I'll believe sharding when I see it. It's a much harder problem than it seems. Also, the dirty little secret of zkrollups is that they don't scale the system

I'll believe sharding when I see it. It's a much harder problem than it seems. Also, the dirty little secret of zkrollups is that they don't scale the system or make the system more trustworthy in this context. First, the prover, while asymptomatically fast (as one would expect from the PCP theorem), incurs constant factors that are still so high that you need a really beefy server to generate proofs in a reasonable amount of time. Second, even if that wasn't a problem, provers are still bandwidth-constrained and require BFT agreement -- each prover has to process all zkrollup transactions, and a malicious prover can still ignore transactions while generating correct proofs. At best, this leaves provers in a similar position as high-bandwidth chains like EOS, where only a few beefy nodes can feasibly process the requisite transaction load, but a BFT majority of nodes need to agree on state in order to prevent malicious censorship. I honestly think zkSNARKs and zkSTARKs are wasted on blockchains. Imagine if AWS, grid computers, and supercomputers offered a zk prover interface, where they run your computations on beefy hardware and give you a proof that it was faithfully executed, thereby removing themselves from the TCB. I'd pay to use that service before I'd pay to use Ethereum.