Some smart ppl on my tl been bullish on #DeSci# and PoUW narratives recently. Spent some time digging deeper and I think DeSci x Proof of Useful Work (PoUW) might be one of the few setups where crypto actually pays for real output instead of just incentives. Why? PoUW is trying to turn energy + GPUs into useful output like AI training, molecular simulation, protein folding, ZK optimization, or biomedical research. Right now, AI infra is eating capex. Big tech is racing for the best models, locking up GPU supply. AI compute is scarce, expensive, and very real. But most workloads aren’t training frontier models. They’re inference, fine-tuning, simulations, scientific compute… stuff that doesn’t need perfect clusters. That kind of compute can run on fragmented, idle, or lower-tier GPUs. So why is crypto still mostly paying people to farm points instead of paying networks to run useful workloads? DeSci might be the first crypto-native buyer class that makes PoUW make sense. Because DeSci needs compute, data, simulations, agents, and ways to fund research before TradFi even cares. And this demand is continuous, not one-off. Projects across the stack already point there: – @opentensor | $TAO → subnets compete to produce useful AI outputs, including scientific workloads. – @dynexcoin | $DNX → miners solve real-world optimization problems instead of hashing. – @gensynai | $AI → distributed training + verification layer. – @akashnet | $ASK → supply side of cheap, flexible GPU compute. On the demand side: – @BioProtocol | $BIO → connecting researchers, funding, and compute. – @vitadao | $VITA → funding biotech research with onchain capital. – @Molecule_sci → turning research into IP that can be owned and licensed. #DeSci# is moving from DAO votes funding science into onchain research pipelines that can buy useful work repeatedly. PoUW gives the supply side of useful compute. DeSci gives the demand side that actually has a reason to buy it. Scientific tasks get posted onchain with a budget, dataset, target metric, and validation requirements. → compute providers run the job offchain through decentralized GPU networks. → validators check results via redundancy, benchmarking, or verification layers. → payment settles only if output passes. → resulting value flows into datasets, models, or tokenized IP. If this loop works: – more research demand → more compute usage. – more compute usage → better network efficiency. – better efficiency → more real output. – more output → stronger demand again. If this thesis really plays out, DeSci x PoUW could become one of the first real demand loops in crypto. Still early, infra not fully mature, verification remains a bottleneck. But this is one of the few areas where I see: – real users – real workloads – real value being created