While DeepSeek can now be considered an established player in the sector, Manus AI presents an ambitious vision in the initial development phase. Like any emerging technology, it deserves both enthusiasm and balanced evaluation.
Between Expectations and Reality
According to TechCrunch, Manus has generated a wave of enthusiasm comparable to “a Taylor Swift concert”, but like any beta closed, it is facing the inevitable initial challenges:
Integrative Approach: Rather than developing models from scratch, Manus seems to use a combination of existing technologies such as Claude by Anthropic and Qwen by Alibaba - an effective strategy for innovation at the application level
Ambitious Vision: The platform promises extraordinary capabilities, from market research to video game programming. While some of these features are still being refined, the direction is fascinating nonetheless
Variable Experiences: Some early users have reported difficulties with specific tasks, but as the developers themselves point out, “the main goal of the closed beta is to test the system and identify problems” - a responsible approach for a nascent technology
Enthusiasm as a Driving Force
The Manus phenomenon demonstrates how enthusiasm can catalyze attention to new ideas:
In a few days, the Discord server reached 138,000 members
Invitation codes became extremely sought after, creating an aura of exclusivity
Media and influencers helped spread the vision of what Manus aspires to become
This enthusiasm, if properly exploited, can accelerate development and attract talent and resources. In a viral video cited by TechCrunch, Yichao “Peak” Ji, Manus' research lead, described the platform as “a fully autonomous agent that bridges the gap between conception and execution” and “the next paradigm in human-machine collaboration” - a vision that could inspire significant progress.
A Future To Be Written
It is still too early to definitively evaluate Manus. Like many revolutionary technologies, the beta version may only be a hint of what it will become. The developers themselves have stated: “As a team, our focus is to continue improving Manus and creating AI agents that really help users solve problems.”
Unlike DeepSeek, which focused on the efficiency of basic models, Manus is exploring new paradigms of interaction. Both approaches are valid and potentially complementary in the innovation landscape
Beyond the Technique: The Disturbing Implications of Globalized AI
The debate on DeepSeek and Manus AI hides far more controversial issues that are not always openly discussed:
Digital Colonialism 2.0
For decades, the West has dictated the rules of technological innovation. The emergence of Chinese AI powers represents a threat to this dominance, triggering reactions reminiscent of colonial dynamics:
The United States imposes restrictions on the export of advanced chips to China, a modern form of “trade blockade”
Western companies label Chinese technologies as “copies” or “theft”, while celebrating their own as “innovations”, even when both are based on common academic research
Western media present Chinese AI as inherently dangerous, subtly suggesting that only the West can develop “ethical” technology
This technological clash is framed within a broader context of commercial tensions, as demonstrated by the recent escalation of the tariff war between the United States and China. With the imposition of American tariffs of up to 125% on Chinese imports and Beijing's response with 84% tariffs, we are witnessing an economic battle that goes well beyond traditional trade.
It is significant that, while Trump announced a 90-day pause on many global tariffs, China was explicitly excluded from this truce, confirming how the AI confrontation is part of a broader strategy to contain Chinese technological and economic influence, reducing the dependence of global supply chains on Beijing. In this scenario, Taiwan emerges as a crucial actor, being home to TSMC, the world's leading manufacturer of advanced semiconductors.
It is no coincidence that, following the announcement of the pause on tariffs, the Taiwanese stock market recorded a jump of 9.2%, highlighting how the technological and commercial tensions between the USA and China have direct repercussions on the island, whose strategic position in the chip supply chain makes it a fundamental element in the confrontation for technological supremacy.
Censorship in Chinese models is often criticized, but the fact that Western models implement equally pervasive forms of censorship is kept quiet:
American models have “guardrails” that prevent discussion of politically sensitive topics, under the pretext of “safety”
Western companies unilaterally decide which values to incorporate into their models, imposing a corporate morality disguised as universalism
“Western censorship” is called “value alignment”, Chinese “authoritarian control” - two different names for similar phenomena
The nature of competition
The competition between Western and Chinese models hides an inconvenient truth:
Both draw on the same body of human knowledge, much of it extracted without compensation
Most advances are incremental optimizations that, by methodically perfecting what already exists, embody the very essence of innovation
The significant difference in market valuation between Western and Eastern AI companies reflects both financial and investment dynamics as well as actual competitive technological advantages
The Limits of “Democratization”
The idea that models such as DeepSeek “democratize” AI deserves critical analysis:
Cheaper access to AI is a step forward, but it is not the same as real democratic power over the technology
Dependence on centralized cloud infrastructures remains, shifting some control from model creators to infrastructure providers
The skills needed to modify or adapt these models remain inaccessible to the majority
While we discuss costs and efficiency, we ignore that:
The data used to train these models contains personal information on billions of people
“Distillation” could perpetuate and amplify this violation of privacy
The New Wave of Innovation from the East
Chinese startups are revolutionizing the global artificial intelligence sector. DeepSeek and Manus AI represent two different but equally innovative approaches that are changing the rules of the game. Let's take a look at what makes these companies so special and why you should pay attention to their development.
DeepSeek: Low-Cost Artificial Intelligence
DeepSeek, founded in July 2023 by Liang Wenfeng, has disrupted the market with a revolutionary approach: creating powerful AI models while spending much less than their competitors.
How did they do it?
A team of about 200 employees with experience in quantitative finance
Optimization of algorithms instead of focusing only on larger models
Open source licenses that encourage knowledge sharing
The result? Their DeepSeek-R1 model, with 671 billion parameters, cost only 6 million dollars (nb. we can't really verify this figure), compared to the 100 million spent by OpenAI for GPT-4!
The impact on the market
When DeepSeek released its model at the beginning of 2025, Nvidia's shares plummeted by 13%, with a loss of 600 billion dollars in market capitalization. Why? DeepSeek has shown that it is possible to create advanced AI without being completely dependent on expensive Nvidia chips.
Their service costs just $2.19 per million tokens (about 30 times less than OpenAI), making advanced AI accessible to many more users.
Model Distillation: The Secret to Success
The Teacher-Student Relationship: When AI Mimics Human Education
One of the key concepts behind these advances is “model distillation”, based on the teacher-student paradigm. This process mirrors surprisingly what happens in classrooms, but with much more controversial implications.
How Knowledge Transfer Works
In human education, a university professor (expert in the subject) teaches less experienced students. The teacher doesn't transmit the entire content of the textbooks, but “distills” decades of study into simplified lessons, highlighting the key concepts and leaving out less relevant details.
Something similar happens in AI:
A larger model (the teacher) is trained on huge amounts of data
This “digital teacher” produces answers that the smaller model (the student) tries to imitate
The student does not learn from the original data, but from the teacher's simplified “lessons”
The Shadow Zones of Distillation
This raises profound ethical questions that are rarely discussed in the sector:
Institutionalized Plagiarism? What happens if the teacher model is proprietary (like GPT-4) and the student model is open source? Isn't this a form of intellectual appropriation masked as innovation?
The Paradox of Inherited Ignorance: Human students can verify and challenge the teacher. The “student” models, on the other hand, blindly inherit the prejudices and errors of the “teacher” model, amplifying them without the possibility of critical correction.
Creating “Cheap Clones”: Western companies invest billions in research, while other organizations could simply “distill” the final result at a fraction of the cost. Is it fair to allow this kind of “technological parasitism”?
Originality vs. Imitation: A distilled model can never surpass its teacher in creativity, only imitate it more effectively. So are we creating an AI that by definition will always be derivative?
The Irony of the Debate: Who is Really “Original”?
However, there is a fundamental irony in these criticisms that is rarely mentioned: even the “original” Western models were trained on content that they did not create themselves. GPT-4, Claude and other AI giants “learned” by analyzing:
Books published by human authors
Scientific articles written by researchers
Open source code developed by programmers
Web content created by millions of people
In this sense, Western companies that accuse distillation of “appropriation” are ignoring that their own models are based on a massive appropriation of human content, often without compensation or explicit consent. Who can really claim “ownership” of knowledge in this ecosystem?
DeepSeek and similar services could therefore be seen not as “pirates” of intellectual property, but as democratizers of knowledge that has already been extracted from collective human creativity. Distillation thus becomes just another level of an appropriation process that is already underway.
DeepSeek has perfected this controversial technique, obtaining performance similar to that of the American giants at a fraction of the cost.
Manus AI: The Assistant that Works Independently
If DeepSeek focuses on efficiency, Manus AI (launched in March 2025) focuses on autonomy. Unlike traditional chatbots that answer questions step by step, Manus claims to be able to:
Perform complex tasks independently
Operate with a single initial command
Continue working even when the user is not connected
According to co-founder Red Xiao Hong, Manus is similar to “a human being” because of its ability to interact with the environment and adapt.
How does it work?
Instead of developing a new base model, Manus integrates existing models such as Claude by Anthropic, focusing on innovation at the application level.
Two Different Paths for the Future of AI
These two approaches represent different visions of the evolution of AI:
DeepSeek: vertical optimization, improving the efficiency of the basic model
Manus AI: horizontal expansion, aiming at task automation
Performance comparison
DeepSeek-R1 is ranked 4th on Chatbot Arena, approaching the performance of GPT-4o but at a much lower cost. Manus AI, on the other hand, does not yet have public benchmarks and some experts have raised doubts about the capabilities shown in the demonstration videos.
Challenges and Doubts
Despite the enthusiasm, some concerns have emerged:
Transparency: How did DeepSeek achieve such impressive results with limited resources? Who knows...
Intellectual property: There is suspicion that these companies may have used data or techniques that are proprietary to other companies
Accuracy of claims: Some of Manus' demonstrations seem exaggerated compared to his real capabilities
The Future is Hybrid
The next phase of AI will probably combine both approaches: the efficiency of DeepSeek with the versatility of Manus. However, it will be essential to address the legal, ethical and technical issues that emerge from these developments.
For you, the user, this all means access to AI tools that are more powerful, cheaper and useful in everyday life. Chinese innovation is accelerating the development of AI globally, bringing benefits but also new challenges to face together.
Did you like this newsletter? Share it with anyone who might be interested and subscribe to receive future updates on the world of artificial intelligence!
Welcome to Electe’s Newsletter - English
This newsletter explores the fascinating world of how companies are using AI to change the way they work. It shares interesting stories and discoveries about artificial intelligence in business - like how companies are using AI to make smarter decisions, what new AI tools are emerging, and how these changes affect our everyday lives.
You don't need to be a tech expert to enjoy it - it's written for anyone curious about how AI is shaping the future of business and work. Whether you're interested in learning about the latest AI breakthroughs, understanding how companies are becoming more innovative, or just want to stay informed about tech trends, this newsletter breaks it all down in an engaging, easy-to-understand way.
It's like having a friendly guide who keeps you in the loop about the most interesting developments in business technology, without getting too technical or complicated
Subscribe to get full access to the newsletter and publication archives.
