Some people think that my criticism of web3 AI is just a one-sided endorsement of web2 AI? In fact, this is not the case. Anything I have written about web2 DeepSeek, Manus, MCP, and the like does not mean that I am "favoring one and disfavoring the other". From the very first day I started cheering for AI+Crypto, I have been thinking about how to create innovation in web3 AI, how the narrative should evolve, where the development problems lie, and how to adjust.
1) My main focus is on the AI+Crypto field, and I have some understanding of the innovative development of web2, but not enough to see the whole picture, so I don't make absolute judgments. Like most people, I just know that after a series of innovations from web2 AI, web3 AI collapsed. After serious research, I found that web2 AI is steadily progressing in terms of talent pipeline, innovative thinking, and pragmatism, which is in line with the market's expectations for AI development, and that's all.
Indeed, I have fantasized about how great it would be if the web3 AI track could produce such innovations. There's no need to compete with the web2 team, just find a differentiated path and do some web3-specific narratives and technical implementations, at least to live up to the technical faith of the 'buy the dips' Holders.
Unfortunately, the expectations have been repeatedly dashed. Seeing the token prices of those mainstream projects approach zero endlessly, seeing old and new believers leave with their flesh cut, I feel incredibly angry. If we can't keep up with the differentiated innovation of web2, are we not even willing to play the narrative and hype of web3 expertise?
2) Sitting on Crypto, I don't have the grand mission of revolutionizing human technology like AI. From the beginning, I understood that AI is an opportunity for the internet, and Crypto also needs this opportunity. The true meaning of AI + Crypto is to endow Crypto with the new soul of AI, so that we can revitalize the "old narratives" that Crypto has once hyped up and become too bubble-like. We just need to treat AI as a new productive Catalyst.
Based on this premise, whether it's the new developers entering the field to do AI, or the original web3 old developers wanting to do innovation with new narratives, they should anchor on a goal: use the Tokenomics and decentralized flexibility and resource advantages of web3 to do web3 Native narrative innovation.
Web3ers must acknowledge that the key to the current prominence of web2 AI Agents is that they can bring disruptive innovation experiences to end-users through "monopoly + scale", which is an advantage that web3 AI Agents cannot possibly have. Trying to compete head-on is not a match for web3 AI.
However, it is precisely some upstream distributed framework layers, some distributed trust frameworks and incentive network designs that web2 is unwilling to touch, as well as the design of components that implement on-chain interaction communication beyond the Agent computing layer, that are the points where web3 is worth focusing on. This is also the point that "future AI" development cannot avoid at a certain stage, and it is the differentiated advantage of web3 AI.
