Three things worth your weekend. 1. A long-form explainer on zkML, if you want the whole picture in one read. Best fact in it: the largest model proven end-to-end in zero-knowledge so far is around GPT-2 scale. Useful reality check on where this actually is today. 2. A practical walkthrough of running AI inference off-chain with proofs. It makes a point we keep repeating: "zero-knowledge" doesn't mean every input, output and piece of metadata is hidden by default. You choose what stays private. 3. Our own Ep. 8, if you missed it — why creating a proof is expensive, why verifying is nearly free, and why that gap is the whole business model. No thread, no takes. Just things we found good this week💜