I really have to complain about @katana.

You’re keeping up with @KaitoAI’s leaderboard. It’s full of foreigners, and a large proportion are 🇰🇷 Korean KOLs.

This is really unfriendly to our great Chinese 🇨🇳 users!

I hope you pay more timely attention to the China 🇨🇳 region. We are the real main force‼️

I’ve posted more than a dozen articles and I’m still at the kids’ table.

Among the new L2 tracks, there are only a few solid ones, and #katana is one of the few that can compete. It’s the only L2 I have my eyes on.

The biggest difference between Katana and traditional L2s is that it wasn’t created to “help Ethereum scale” at all, but instead has shifted its focus directly to AI—it’s an execution layer built specifically for AI Agents.

Traditional L2s aim for low cost, speed, and higher TPS, while Katana solves the fundamental problem of AI being unable to run on-chain: long-running computations, high-frequency loop executions, external calls, model inference, and composability between agents.

Simply put, traditional L2s are like the “express lanes” of blockchain, while Katana is a “city-level infrastructure” designed specifically for AI, allowing AI applications to truly run on-chain, run efficiently, be affordable, and be composable with each other.