using AI" that is "actually advancing the frontier." | ||
* '''Andrew Wilkinson''', co-founder of [[Tiny]], stated it was "absolutely insane" and felt like "time travelling six months into the future" after testing its ability to analyze job applicants. | * '''Andrew Wilkinson''', co-founder of [[Tiny]], stated it was "absolutely insane" and felt like "time travelling six months into the future" after testing its ability to analyze job applicants. | ||
| Line 69: | Line 65: | ||
* '''Kyle Wiggers''' (TechCrunch) and '''Alexander Doria''' (AI researcher) reported experiencing errors, execution failures, and endless loops while testing the system, particularly with tasks like ordering food or booking flights. | * '''Kyle Wiggers''' (TechCrunch) and '''Alexander Doria''' (AI researcher) reported experiencing errors, execution failures, and endless loops while testing the system, particularly with tasks like ordering food or booking flights. | ||
* '''Pierre-Carl Langlais''' (co-founder of AI startup Pleias) criticized its "deceptive communication" and what he described as "hunger marketing" tactics—creating artificial scarcity through limited invitations to generate hype. | * '''Pierre-Carl Langlais''' (co-founder of AI startup [[Pleias]]) criticized its "deceptive communication" and what he described as "hunger marketing" tactics—creating artificial scarcity through limited invitations to generate hype. | ||
Several points of controversy emerged following Manus's launch: | Several points of controversy emerged following Manus's launch: | ||
=== Technology Claims === | ===Technology Claims=== | ||
Some AI researchers questioned whether Manus represented a genuine breakthrough, noting its reliance on existing models like Claude 3.5 Sonnet rather than novel proprietary technology. Critics argued that Manus primarily orchestrated existing AI capabilities rather than advancing fundamental AI research. Unlike [[DeepSeek]], which developed its own foundation model, Manus leverages and fine-tunes pre-existing models, leading to debates about its originality. | Some AI researchers questioned whether Manus represented a genuine breakthrough, noting its reliance on existing models like Claude 3.5 Sonnet rather than novel proprietary technology. Critics argued that Manus primarily orchestrated existing AI capabilities rather than advancing fundamental AI research. Unlike [[DeepSeek]], which developed its own foundation model, Manus leverages and fine-tunes pre-existing models, leading to debates about its originality. | ||
=== Performance Consistency === | ===Performance Consistency=== | ||
Early users reported inconsistent performance, with some tasks completed impressively while others resulted in errors or repetitive loops. This raised questions about the system's reliability for mission-critical applications. Some testers found Manus excelled at structured tasks like resume screening but struggled with more ambiguous requests. | Early users reported inconsistent performance, with some tasks completed impressively while others resulted in errors or repetitive loops. This raised questions about the system's reliability for mission-critical applications. Some testers found Manus excelled at structured tasks like resume screening but struggled with more ambiguous requests. | ||
=== Privacy and Security === | ===Privacy and Security=== | ||
Researchers including '''Luiza Jarovsky''' raised concerns about Manus's data handling practices, questioning where user data was stored and whether Chinese authorities might have access to information processed by the system. The sandbox execution environment and isolated sessions were designed to prevent unauthorized access, but the system's reliance on external models raised questions about data privacy and scalability. | Researchers including '''Luiza Jarovsky''' raised concerns about Manus's data handling practices, questioning where user data was stored and whether Chinese authorities might have access to information processed by the system. The sandbox execution environment and isolated sessions were designed to prevent unauthorized access, but the system's reliance on external models raised questions about data privacy and scalability. | ||
=== Comparison to DeepSeek === | ===Comparison to DeepSeek=== | ||
Manus's launch prompted comparisons to [[DeepSeek]], another Chinese AI development that generated significant attention in early 2025. While some observers like '''Rowan Cheung''' of The Rundown AI newsletter described Manus as "China's second DeepSeek moment," others, including AI researcher Dean Ball, argued that Manus represented a more significant advancement, potentially pushing the technological frontier rather than simply replicating capabilities already demonstrated by American companies. | Manus's launch prompted comparisons to [[DeepSeek]], another Chinese AI development that generated significant attention in early 2025. While some observers like '''Rowan Cheung''' of The Rundown AI newsletter described Manus as "China's second DeepSeek moment," others, including AI researcher Dean Ball, argued that Manus represented a more significant advancement, potentially pushing the technological frontier rather than simply replicating capabilities already demonstrated by American companies. | ||
== Market Position and Future Development == | ==Market Position and Future Development== | ||
As of March 2025, Manus remained in closed beta with limited access. The company indicated plans to expand availability but cited server capacity limitations as a constraint. According to Zhang Tao, Manus AI's product partner, "The current invite-only mechanism is due to genuinely limited server capacity at this stage... The team underestimated the enthusiasm of the public response, and our server resources were only planned for a demonstration level." | As of March 2025, Manus remained in closed beta with limited access. The company indicated plans to expand availability but cited server capacity limitations as a constraint. According to Zhang Tao, Manus AI's product partner, "The current invite-only mechanism is due to genuinely limited server capacity at this stage... The team underestimated the enthusiasm of the public response, and our server resources were only planned for a demonstration level." | ||
| Line 109: | Line 100: | ||
== References == | == References == | ||
# Manus AI Official Website (2025). ''Leave it to Manus''. Retrieved from manus.im. | # Manus AI Official Website (2025). ''Leave it to Manus''. Retrieved from manus.im. | ||
# Burman, T. (2025, March 6). ''What is Manus? China's World-First Fully Autonomous AI Agent Explained''. Newsweek. | # Burman, T. (2025, March 6). ''What is Manus? China's World-First Fully Autonomous AI Agent Explained''. Newsweek. | ||
| Line 118: | Line 108: | ||
# Smith, C. S. (2025, March 8). ''China's Autonomous Agent, Manus, Changes Everything''. Forbes. | # Smith, C. S. (2025, March 8). ''China's Autonomous Agent, Manus, Changes Everything''. Forbes. | ||
# LLMhacker. (2025, March 6). ''Manus AI: The Best Autonomous AI Agent Redefining Automation and Productivity''. Hugging Face. | # LLMhacker. (2025, March 6). ''Manus AI: The Best Autonomous AI Agent Redefining Automation and Productivity''. Hugging Face. | ||
[[Category:Terms]] [[Category:Artificial intelligence terms]] | |||
Revision as of 21:06, 11 March 2025
- See also: artificial intelligence terms
Manus AI is an autonomous artificial intelligence agent that was publicly unveiled in March 2025. Named after the Latin word for "hand," Manus is designed to bridge the gap between human thought and action, independently executing complex tasks with minimal supervision. Developed by the Chinese startup Butterfly Effect, Manus operates as a general AI agent that doesn't just generate ideas but delivers tangible results. The system attracted significant attention upon its release for its ability to autonomously navigate digital environments, complete multi-step workflows, and function asynchronously in the cloud without continuous human guidance.
History and Development
Manus AI was developed by a company called Butterfly Effect, which reportedly has offices in Beijing and Wuhan, China. The company is led by Xiao Hong, a 33-year-old entrepreneur and 2015 graduate of Wuhan's Huazhong University of Science and Technology. The project's research team is led by Ji Yichao (also known as "Peak" Ji), who serves as the co-founder and chief scientist.
Prior to launching Manus, Xiao Hong developed WeChat-based applications that were later acquired and subsequently launched Monica.ai, an AI assistant available as a browser extension and mobile app. While Butterfly Effect appears to be the primary company behind Manus, some reports have linked the product to a Singapore-registered entity of the same name, as well as to Monica.im, which is developing autonomous agents.
Manus was released in limited preview form on March 6, 2025, with access restricted to invitation-only participants. The launch was marked by a demonstration video hosted by Ji Yichao that went viral, amassing over 200,000 views within 24 hours. Within days of its announcement, the official Discord server for Manus reportedly grew to over 138,000 members, and invitation codes were allegedly being resold for thousands of dollars on Chinese secondary markets such as Xianyu and international platforms like eBay.
Technology and Architecture
Unlike many advanced AI systems that rely on a single large language model (LLM), Manus operates as a multi-agent system composed of several specialized AI models working in concert. According to statements from Ji Yichao, Manus utilizes existing AI models including Claude 3.5 Sonnet from Anthropic and fine-tuned versions of Alibaba's Qwen models rather than using a proprietary foundation model. The team has indicated they are testing integration with newer models such as Claude 3.7 to enhance reasoning capabilities.
Manus employs a structured agent loop that follows these general steps:
- Analyzing events and understanding user requests
- Selecting appropriate tools or APIs for specific tasks
- Executing commands in a Linux sandbox environment
- Iterating and refining actions based on new data
- Submitting structured results to users
- Entering standby until further input is required
This architecture features a central executor agent that coordinates with specialized sub-agents (planners, knowledge retrievers, code generators) to break down complex tasks into manageable components. The system reportedly integrates with 29 tools and open-source software, enabling it to browse the web, interact with APIs, run scripts, and develop software independently.
Key technical features of Manus include:
- A Linux sandbox execution environment
- Shell and command-line execution capabilities
- Integrated web browser control for navigating websites and extracting data
- File system management for organizing documents and data
- Deployment capabilities for publishing content to the web
Capabilities and Use Cases
Manus AI is marketed as a general-purpose AI agent capable of handling diverse tasks across various domains. According to its website, Manus can perform functions including:
- Travel Planning: Creating personalized itineraries and generating custom travel handbooks for destinations like Japan, with comprehensive information about accommodations, transportation, and cultural highlights.
- Financial Analysis: Delivering in-depth analyses of stock performance, such as detailed evaluations of Tesla stocks with interactive visual dashboards that showcase market trends and financial outlooks.
- Educational Content: Developing engaging step-by-step educational courses and presentations, such as interactive lessons on the momentum theorem for middle school students.
- Document Comparison: Analyzing complex documents like insurance policies to generate structured comparison tables highlighting key differences.
- B2B Research: Conducting supplier sourcing and industry research to identify optimal business partners.
- E-commerce Analysis: Processing Amazon store sales data to deliver actionable insights and performance optimization strategies.
- Web Development: Building and deploying interactive websites from scratch, with real-time demonstrations of the creation process.
- Data Analysis: Processing and visualizing information from multiple sources to generate reports and insights.
A distinguishing feature of Manus is its ability to function with minimal human guidance, planning and executing multi-step workflows autonomously. The system operates asynchronously in the cloud, allowing users to assign tasks and log off while Manus continues working, notifying them when complete. It can browse websites, leverage various software tools, and display its workflow in real-time through "replays" that demonstrate how it collects web data, generates documents, and refines its approach—functioning similarly to a human assistant completing digital tasks.
Performance and Benchmarks
According to information published on the Manus website, the system has achieved state-of-the-art performance on the GAIA benchmark, a third-party evaluation framework for general AI assistants developed to test how well AI systems can solve real-world problems.
The benchmark results claimed by Manus showed performance across three difficulty levels:
- Level 1 (basic tasks): 86.5%, compared to OpenAI Deep Research at 74.3% and previous SOTA at 67.9%
- Level 2 (intermediate tasks): 70.1%, compared to OpenAI Deep Research at 69.1% and previous SOTA at 67.4%
- Level 3 (complex tasks): 57.7%, compared to OpenAI Deep Research at 47.6% and previous SOTA at 42.3%
Manus was reportedly evaluated in standard mode using the same configuration as its production version for reproducibility. However, these benchmarks have not been independently verified by third parties, and real-world performance reports from users have varied considerably, with decreasing scores at higher difficulty levels indicating that even the best AI models still struggle with the most complex, multi-step reasoning tasks.
Reception and Controversy
The launch of Manus generated significant discussion within the AI community, with mixed reactions from experts and early users. Some AI researchers and industry figures described Manus as groundbreaking:
- Victor Mustar, head of product at Hugging Face, called it "the most impressive AI tool I've ever tried," adding that its agentic capabilities "redefine what's possible."
- Dean Ball, an AI policy researcher, argued that it was "wrong to call Manus a 'DeepSeek moment'" because it goes further, describing it as "the most sophisticated computer using AI" that is "actually advancing the frontier."
- Andrew Wilkinson, co-founder of Tiny, stated it was "absolutely insane" and felt like "time travelling six months into the future" after testing its ability to analyze job applicants.
However, other experts expressed skepticism about Manus's capabilities:
- Kyle Wiggers (TechCrunch) and Alexander Doria (AI researcher) reported experiencing errors, execution failures, and endless loops while testing the system, particularly with tasks like ordering food or booking flights.
- Pierre-Carl Langlais (co-founder of AI startup Pleias) criticized its "deceptive communication" and what he described as "hunger marketing" tactics—creating artificial scarcity through limited invitations to generate hype.
Several points of controversy emerged following Manus's launch:
Technology Claims
Some AI researchers questioned whether Manus represented a genuine breakthrough, noting its reliance on existing models like Claude 3.5 Sonnet rather than novel proprietary technology. Critics argued that Manus primarily orchestrated existing AI capabilities rather than advancing fundamental AI research. Unlike DeepSeek, which developed its own foundation model, Manus leverages and fine-tunes pre-existing models, leading to debates about its originality.
Performance Consistency
Early users reported inconsistent performance, with some tasks completed impressively while others resulted in errors or repetitive loops. This raised questions about the system's reliability for mission-critical applications. Some testers found Manus excelled at structured tasks like resume screening but struggled with more ambiguous requests.
Privacy and Security
Researchers including Luiza Jarovsky raised concerns about Manus's data handling practices, questioning where user data was stored and whether Chinese authorities might have access to information processed by the system. The sandbox execution environment and isolated sessions were designed to prevent unauthorized access, but the system's reliance on external models raised questions about data privacy and scalability.
Comparison to DeepSeek
Manus's launch prompted comparisons to DeepSeek, another Chinese AI development that generated significant attention in early 2025. While some observers like Rowan Cheung of The Rundown AI newsletter described Manus as "China's second DeepSeek moment," others, including AI researcher Dean Ball, argued that Manus represented a more significant advancement, potentially pushing the technological frontier rather than simply replicating capabilities already demonstrated by American companies.
Market Position and Future Development
As of March 2025, Manus remained in closed beta with limited access. The company indicated plans to expand availability but cited server capacity limitations as a constraint. According to Zhang Tao, Manus AI's product partner, "The current invite-only mechanism is due to genuinely limited server capacity at this stage... The team underestimated the enthusiasm of the public response, and our server resources were only planned for a demonstration level."
According to statements from Ji Yichao, Butterfly Effect plans to open-source some of the models behind Manus later in 2025, potentially allowing developers to integrate parts of its AI system into their own applications.
Manus entered a competitive landscape of AI agents, with comparable products including OpenAI's Deep Research and Operator, Google's AI assistants, and Anthropic's Claude Computer Use mode. Unlike some competitors, Manus emphasized its ability to work asynchronously and execute complete workflows autonomously.
The launch of Manus happened amid rumors of OpenAI planning to launch advanced AI agents with subscription fees ranging from $2,000 to $20,000, suggesting it could disrupt pricing models in the AI industry if it delivers on its promise of low-cost, open access to autonomous AI.
The long-term business model for Manus remains unclear, as does its potential impact on the broader AI industry and employment landscape. If it can overcome current technical limitations and scale effectively, it could represent a significant advancement in agentic AI with implications for workflow automation, AI competition, labor markets, and AI development costs.
See Also
References
- Manus AI Official Website (2025). Leave it to Manus. Retrieved from manus.im.
- Burman, T. (2025, March 6). What is Manus? China's World-First Fully Autonomous AI Agent Explained. Newsweek.
- Wiggers, K. (2025, March 9). Manus probably isn't China's second 'DeepSeek moment'. TechCrunch.
- Webb, E. (2025, March 11). China's newest AI model Manus is dividing opinion over DeepSeek comparisons. Business Insider.
- Franzen, C. (2025, March 10). What you need to know about Manus, the new AI agentic system from China hailed as a second 'DeepSeek moment'. VentureBeat.
- Nguyen, B. (2025, March 11). China has a new 'completely autonomous' AI agent. Here's what to know about Manus. Quartz.
- Smith, C. S. (2025, March 8). China's Autonomous Agent, Manus, Changes Everything. Forbes.
- LLMhacker. (2025, March 6). Manus AI: The Best Autonomous AI Agent Redefining Automation and Productivity. Hugging Face.