The DUSK project is quite interesting—it's not following the hype with concepts, but instead focusing on bottom-layer optimization. The PLONKup proof system developed in collaboration with the University of Barcelona is a prime example, optimizing multiple polynomial commitment processes based on PLONK. The actual test data is very intuitive: generating a multi-block aggregation proof takes only 1.7 seconds, which is 40% faster than native PLONK, and it also supports recursive verification—meaning the validity proofs of multiple blocks can be aggregated into a single one, significantly reducing on-chain verification pressure.

Pairing with the rkyv zero-copy deserialization framework, the data loading efficiency is notably improved. Previously, I tested a normal ZK project loading 100 transactions, which took 1.2 seconds; DUSK only needs 0.5 seconds, and it also saves 35% of memory usage. This is a very tangible improvement in the experience for light nodes.

Additionally, their customized implementation of Poseidon Hash is specifically designed for zero-knowledge proof scenarios. In collision resistance tests, it far exceeds SHA-256, with hash computation time 28% less than similar functions, fully capable of handling the high-frequency calculations required by privacy contracts.

But finally, here’s an interesting detail—during the initial test, they didn't enable the std feature for rayon parallelism, and as a result, proof generation time doubled. Later, after checking the crates.io documentation, they found they needed to manually enable parallel compilation parameters. After adjustment, performance immediately met expectations. This kind of meticulous attention to low-level details is quite impressive. Not relying on hype, but instead on academic collaboration and engineering optimization to refine the product, making it a true tech-driven force in the privacy public chain sector.