Cold Thinking: What are the differences between AI and the Crypto track?

Author: Haotian

Everyone says that Ethereum's Rollup-Centric strategy seems to have failed? And deeply despises this L1-L2-L3 nesting game, but interestingly, the development of the AI track over the past year has also gone through a rapid evolution of L1—L2—L3. Comparing the two, where exactly is the problem?

1. The layered logic of AI is that each layer addresses core issues that the upper layer cannot solve.

For example, L1's LLMs address the basic capabilities of language understanding and generation, but logical reasoning and mathematical calculations are indeed weaknesses; thus, at L2, reasoning models specifically tackle this shortcoming. DeepSeek R1 can solve complex math problems and debug code, directly filling the cognitive gaps of LLMs. After completing these foundations, the AI Agent at L3 naturally integrates the capabilities of the first two layers, allowing the AI to transition from passive responses to proactive execution, capable of planning tasks, calling tools, and handling complex workflows.

Look, this kind of layering is "capability progression": L1 lays the foundation, L2 fills in the gaps, and L3 integrates. Each layer makes a qualitative leap based on the previous one, and users can clearly feel that AI is becoming smarter and more useful.

2. The layered logic of Crypto is that each layer patches the issues of the previous layer, but unfortunately brings about a brand new and larger problem.

For example, when the performance of L1 public chains is insufficient, it's natural to think about using layer 2 scaling solutions. However, after a wave of competition in layer 2 infrastructure, it seems that while Gas has decreased, TPS has increased cumulatively, liquidity has become dispersed, and ecological applications remain scarce, making the excessive layer 2 infrastructure a significant problem. Consequently, there has been a move to create layer 3 vertical application chains, but these application chains operate independently, unable to enjoy the ecological synergy effects of the universal infrastructure chains, resulting in an even more fragmented user experience.

In this way, this layering has become a "problem shift": L1 has bottlenecks, L2 provides patches, and L3 is chaotic and fragmented. Each layer merely shifts the problem from one place to another, as if all solutions are focused solely on the issue of "issuing tokens."

At this point, everyone should understand what the crux of this paradox is: AI layering is driven by technological competition, with OpenAI, Anthropic, and DeepSeek all striving to enhance model capabilities; Crypto layering is hijacked by Tokenomics, where the core KPIs of each L2 are TVL and token prices.

So, essentially one is solving technical problems while the other is packaging financial products? There may not be a definitive answer as it depends on personal perspectives.

Of course, this abstract analogy is not so absolute; it just seems very interesting to compare the development contexts of the two, a little mental massage for the weekend.