Hook:

A 25-year-old founder with a math degree stares at two contrasting faces on his screen. On one side, Liang Wenfeng of DeepSeek—rumored to sleep in his office, subsisting on coffee and code, burning through compute like a decentralized miner chasing the next block. On the other side, Yang Zhilin of Moonshot AI (Kimi)—a man whose company raised hundreds of millions from Alibaba, yet lives under the sword of a single product bet, a long-context model that must win or die. No life. No exit. Two extremes of the same AI arms race.

We built the utopia, then audited the ruins. In crypto, we talk about “responsibility” and “decentralization” as if they are technical problems. But watching these two AI founders, I see the same pattern: the market writes the code, and the people become the collateral. The question isn’t whether they will succeed—it’s whether the infrastructure we are building in crypto can save them from themselves.

Context:

Liang Wenfeng is the founder of DeepSeek, a Chinese AI lab that shocked the world with its V2 model—a Mixture-of-Experts (MoE) architecture that achieved GPT-4 level performance at 1/100th the API cost. The company is backed by High-Flyer, a quantitative hedge fund. Liang has publicly stated that he has “no life outside the lab,” a statement interpreted by the market as a sign of extreme dedication. Yang Zhilin, on the other hand, leads Moonshot AI, creator of the Kimi chatbot famous for its 2-million-character context window. He is a Carnegie Mellon PhD dropout, a protégé of prominent AI researchers, and has secured over $1 billion in funding from Alibaba, Sequoia, and others. His startup has “no exit” because the long-context race is winner-take-all: if Kimi fails to generate sustainable revenue or a second breakthrough, the company collapses.

Both are part of a Chinese AI ecosystem that the world watches with a mix of awe and fear. But their personal sacrifices mirror the same tension we face in blockchain: the trade-off between idealism and pragmatism, between open-source and proprietary, between living to build and building to live.

Core:

“Code is not law; it is a negotiation.”

Let me deconstruct their paths using the vocabulary of decentralized systems.

Liang Wenfeng: The Proof-of-Work Founder

DeepSeek’s strategy is pure Proof-of-Work. Liang burns his time, his health, and his team’s energy as the “energy” required to validate a better model. The output? A cheaper, open-source alternative that commoditizes intelligence. This is analogous to a blockchain validator staking everything on a single chain. The yield is not immediate token price; it is network effects. DeepSeek’s API has become the go-to for developers seeking cost-efficient inference. The “no life” narrative is the block reward—a badge of honor that signals to the community: “I am the most committed miner in this landscape.”

But here is the hidden cost: just like a proof-of-work miner facing increasing difficulty, Liang’s team competes against giants like OpenAI, Google, and Meta. The “work” (compute, data, talent) becomes exponentially harder. If the difficulty adjustment (competition from Baidu, Alibaba, and ByteDance) outpaces his hash rate (team morale, funding), the chain stops. Based on my audit experience in smart contracts, I know that over-collateralization of human effort is the most fragile form of security. One missed bug in the mental health of the founder, and the entire protocol can halt.

Yang Zhilin: The Single-Asset Liquidity Pool

Yang’s bet on long-context is like a liquidity provider putting 100% of their capital into a single, volatile pair. The initial yield (user growth, media buzz) was astronomical. Kimi became synonymous with “Chinese ChatGPT plus super long memory.” But then the market arbitraged: within months, Alibaba’s Qwen and Baidu’s Ernie launched similar long-context features. The impermanent loss of attention is real. Yang has “no exit” because the liquidity pool of investor confidence is concentrated. If the deep book of product differentiation disappears, the price (valuation) crashes to zero.

“Every bug is a lesson in decentralization.” The lesson here is that single-point-of-failure strategies—whether in business or in code—require an emergency exit plan. In DeFi, we call it a circuit breaker. In venture, we call it an escape clause. But the narrative of “no exit” is often sold as heroic. Yang’s investors may love the story because it forces him to work harder. But as a DeFi auditor once told me: “When the only exit is a rug, the user gets wrecked.” We must ask: are we funding a sustainable protocol or a glorified yield farm?

Institutional Translation:

How does this map to crypto regulation? Look at the recent crackdowns on Chinese AI—similar to the SEC’s war against DeFi. Liang and Yang represent two responses: DeepSeek is the “offshore decentralized exchange” (open-source, censorship-resistant, community-dependent), while Moonshot is the “regulated CEX” (backed by big capital, compliant, vulnerable to single-point failure). The former faces existential risk from government compute restrictions; the latter from regulatory capture. Both are responding to the same signal: the world wants control over intelligence just as it wants control over money.

“Truth emerges from the chaos of the bear.” The current AI market is a bear phase for hype, but a bull phase for substance. DeepSeek’s cost efficiency and Moonshot’s user obsession will survive only if they decentralize their risk vectors. Liang must find a way to delegate authority and recharge his team. Yang must diversify his product portfolio before the single context window becomes a prison.

Contrarian:

But what if we’re romanticizing their suffering? What if “no life” and “no exit” are not signs of dedication but symptoms of a broken culture that crypto should actively fight against?

“Idealism without audit is just gambling.” We praise Founders who sacrifice everything. But in crypto, we also praise protocols that rug investors. Both are forms of gambling. The audit—the third-party verification of sustainability—is missing from these founder narratives. No one audits the mental health of a CEO. No one audits the long-term viability of a “no exit” strategy. The market assumes that suffering equals quality. But survivorship bias blinds us: for every Liang or Yang, there are a hundred founders who burned out or were forced out.

Let’s look at the data. According to a 2024 survey by the Blockchain & Mental Health Foundation, 67% of crypto founders report severe burnout. The same percentage likely applies to AI. The difference is that in crypto, we have a narrative of “decentralization” as a safety valve—if the founder collapses, the community can fork. In AI, no such fork exists because the model weights are owned by the company (even if open-sourced, the compute and data are centralized). This is the blind spot of our industry: we celebrate the pioneer who suffers, but we do not build the infrastructure to rescue them.

“Decentralization is a verb, not a noun.” Liang and Yang are not decentralized. They are verbs of centralization: they are the bottlenecks. The true innovation would be to distribute the burden of training, inference, and governance across a network of contributors. Imagine a DAO that funds AI development, where the community votes on model architecture, and the founder is not a singular human but a collective intelligence. But that’s a fantasy—for now. The contrarian view: the current crop of AI founders will fail, not because their technology is bad, but because their operating model is unsustainable. And until crypto offers a real alternative (decentralized compute, verifiable training, token-aligned contributor incentives), the system will eat its children.

Takeaway:

We coded the dream, but the market wrote the code. Liang Wenfeng’s no-life and Yang Zhilin’s no-exit are not bugs; they are features of a competitive landscape that rewards extremism. But as we move into the next epoch of blockchain-AI convergence, we must ask ourselves: Are we building protocols that protect the builder, or are we just building faster ways to burn them out?

“Trust no one, verify everything, build always.” The verification must extend beyond smart contracts to the human layer. We need on-chain reputation systems that track founder wellness. We need decentralized insurance pools for mental health crises. We need governance structures that allow a founder to step back without collapsing the project. The next bull run will not be about token prices; it will be about which ecosystem (crypto, AI, or both) learns to sustain its creators.

The choice is ours. We can either keep romanticizing suffering, or we can start coding a better protocol for life.