What is DeFAI & Reflection on DeFAI

In recent months, people have started talking about DeFAI—DeFi combined with AI/AI agents. Based on Haotian’s article on PANews, DeFAI can be categorized into four main areas:
1. AI models in user interfaces let users interact via natural language, automating transactions.
2. AI-driven DeFi investment tools optimize returns and portfolio management.
3. AI market analysis tools that assist traders like aixbt.
4. AI-native platforms/frameworks like ai16z, Virtual, and Bittensor, etc.
Why Now?
Some aspects of DeFAI, such as AI-driven yield optimization, have existed before. For example, Alpaca Finance V2 claimed to use machine learning to enhance its vault strategies. However, AI's role in yield optimization has been minimal because the underlying revenue sources haven't changed. A small percentage increase in yield doesn’t necessarily attract users, especially in a DeFi market where security, reputation, and historical performance matter more. The DeFi market operates in an 80/20 dynamic—most users prefer stability over slightly higher yields (unless the yields are extraordinarily high).
The Key is User Experience
DeFi’s market hasn’t shown consistent growth; instead, it has fluctuated significantly. According to DefiLlama, total value locked (TVL) peaked at $180 billion in December 2021 but has since dropped to $108 billion as of February 2025. The reason? Liquidity is scarce.
So, what’s next? Generative AI has advanced to the point where it can handle tasks that previously required human intervention. My hypothesis is that the DeFi projects that will succeed in this cycle are those that enhance user experience and drive broader adoption. AI isn’t a magic bullet, but it can lower barriers to entry, automate complex processes, and ultimately make DeFi more accessible to a wider audience.