Killer apps that want to use ZK often don't care about ZK for scalability. They care about ZK for privacy or for verifiable compute where they otherwise would not have it.

Non-uniform incremental verifiable computation (NIVC) is AWESOME! It is the bedrock, that supports apps that use ZK for consumer-side / privacy / and DePin. 
 
Let's understand what it is and why its useful. 🤯

Read:

https://blog.icme.io/pick-prove-profit-the-nivc-singularity-2/

Or read thread 1/n

NIVC was introduced in 2022 in SuperNova. It differs from IVC in that you do not prove a single universal circuit over and over, but rather can pick a-la-carte' which opcodes to include for each IVC iteration.

Other than just for zkVM.. it is actually useful for prover networks.

https://eprint.iacr.org/2022/1758.pdf

2/n

Memory efficient ZKP at https://NovaNet.xyz

Off-chain compute triggers on-chain actions, on any device.
https://blog.icme.io/

Private Basename
zkbasename.com

1. You can pick the memory usage of your zkVM to match the machine specs of your users.

In a more recent paper, Nebula, you can pick how many opcodes you want to prove per IVC iteration. This means that you get speed and memory specs that match the users machine. We implemented Nebula and its SUPER fast. Moreover, it works on a large range of devices, from very small to production grade hardware. 

https://eprint.iacr.org/2024/1605.pdf

3/n

2. You only pay for the circuits that you use. In the prover network setting, this means custom & specialized provers can undergo natural selection. With creators getting paid a premium for winners. Create a new proving scheme for AI, FHE, or DA? Get paid for that implementation in an NIVC prover network (https://x.com/NovaNet_zkp).

How many schemes have been implemented that were superior for specific tasks but never adopted? Never more as long as there is a demand for them 🙂.

4/n

3. Not all proving schemes need to do the tedious work of proof composition to get onto L1, or even re-work themselves to be privacy preserving. This all can be done at the NIVC prover network or VM level. The current plans for many zkVM actually introduce 'folding' for continuations and privacy. This applies more broadly to any proving scheme that fits into the NIVC framework.

5/n

4.  Prover networks that use NIVC are based on maths and not re-execution. This means succinct verification of proofs and fair reward distributions. 

6/n

5. Any hardware advanced or specilized zkVM or scheme can be used in an NIVC based network as long as its verifier can be composed into a circuit. This 'composition' work can be done at the network level rather than each team doing it themselves.

For consumers this means they get the best ZKP for their needs.

7/n

6. NIVC can be used to aggregate locally created privacy proofs, resulting in a single succinct proof for on-chain verification. i.e. Many N circuits becoming one circuit posted on-chain. 

8/n

7. NIVC can be used with continuations to break large computations into smaller ones running in parallel. This can also be done with multifolding.

9/n

I can keep going on the benefits of NIVC, they don't stop here. 🤗

But for now, FIN.

Killer apps prioritize ZK for privacy, not scalability. NIVC is essential for consumer-side verifiable computation. 🚀🔒

(40 characters)