Tokens, Models, and Bubbles: The Crypto × AI Game in the Tier 1 Market
author:Lao Bai
Original title: Looking at Crypto × AI from the Primary Market: A Tokenization Illusion Experiment
After a lapse of two years, I posted Twi again. I'll follow the research report from two years ago. Even the time was exactly the same, February 10th. (Related reading:ABCDE: Sorting out AI+Crypto from a Primary Market Perspective）
Two years ago, God V had actually vaguely expressed that he wasn't very optimistic about the various Crypto Helps AI that were popular at the time. The popular troika in the industry at the time was computing power capitalization, data capitalization, and model capitalization. My research report from 2 years ago was mainly about some of the phenomena and questions that the troika observed in the primary market. From the perspective of God V, he is still more optimistic about AI Helps Crypto.
Some of the examples he gave at the time were:
AI as a participant in the game;
AI as a game interface;
AI as a rule of the game;
AI as the goal of the game;
Over the past two years, we've actually tried a lot with Crypto Helps AI, but there have been few results. Many tracks and projects have been successful - just one coin transaction, no real commercial PMF. I call it the “tokenization illusion.”
1. Capitalization of computing power - Most are unable to provide commercial-grade SLAs, are unstable, and frequently disconnected. It can only handle simple small to medium model inference tasks. Most of them serve marginal markets, and revenue is not linked to tokens...
2. Data capitalization - High supply-side (retail) friction, low will, and high uncertainty. On the demand side (enterprise), what is needed is a structured, context-dependent, professional data provider with trust and legal responsibility. Web3 projects that are DAO entities are difficult to provide.
3. Model capitalization - The model is inherently a process asset that is not scarce, can be replicated, fine-tuned, and rapidly depreciated, rather than a final asset. Hugging Face itself is a collaboration and communication platform, more like GitHub for ML, rather than App Store for models, so the so-called “decentralized Hugging Face” tokenizing models basically ended in failure.
Furthermore, we have tried various “verifiable deductions” in the past two years. This is also a typical story of finding a nail with a hammer. From zkml to OPML to Gaming Theory, etc., even EigenLayer turned his restaking narrative into verifiable AI.
But it's basically similar to what happened at the Restaking Track - few AVS are willing to keep paying for additional verifiable security.
Similarly, verifiable reasoning is basically verifying “something no one really needs to be verified”, and the demand-side threat model is extremely vague - who is being protected?
There are far more AI output errors (model capability issues) than AI output being maliciously tampered with (countering issues). Some time ago, everyone also saw various security incidents on OpenClaw and MoltBook. The real problem comes from:
Strategy design is wrong
More permissions
I didn't think the boundaries were clear
The toolset interacts unexpectedly
...
There is almost no imaginary nail where “the model has been tampered with” or “the inference process has been maliciously rewritten.”
I posted this picture last year; I don't know if Lao Tie remembers it.
The ideas given by God V this time are clearly more mature than two years ago, also because of our progress in various directions such as privacy, X402, ERC8004, and forecasting the market.
It can be seen that half of the four quadrants he divided this time belongs to AI Helps Crypto, and the other half belongs to Crypto Helps AI, and is no longer clearly biased towards the front two years ago.
Top left and bottom left - Using Ethereum's decentralization and transparency to solve AI trust and economic collaboration issues
1. Trustless and private AI interaction (infrastructure + survival): Using technology such as ZK and FHE to ensure the privacy and verifiability of AI interactions (I don't know if the verifiability reasoning I mentioned earlier counts).
2. Ethereum as an economic layer for AI (infrastructure + prosperity): Enables AI agents (agents) to make economic payments, recruit other robots, pay security deposits, or establish a reputation system through Ethereum, thereby building a decentralized AI architecture without being limited to a single giant platform.
Top right and bottom right - Using AI's intelligent capabilities to optimize the user experience, efficiency, and governance of the crypto ecosystem:
3. Cypherpunk mountain man vision with local LLMs (impact + survival): AI as a “shield” and interface for users. For example, a local LLM (Big Language Model) can automatically audit smart contracts and verify transactions, reduce reliance on centralized front-end pages, and protect individual digital sovereignty.
4. Make much better markets and governance a reality (impact + prosperity): AI is deeply involved in prediction markets (Prediction Markets) and DAO governance. AI can act as an efficient participant, amplify human judgment by processing information on a large scale, and solve various market and governance problems such as insufficient human attention, excessive decision-making costs, information overload, and indifference in voting.
Before we frantically wanted Crypto to help AI, God V was on the other side. Now we've finally met in the middle, but it's just a visual guess that it has nothing to do with various XX tokenizations, or any AI Layer 1. I hope to look back at today's post in two years, and there will be some new directions and surprises.
Twitter:https://twitter.com/BitpushNewsCN
Compare the TG exchange group:https://t.me/BitPushCommunity
Compare TG subscriptions:https://t.me/bitpush
