E06 with Jens Groth: Scaling Zero-Knowledge Proofs with Instruction Sorting

Scalable ZK at internet scale — with Jens Groth (pronounced “Yens”), Chief Scientist at Nexus and creator of Groth16. We dive into Nexus’s zkVM (RISC-V), its distributed prover network for low-latency proofs, and why “proofs-of-proofs” matter. Jens previews “instruction sorting” (grouping adds/XORs/etc. to remove redundant constraints), compares ZK vs TEE vs FHE in real systems, and outlines what real maturity in ZK stacks looks like (open source, audits, formal verification of verifiers, and reproducible builds). If you care about rollups, privacy-first apps, and verifiable backends, this episode is a clear blueprint.

HIGHLIGHTS
• How a zkVM compiles, runs, and emits an extended witness + proof—and what the verifier actually checks
• Why an orchestrator + many workers wins on latency (and where privacy trade-offs appear)
• TEEs (Intel TDX, AMD SEV) for speed vs ZK for public verifiability vs FHE/MPC for data-in-use privacy
• Instruction sorting: reorder the trace into per-op buckets to shrink constraints—especially powerful with proof composition
• What “mature” ZK looks like in practice: audits, formal verification for verifiers, supply-chain discipline, reproducible builds

LINKS MENTIONED
Nexus — zkVM v3, prover network, and blog:
• [https://specification.nexus.xyz/](https://specification.nexus.xyz/)
• [https://docs.nexus.xyz/zkvm](https://docs.nexus.xyz/zkvm)
• [https://blog.nexus.xyz/nexus-zkvm-3/](https://blog.nexus.xyz/nexus-zkvm-3/)
• [https://blog.nexus.xyz/nexus-launches...](https://blog.nexus.xyz/nexus-launches...)

Confidential compute & TEE references:
• Phala Network overview (TDX direction): [https://docs.phala.com/overview/phala...](https://docs.phala.com/overview/phala...)
• dstack-TEE (attested, policy-driven confidential compute): [https://phalanetwork.mintlify.app/doc...](https://phalanetwork.mintlify.app/doc...)
Additional Background / Context Material:
• RISC-V ISA (how zkVMs model CPU steps): [https://riscv.org/specifications/rati...](https://riscv.org/specifications/rati...)
• STARKs primer and resources: [https://starkware.co/stark-101/](https://starkware.co/stark-101/) • [https://starkware.co/stark/](https://starkware.co/stark/)

ABOUT THE SHOW
Blackbox Podcast — Confidential AI, Zero Trust & Privacy. Technical insights to navigate the AI Agent Decade and build toward a safe AGI future. Each episode adds concrete patterns you can apply to scale verifiable, privacy-preserving systems.

If this helped, drop a like, subscribe, and share with a builder wrestling with verifiable compute.