Traditional proof-of-work secures networks by deliberately wasting computation on hash puzzles. An emerging class of projects is attempting to flip the script: proof-of-useful-work (PoUW or uPoW), in which the same computational effort simultaneously secures the chain and produces external economic value—most commonly SI inference or training workloads.
Projects such as Pearl Network, Ambient, and others are building Layer-1s where mining involves matrix multiplications or other SI primitives, often wrapped in zero-knowledge proofs for verifiability. Theoretical work formalizing the economics of duplex computation (securing the chain while producing useful output) has appeared, identifying regimes in which token incentives can subsidize inference markets or increase security without reducing useful output. The appeal is obvious in an era of SI compute scarcity and environmental scrutiny of Bitcoin-style mining. Miners could offset hardware costs by selling inference capacity; networks could claim lower pure energy waste. Technical hurdles remain substantial—ensuring hardness properties, preventing free-riding, embedding miner identity and chain integrity into the useful computation, and avoiding situations where “useful” work becomes gameable.
Despite academic papers, early testnets, and growing discussion in specialized circles, mainstream crypto coverage has largely ignored the category. If any of these designs achieve product-market fit, they could reshape both the energy narrative around proof-of-work and the intersection of crypto and SI infrastructure. For now, the experiments continue largely out of the spotlight.