App-specific L3s on zkEVM L2s and why zk is the endgame
Quote180 words
Thesis that high throughput apps need app-specific L3s with parallelized VMs like Sealevel or Move, verifying plonky3 proofs on a cheap zkEVM L2. Argues optimistic rollups' edge fades as DA cost dominates proving cost and proof systems mature.
Key claims
DA cost likely overshadows proving cost even post EIP-4844
The optimistic rollup advantage will disappear as proof systems and hardware mature
Long term zk is the endgame
for high throughput on-chain applications what we need is app-specific L3s that run highly performant, parallelized VMs like Sealevel VM or Move VM and that verify plonky3 validity proofs on the already cheap zkEVM L2
probably DA on L2 is already sufficient to get to almost no cost but if needed, that can also be scaled by opting for a specialized DA layer like Celestia, Avail or EigenDA which can elastically scale data capacity with the number of (light) nodes in the network
L2 settlement layer enables cheap L2<>L3 and L3<>L3 interoperability already & zk makes bridging/interoperability even more seamless, especially in a system like Polygon 2.0 where proofs from rollups are aggregated and an overall network state is settled/verified on L1 recursively
as long as DA cost overshadows proving cost (which it likely will post-EIP4844 as well), the edge that abstracting the proving cost allows for is limited for ORs. And mid-long term this advantage will disappear as well as proof systems & hardware mature and make proof generation & verification more efficient
long term zk is the endgame
QUOTING @