External Innovation vs. Internal Dilemma: Crypto Sailing Through the Fog

Original Title: External Innovations and Internal Dilemmas: Crypto Navigating Through Uncertainty

Original Author: YBB Capital Researcher Zeke

Chapter 1: AI is Inevitable, but Crypto x AI?

At the beginning of 2025, the DeepSeek created by Magic Square Quantification dropped a "nuclear bomb" on the AI community. A Chinese AI model, using only 2048 NVIDIA H800 GPUs with a training cost of $5.58 million (about one-tenth of Meta's), directly outperformed GPT-4o and Llama 3.1 in benchmark tests such as MMLU and GPQA, even slightly surpassing these top-tier Silicon Valley models in areas such as complex reasoning and Chinese semantic understanding. Despite the U.S.'s years-long chip blockade against China, the American fortress faced a dramatic deconstruction in the horsepower battlefield of DeepSeek. Under siege, the Chinese-style AI finally developed a technology path that was in line with the national conditions yet rivalled cutting-edge capabilities. An open-source, cost-effective, homogenous combination swiftly breached America's computational moat.

I have always had a stereotypical impression that Chinese tech products, which perennially lag behind the U.S. in performance, are cheap and imitative. I believe this is also the intuitive perception of most people regarding Chinese internet products. However, this time, DeepSeek is indeed exceptional. Let's put aside, for now, whether the user experience is superior to subjective sensations like ChatGPT. Merely from the tense reactions in American politics and from the frequent mentions by tech giants, it is apparent that China is no longer playing the role of a mere technology follower this time. The global shockwave triggered by this event is also very strong.

Although my wallet was the first to be affected, to a certain extent, this was indeed a misjudgment of traditional AI development. Nevertheless, I would like to share my thoughts, particularly on DeepSeek's impact on Crypto.

1. NVIDIA suffered the most in this event. Firstly, its AI computing power demand was questioned, and secondly, NVIDIA's software-hardware unified computing architecture "CUDA" was bypassed. Friends familiar with the AI field should be no strangers to CUDA, which is one of the key cornerstones driving modern AI development. When developers of large models use NVIDIA GPUs, they generally do so based on CUDA. The use of CUDA lowers the bar for developers because some functions are already encapsulated in CUDA, obviating the need to worry about many details during use, although it certainly sacrifices execution efficiency.

Since CUDA is a general-purpose programming framework, it can lead to some loss of flexibility when training models. DeepSeek's approach is to directly use PTX (NVIDIA's intermediate instruction set architecture designed for the GPU) to bypass hardware limitations on training speed, which can shorten the training time. While other model training may take 10 days, DeepSeek can complete it in 5 days. This also means that if DeepSeek intends to adapt to domestically produced GPUs in the future, it will be more adept at hardware adaptation, potentially shaking NVIDIA's AI chip dominance. (This paragraph is taken from Future Asset Securities' analysis of DeepSeek's training process)

In addition to the impact of a stock price drop on the cryptocurrency market strongly associated with the U.S. stock market, I personally believe that, in the long run, it is actually a good thing for decentralized computing power projects. First, more individual GPUs will be able to contribute to the ecosystem in the future. Second, if DeepSeek's small yet powerful open-source model approach proves successful, it will force many AI companies to open-source their technologies. The demand for computing power for local deployment and secondary development will continue to grow. Looking at the hardware requirements of DeepSeek R1 ranging from a minimum of NVIDIA GeForce GTX 1660 Super to the 40 series, 50 series, and professional-grade A100, H800 GPUs, there will be opportunities to contribute excess computing power. For currently transparent decentralized computing power projects with somewhat redundant capacity, this may be an opportunity for a turnaround (of course, provided that the latency is low enough).

2. AI framework projects were the most hotly anticipated emerging track within Crypto before DeepSeek dropped its "nuclear bomb," and it was the last track I wrote about before the Spring Festival. Now, under the bombardment of DeepSeek, almost all of them are heading towards zero. After all, others can compete head-to-head with OpenAI for less than 6 million, while our leading projects worth tens of billions have not yet produced any AI Agents with practical value.

Since the era of inscriptions, we have almost obsessively pursued securitization. Currently, the cryptocurrency community's tolerance for securitization is quite open. AI framework projects that are not completely on-chain only need an open-source Github code repository and a social account to create a token. The cost of such "library coin" is that one must be prepared to accept the reduction blow of a traditional AI company's foil one day.

In the golden age of AI development, traditional internet companies will not only have DeepSeek as their trump card. The development of AI between the U.S. and China will only accelerate, and Crypto's integration into the upstream and downstream of AI will showcase the advantages of decentralization. Figuring out which direction to combine Crypto with AI to highlight decentralization's strengths and not be suddenly defeated by some mysterious AOE is the key issue. Broadly categorizing the Crypto x AI technical stack, we can divide it into the computing power layer, data layer, middleware layer, and application layer.

In the current landscape, I have always failed to see the necessity of Crypto's existence. However, looking from a future perspective, privacy and security could be a good angle to consider. After all, AI agents replacing or assisting human work has already become a reality today. The issue of how AI processes work data and personal data while ensuring privacy may be a problem that traditional internet companies cannot solve. Furthermore, if an AI agent has payment capabilities, ensuring the security of the wallet would also become a concern. Using blockchain as a compliance and audit layer for AI models should be our main direction for future development.

On the other hand, the question is what to incentivize and in which direction to incentivize. Incentives not only drive the computational layer and model sharing but also teach AI how to interact with the virtual world. Unlike language models trained on decades of worldwide internet textual data, teaching AI how to act correctly requires continuous human annotation. It's like teaching a vision model to recognize what is an animal and what is a car. This task cannot be accomplished simply by outsourcing to a group of college students. Fostering an AI agent capable of interacting with the virtual world requires a complex decentralized network, where numerous individual humans teach AI. This is also a direction. I have elaborated more on this point in my previous articles. What else can be done around incentives?

Combining DeFi with AI to teach intelligent agents to interact with reality, incentivizing AI to gain attention, incentivizing AI-related creations (such as Bittensor's incentive mechanism is a good example), and having a token incentive mechanism automatically adjusted by AI (derived from a question I raised in a previous article, where I stated: When a decentralized project becomes massive and enters the mainstream. concerning its inflation-deflation mechanism, should it rely on code with simple rules, listen to a few or dozens of project members, or listen to those influential figures? Oh, right, we also have governance tokens. However, governance tokens are meaningless before solving the Sybil problem. Democratic voting can never truly reflect governance proposals, as a few wallets held by a16z can veto a large community's approval votes. Then what's the point of voting? etc.)

We indeed cannot aggregate a group of top-tier AI talents like traditional internet companies, purchase or rent a large-scale GPU cluster for training, and dream of recreating a DeepMind within the blockchain space. The meaning of Crypto's existence lies in bestowing another field with the irreplaceable decentralized attribute, much like how we once endowed finance with freedom. AI is humanity's inevitable narrative, but what role can Crypto truly play in it?

3. This is the first time I have mentioned Wordcoin in an article. The cryptographic utopia project initiated by Sam Altman seems somewhat absurd even now when I think about it. Whether to input your iris seems to mean you need to make a choice between national surveillance and corporate surveillance, as if you were in the Matrix deciding whether to take the red pill or the blue pill.

However, the concept of universal basic income or inclusive finance no longer seems like a joke at this stage. The ability to locally deploy DeepSeek, which rivals cutting-edge large models, has already enabled AI to appear in Chinese hospitals and government agencies. According to McKinsey's 2024 forecast report, up to 50% of jobs could be replaced by AI in the next six years. Future versions of Wordcoin might even be issued by the government, and if this trend intensifies, the related tokens for inclusive finance could also see widespread adoption and repeated hype. Considering a window of five to six years, which coincides with Trump's term, could this crypto president issue a similar token? I believe it is highly possible.

4. Based on recent statements by Elon Musk, AI may dominate the Nobel Prize for the next 25 years. Therefore, I find it more interesting and effective to drive AI research through blockchain fundraising (or even contribute resources such as computing power, storage, methods, etc.) than the current DeSci. Perhaps I could call it Decentralized AI Science? That is, DeAIS.

II. Meme Coin is No Longer Just a Meme

Previously, our analysis of Meme Coins revolved around subcultures, community consensus, and virality. However, as I sit in front of GMGN, my daily analysis now focuses on conspiracy groups, rug pulls, and developer exit scams. When a contract address is shared in various community groups, it's a moment of reckoning. Today's meme culture is more absurd than ever before, and in the current Pump.fun environment, it's challenging to find a token that allows you to sleep soundly. You might just step away for a moment, and the chart will plummet.

The continuously simplified asset issuance process and the high degree of anonymity in blockchain have fueled this frenzied gambling culture, where an unknown project team can treat the crypto world as an ATM. The evolution of memes has become increasingly arbitrary, covering not only meme coins based on algorithms but anything from events to individuals and even AI.

Without a cultural core and consensus cohesion, so-called pioneering projects can be forgotten in just a few weeks. The wave of celebrity coins starting from Trump lasted only a month. Following a tweet from President Mile, billions of funds once again flowed out of Sol to the outside world, marking the beginning of the meme culture's decline. President Mile's response to this was simple: delete the tweet and reply to everyone, "I have no knowledge of this."

The rapid development of AI has taken away too much attention from this world, making it difficult for technocrats to progress. Retail investors who have abandoned value investing can only gamble in a bear trap, believing they are among the few lucky ones. The increasingly scarce liquidity is repeatedly drained, reflected in the continual red candles on both Cex and Dex platforms, as well as the disdain of traditional capital and outsiders towards altcoins.

III. Continuing to Dig for Abandoned Swords

The periodic law has clearly failed, and all efforts to dig for the sword have been in vain. A bull run for BTC does not mean a bull run for altcoins, but a bear run for BTC definitely means a bear run for altcoins. Our understanding of altcoins must be revolutionized; the altcoin market is no longer one that can be supported by just a whitepaper. A large project listed on a top-tier CEX must be mature enough to sustain its own price.

Reflecting on the seven years of token growth, in 2017, there were fewer than 2,000 listed tokens in the market, while in 2024, the number of listed tokens has approached 25,000 (data from CoinGecko, including delisted tokens). The exponential expansion of tokens is essentially the irreversible evolution of the blockchain's low-entropy value system into a high-entropy noise system. While in 2017, each token embodied the ideal of "disrupting the world," by 2024, tokens have evolved into chips for liquidity exit. The emergence of more tokens has not brought about more innovation and real-world adoption, but the sky-high valuations of flagship projects have exponentially increased the market's liquidity demands.

As mentioned earlier, without recognition from the external world, retail investors cannot support the valuations of these projects. The listing of most altcoin projects often marks the historical all-time high, with Binance being the final stop. The crypto space needs an innovation; flagship projects should be able to justify their massive funding, and Bybit's experiment with publicly disclosing project financial reports may be a solution. However, in my humble opinion, the market needs a deep bear market to reshape the valuation system and listing standards of current altcoin projects.

IV. Confusion

I once saw a glimmer of light on Ton, thinking that the beginning of consumer-grade Crypto applications had arrived. However, this brief light faded away and vanished with the trend of Tap to Earn. Five years ago, the liquidity mining derived from DeFi brought the crypto space to an unparalleled peak, yet after five years, our only successful area is still DeFi.

The topics of conversation among insiders today are very simple: Have you bought BTC? Have you shorted? Give me a CA. Everyone is perplexed; we can no longer find the right direction. Aside from BTC, holding any token causes sleepless nights. Diamond Hands is not a term of praise today; if you don't buy BTC, it's more like a synonym for a fool.

When I open various crypto media on my phone, it feels like reading The New York Times and tabloids, with all phenomena reflecting that most of the hope in this circle has been placed on policies and eyeballs. From a VC's perspective, we may only invest in utility products in the future, where an asset issuance platform becomes a tool vendor, a rent collector, just to survive.

Epilogue

Clearly, this is not the situation we would like to see. Although the current state of Crypto seems to be lost in a fog of uncertainty, the success of DeepSeek has proven to us that innovation is still the most effective way to break through the current dilemma. Crypto is currently in the best-ever policy environment, with the highest level of attention, funding, and robust infrastructure. In the near future, there will be many more altcoin ETFs that can inject liquidity back into the market. We are clearly in the mainstream, but we are trapped within our own walls. The ebb of the Meme Coin wave may be hiding a turning point, and the future of humanity may not be just AI.

Original Article Link