GuShiio.com DeepSeek: Hello Crypto, big hammer 80 small hammer 40, what hammer do you want?
Authors: Bubble, BlockBeats
In January 2025, the launch of DeepSeek R1 caused a shock in the AI community, and it also really changed the Crypto AI ecosystem. In the past cycle, Crypto AI has mainly focused on AI Agents, and DeepSeek R1 and its open source strategy have completely changed the rules of the game: extremely low training costs and breakthrough adaptive training methods have made the vision of decentralizing the AI industry no longer Empty talk, but a reality at your fingertips. This change has far-reaching implications. The total market value of the Crypto AI market has shrunk significantly, and many AI tokens have experienced a 70% correction. But is this really a crisis? Does it mean a complete reshuffle for Crypto AI? Is DeepSeek the “terminator” that breaks the Crypto AI narrative, or is it the “breaker” that accelerates its entry into the era of practicality?
DeepSeek, the barbaric growth
DeepSeek’s development dates back to 2021. At that time, hedge fund Magic Square, which focused on quantitative trading, began to recruit AI talents on a large scale. Few quantitative companies turned into AI, and most of their recruitments explored cutting-edge directions, including Big Model (LLM) and Venson Model. AI researchers in fields such as AI researchers, although there are rumors that Magic Square was transforming to make better use of idle GPU resources in the company, most of the reasons should be decisions made to seize the commanding heights of cutting-edge AI technologies such as Big Models.
By the end of 2022, Magic Square has absorbed more and more top AI talents, mainly students from Tsinghua University and Peking University. Stimulated by ChatGPT, Magic Square CEO Liang Wenfeng was determined to enter the field of general artificial intelligence and established DeepSeek in early 2023. However, with the rapid rise of AI companies such as Intelligent Spectrum, Dark Side of the Moon, and Baichuan Intelligence, DeepSeek is facing great difficulties in independent financing as a pure research institution and lacks star founders. Therefore, Magic Square chose to divest DeepSeek and fully fund its development. Although this decision is extremely risky, DeepSeek does not need to be subject to profit commitments or valuation pressure from the financier. At the same time, it has relatively sufficient GPU resources, allowing the team to focus on technological breakthroughs. A group of innovative young people can run amok in a promised land. At this moment, DeepSeek is more like a research institute than a company.
Just like in the early days of OpenAI, no one would have thought of how a company that researched robot playing Rubik’s Cube would eventually develop ChatGPT, nor would anyone imagine how a quantitative company used DeepSeek to break through the current AI bubble. The former used it for 7 years, and the latter only used it for 2 years. In November 2023, DeepSeek LLM with 67 billion parameters and performance close to GPT-4 was launched. DeepSeek V2 was launched in May 2024. DeepSeek V3 released in December of the same year performed on par with GPT-4o and Claude 3.5 Sonnet in benchmark tests. DeepSeek’s rapid technological transition is not due to the company’s financial resources or high education, but because after a technological singularity occurred,”ChatGPT affected the world’s AI industry.” Singularities, large and small, accelerate in any place that can satisfy the imagination. In the soil until the next key singularity appears.
Finally, in January 2025, DeepSeek accelerated through the singularity, using the first-generation large model DeepSeek-R1 with reasoning capabilities they cultivated to open the door at a training cost and excellent performance much lower than ChatGPT-O1.
Use open source to distribute the key to the stargate to the world
Just one day after DeepSeek R1 released and announced the open source model, U.S. President Trump officially announced the start of a US$500 billion ultra-large-scale investment “Stargate” program at a White House press conference. A joint venture called Stargate was jointly formed by OpenAI, SoftBank, Oracle and investment company MGX to build a new artificial intelligence infrastructure for OpenAI in the United States.
This kind of investment is even comparable to the “Manhattan Project”. It requires the whole country to use algorithm stacking to push closed-source AI to a climax and monopolize the AI market to ensure the leading position of the U.S. AI industry. However, when the plan was released, it should not have been imagined that a few days later, this open source model on the other side of the ocean would no longer open the door. It not only brought a hammer at the door to smash the wall, but also handed out hammers to others.
DeepSeek is an open source model that can compete with top closed-source models. Its new training architecture has triggered a chain reaction, making closed-source AI difficult. Closed-source models that cannot beat DeepSeek R1 will be directly eliminated by the capital market. Even A16z “OpenAI Investor” Marc Andreessen, founder of the “OpenAI Investor”, has publicly stated that he needs to pay more attention to open source AI rather than focusing on closed source AI. In the industry, whether supporting AGI may produce or supporting AI can only be used as an upgraded version of the SaaS industry. Everyone believes that the harm of closed source is far greater than that of open source. Whether it is black boxes, industrial monopoly, information security, or capital control attention, any one of them is a very dangerous development direction.
Although some industry insiders believe that V3 ‘s hybrid expert technology “MoE” requires a huge data set, it is suspected of using OpenAI’s model for distillation. As well as the reinforcement learning-based method in the reinforcement learning “RL” for R1 requires a large amount of hardware resources, it is suspected that the number of training chips used has been false. However, it will not affect the industrial structural reform it brings.
DeepSeek R1 ‘s open source breaks OpenAI’s closed-source large model business logic in terms of training architecture, using the logic that allows the model to self-evolve to avoid the computing power and large investment in data annotation of traditional paradigms. Although the training model still opens blind boxes, the cost of blind boxes is much lower.
At the AI hardware level, DeepSeek’s V3 open source directly challenges Nvidia’s market dominance. Nvidia’s GPU moat is largely based on its bottom parallel computing platform and programming model CUDA. Its extensive ecosystem and enough developers make the learning cost of using non-Nvidia chips for training is too high, and high threshold purchase conditions and political restrictions have fragmented global AI development.
For us, in the short term, U.S. stock AI will shrink sharply, the total market value of Crypto AI will be almost cut, and the market will enter a bear market. But in the long run, the most recognized AI industry is moving towards an open source, transparent, and decentralized development path. No matter from which perspective, the combination of Crypto and AI will be more tacit.
During this round of Crypto AI bubble burst, many AI concept Tokens accepted a 70% correction, and the Crypto AI market shrank significantly. Some people joked that “a big model can be trained for US$5.5 million. The market value of these AIs exceeds, so why should we buy Crypto AI?” It is true that Crypto is a market dominated by funds, not products, and 90% of AI tokens have no practical significance.
But in fact, with the improvement of the encryption market regulatory system, the encryption market is still the most suitable soil for small and medium-sized AI companies to start businesses. The model cost brought by DeepSeek is 1/100 of that of ChatGPT O1 and the model training method will bring ecological growth more than 10,000 times compared to the current market.
Directly speaking, what DeepSeek brings to crypto is that a decentralized training model makes Depin-type projects more reasonable, makes the training process and information feeding more transparent, and the reward mechanism for contributors to the data set to obtain value is more reasonable, making settlement between the supply and demand sides of model training easier. The surrounding ecological development of the AI industry more than 10,000 times has further improved the industrial richness downstream of Crypto AI. When enough competitive and creative product narratives appear in the market, as long as one of them truly breaks the circle, external funds will naturally return value to Crypto. The market has been suffering from PVP for a long time. A series of celebrity coin harvests after TrumpCoin have broken the original balance of abundant liquidity and positive feedback in the AI market. Therefore, the bubble burst by DeepSeek is actually a greater benefit.
There are currently many Crypto AIs that have either quickly integrated DeepSeek or updated their architecture, including ElizaOS, Argo, Myshell, Build, Hyperbolic, Nillion Network, infraX, etc. Some of these projects have been optimized directly on the product side through DeepSeek.
Myshell
V3, R1 and even the image generation model Janus-Pro were added to the production flow of chat robots, application plug-ins. Myshell technicians completed the model integration in almost half a day. As a rare project in the blockchain, they have always insisted on polishing products, and even become famous in Web2AI products but have been reluctant to issue coins. This time DeepSeek’s open source will bring good news to Myshell users on the cost side. Lower costs will bring more Agent developers to Myshell, which has already been improved.
Argo
Argo developer Sam Gao has DeepSeek’s important functions in the early stage of product design. As a workflow system, Argo built LLM into the standard DeepSeek R1 and handed over the original workflow generation to DeepSeek R1. Also due to WorkFlow, the amount of Token consumption and context information will be very huge “average = 10k Token”, and Argo has also integrated CoT “Chain-of-Thought” into the WorkFlow thinking process. After DeepSeek’s open source, it not only reduces the cost of workflow products, but also allows LLM to be deployed locally in Argo, and users ‘privacy and security can also be guaranteed.
Whereas, Argo chose GoT, the only Crypto AI Workflow currently using this model, thereby achieving a more reliable and transparent process. This innovative approach directly affects the security and trust of transactions on the Argo platform. Integrating Mind Maps (GoT) into Web3 AI Agents puts Argo at the forefront of AI crypto transactions. CoT’s structured reasoning not only enhances the security of financial transactions, but also ensures transparent and reliable decision-making, which is crucial in decentralized finance (DeFi).
It is worth noting that Sam, the core developer of Argo, co-wrote a paper titled “EraseAnything: Enabling Concept Erasure in Rectified Flow Transformers” written by Shaw on how to remove unwanted concepts from large-scale text-to-image diffusion models without compromising the overall generation performance of the model. He received help from DeepSeek researcher Xingchao Liu.
Hyperbolic
Hyperbolic Labs is also the first to announce hosting DeepSeek-R1 models on its GPU platform. Users can rent Hyperblic GPU resources to run DeepSeek-R1 models locally or in designated data centers without having to send sensitive data to DeepSeek’s servers. This approach not only ensures data privacy, but also takes advantage of the excellent reasoning performance of the DeepSeek model. At the same time, through Hyperbolic’s decentralized computing network, users can obtain the efficient reasoning capabilities of the DeepSeek model at a lower cost. It will be a very competitive solution for startups or even pure efficient AI users.
This is not the end, but evolution. Crypto AI needs to move forward faster and more aggressively./ accelerated
