The journey of building infrastructure and applications for Web3 AI is a long marathon.

Written by: Haotian

Recently, I had conversations with many developers at the forefront of web3AI Build and found that working around web3AI infrastructure is much more complex than I imagined.

1. Currently, most of the active AI projects in web3 are generally meme-driven, boasting a lot of stories that are impossible to realize and implement. The key point is that they have attracted most of the attention and liquidity by quickly issuing tokens to enter the market, along with a mess after the short-term bubble burst (negative EV). This is mainly due to the overly appealing narrative of AI + Crypto, while the challenges of its actual implementation are too great, naturally making it a major disaster zone for token issuance based on narrative from the start.

2. The web3AI infrastructure is essentially a reconstruction of the web2 AI infrastructure, which is often a thankless task. It is similar to how Crypto initially challenged centralization in the name of decentralization; for a long time, decentralized network architectures were criticized for being meaningless and redundant until later DeFi application scenarios found some value capture points.

The current predicament of web3AI is no different from the initial vision of decentralized Crypto. Most people still casually say, "What’s the use of web3AI?" But let’s not forget that decentralized computing power aggregation, distributed reasoning, and distributed data labeling networks can all find entry points in terms of training costs, performance, and practicality. It can only be said that the road ahead is long and arduous, but it is of great significance.

3. The construction and expansion of web3AI infrastructure has a high trial-and-error cost, requiring strong rationalist support. For example, it is well-known that web3AI requires the construction of a data layer, but cleaning large amounts of on-chain and off-chain data requires substantial server operation and maintenance costs, as well as development costs. At the same time, the costs associated with integrating mature web3AI APIs, computing power, and algorithm fine-tuning also need to be considered. If these costs are focused on Agent applications, it can lead to quick explorations of commercial monetization models. However, if the focus is on the infrastructure level, it poses a challenge for many developer teams in the current market context where the technical narrative is not very popular.

The more troublesome aspect is that, unlike traditional web2 infrastructure, web3 AI also needs to solve the collaborative problem between off-chain data and on-chain verification, the model distribution and update mechanism under P2P networks, and the complex design of replacing traditional business models with Tokenomics incentives, among others. The short-sightedness of capital and the market's speculative preference have led to some hot money flowing into Agent applications that were hastily launched purely to ride the trend, resulting in teams that are genuinely working on the infrastructure layer finding it difficult to gain sufficient support.

4. The illusion problem of large models compatible with the "black box" attribute in web3AI infrastructure poses significant challenges to their security and credibility in specific scenarios. See @SlowMist_Team

Recently, the output regarding the MCP security vulnerabilities feels like the professional security audits surrounding MCP can already support Slow Fog's future positioning as an AI audit company. This is just a concrete case that verifies the various unknown security challenges of integrating AI LLMs as foundational data sources into web3 AI infrastructure. However, the issues surrounding web3 AI infrastructure are far from limited to this; there are also directions such as building a verifiable computing framework through web3 cryptography verification and on-chain consensus mechanisms to ensure that the AI reasoning processes can be traced and verified.

In fact, the credible verification and computation framework of AI is the core area that web3AI infra needs to tackle. Currently, large models face significant limitations in adoption within professional fields such as finance, healthcare, and law due to their inability to provide verifiable reasoning processes when handling highly sensitive information. The maturity of web3 AI infra, such as zkVM at the base layer, decentralized Oracle networks, decentralized Memory solutions, etc., can provide AI with a verifiable and provable computation framework, fundamentally helping AI achieve rapid expansion in vertical scenarios.

Above.

The infrastructure construction and application development journey of web3AI will not be achieved overnight, but rather it is a long marathon. Those who can truly build infrastructure and application ecosystems that solve real-world problems, those who can balance hype and value in the Go-To-Market process, and those who can find practical business closures while maintaining technological foresight will be the ones who ultimately succeed in the industry.