I think the classic AMM model x*y=k is great for bootstrapping, but it’s terrible for capital efficiency compared to an order book. The problem has always been that running an order book on-chain is too expensive gas-wise. I found a deep dive into the data architecture of scalable DEXs and the shift toward hybrid models: https://coinjournal.net/news/the-data-architecture-of-scalable-dexs-solving-for-liquidity-latency-and-mev-protection/. It discusses solving for liquidity fragmentation by rethinking how state is stored. If the backend architecture can handle the throughput without clogging the L1, we might finally see order books that don't cost $50 to place a limit order.