Chainfeeds Summary:
The next stage will no longer be about theoretical breakthroughs, but about operation and scaling: proving whether it is easy to use enough, whether the supply is truly decentralized, and whether the pricing continues to converge towards general-purpose computing power.
Article source:
https://x.com/castle_labs/status/2019088487119544577
Article Author:
Castle Capital
Opinion:
Castle Capital: In 2018, ZK's focus shifted from privacy to throughput and scalability. Ethereum's scaling path became clearer: for many workloads, "verifying a proof once" was much cheaper than "re-executing all computations." Thus, ZK became a tool for compressing massive computations into concise proofs. This directly shaped the development direction of zk Rollups and privacy systems: instead of requiring each validator to re-execute every state change, only a concise proof of validity is verified. In the Rollup architecture, execution occurs off-chain; the proof of validity is submitted on-chain, allowing Ethereum to accept new states without replaying transactions. During this phase, privacy protocols like Aztec, and general-purpose zk Rollups such as zkSync, Starknet, and Scroll, drove the large-scale engineering of ZK technology. In the mid-2020s, ZK adoption shifted again: from "designing circuits for a single purpose" to a general-purpose proof infrastructure. This phase saw the emergence of three key components: zkVM, capable of generating proofs for any program; coprocessors, which perform specific queries on the on-chain state and generate proofs; and proof networks, which industrialize and scale proof generation. These systems were driven by teams such as Brevis, Axiom, Lagrange, Succinct, RISC Zero, and Cysic. As ZK evolved from a "privacy primitive" to a "general-purpose proof tool," the entire technology stack began to break down into multiple specialized layers. The ZK Stack has developed into an ecosystem composed of various specialized layers. Previously, verifying anything on-chain typically required significant human and financial resources. Now, this cost has been shifted to the proof phase. The process has become: the validator performs a quick check, the prover performs a large amount of computation, the computation is compressed into a single proof, and the cost is reflected in hardware, energy, and latency. In July 2025, the Ethereum Foundation proposed the "real-time proof" goal for L1 zkEVM: to prove 99% of mainnet blocks within 10 seconds using local hardware with a power consumption of no more than $100,000 and a power consumption of 10kW, under open-source software conditions. By December 2025, the Foundation reported that the proof latency had decreased from approximately 16 minutes to 16 seconds, and the cost had decreased by 45 times. zkVM could now prove 99% of blocks within 10 seconds under the target hardware configuration. The essence of proof is to allow the chain to verify results without repeating the work. If the proof cost is too high, Rollups must subsidize users; only when the proof cost decreases can the fee potentially decrease without compressing profits. Ethproofs benchmark data shows that in 2025, the average proof latency decreased from approximately 16 minutes and 44 seconds to 60 seconds, and the average cost decreased from $1.69 to $0.0376. This indicates that proof is moving from "laboratory work" to commercial infrastructure. zkVM enables the proof-of-concept nature of any program, forming the core of the "proof, not execution" approach. Vitalik Buterin recently pointed out that zkEVM has entered the alpha stage, with performance already production-ready; the remaining focus is on security. The key going forward is no longer just faster proof, but proving the system's reliability, its large-scale deployment capability, and the true decentralization of the supply chain.
Content source 
