Archetype(@archetypevc):zkML still feels impractical for broad use cases But in Session 11 of Proof is in the Pudding, @cryptodavidw explains why it’s less ridiculous than it seems, and walks us through what’s needed for it to become a more widely useful technology ⌛️ TIMESTAMPS 0:00 Intro 1:44 What is zkML 5:32 Inference vs training 7:52 Alternatives to ZK for verifiable inference 10:44 Inside a transformer 15:25 Proving transformer computation 27:12 Quantization & SNARK-friendly models 31:58 KV caching & optimization opportunities 35:08 Agent-to-agent proofs & dynamic circuits 37:18 How zkML scales 38:50 Market for verifiable ML Proof is in the Pudding is produced in collaboration with @zksecurityXYZ

2026.09.30 15:00

zkML still feels impractical for broad use cases But in Session 11 of Proof is in the Pudding, @cryptodavidw explains why it’s less ridiculous than it seems, and walks us through what’s needed for it to become a more widely useful technology ⌛️ TIMESTAMPS 0:00 Intro 1:44 What is zkML 5:32 Inference vs training 7:52 Alternatives to ZK for verifiable inference 10:44 Inside a transformer 15:25 Proving transformer computation 27:12 Quantization & SNARK-friendly models 31:58 KV caching & optimization opportunities 35:08 Agent-to-agent proofs & dynamic circuits 37:18 How zkML scales 38:50 Market for verifiable ML Proof is in the Pudding is produced in collaboration with @zksecurityXYZ