The ZK endgame: verifiable compute, field size and verification cost metrics
Quote651 words
Outlines Gevulot's vision that most global compute eventually ships with proof of execution, and the two metrics it proposes for tracking progress: proof field size and on-chain verification cost. Frames a modular ZK stack of proving, verification and messaging layers.
Key claims
Plonky3 uses 31-bit fields while Binius uses a 1-bit field
Plonky2 and Plonky3 verification costs remain high compared to Plonk without EIP-4844
As @VitalikButerin pointed out, ZK is likely as important of a technology as blockchain themselves.
However, ZK is much bigger than just rollups.
So what does the ZK endgame look like and how do we get there? 🚀✨
Let me give you an overview of @gevulot_network's vision ⤵️
As @gevulot_network state in their recent blog, the ZK endgame is not limited to blockchains.
Teams like @gevulot_network are continuously working to make it more cost-effective and feasible for diverse use cases to operate without dependence on trusted computing.
Eventually, most global compute could come with proof of execution.
The real ZK endgame is nothing short of all the world's compute being verifiable
🚀 But how do we get there?
The end goal of ZK builders like @gevulot_network is for the compute and cost overhead of verifiability to approach zero.
To make this a reality, we need to get on a trend line.
This means it is vital to begin tracking the improvements in cost of verifiability over time.
To be able to track high level trends, we need to find metrics for cost of verifiability that are
✦ Simple
✦ Easy to track
✦ Indicative of underlying progress over time
zkBench did some great work here already, with the below benchmarks.
In their ZK endgame article, @gevulot_network specifically propose two key metrics. 👇
📊 Field Size
A key element of of cryptographic proof systems like SNARKs & STARKs is the field size of the underlying mathematical field utilized in the proofs, which is closely tied to performance improvements (smaller field = better for performance).
📊 Verification Cost
Another crucial metric of proof systems is the cost of verification on Ethereum, which significantly influences the scalability and practical use of blockchain applications.
However, an in-depth analysis should also consider:
✦ Proof size
✦ Memory usage
✦ Hardware requirements
✦ Prover and verifier time
As well as the impact of these factors on overall system performance.
What are your takes on this? Please share in the comments! 💬
🛣 The Road Ahead
What will take us there is further improvements in ZK hardware acceleration and more efficient proof systems.
Proof systems like plonky3, that work with smaller fields and leverage things like FRI.
Plonky3 for example features 31-bit fields, continuing a trend toward efficiency since the Goldilocks field was discovered.
But things are moving at the speed of light.
The most recent proof system, Binius, already uses a 1-bit field size!
On the verification side of things, Plonky3 aims to support some polynomial commitment schemes such as Brakedown, optimized data structures and parallel computation.
However, despite these advancements, the verification costs for Plonky2 & Plonky3 (without EIP-4844) remain relatively high compared to Plonk.
But over time, this will change, and such systems designed to adjust its proof sizes based on the requirement (pioneered by plonky2) will take the lead.
Okay cool, but what is @gevulot_network actually building?
Gevulot is building the zkCloud, a decentralized network optimized for offloading arbitrary provable compute.
The universal proving layer of the modular stack.
🤔 Why does this matter?
While hardware acceleration and more efficient proof systems are key to bring ZK to use cases everywhere, it's not all that is needed.
What we also need, is a modularization of the ZK tech stack.
Think:
✦ Proving Layer: @gevulot_network
✦ Verification Layer: @horizenglobal
✦ Messaging Layer: @Polymer_Labs
Easy plug-and-play infrastructure, ranging from proving layers to verification layers that can work in tandem with battle-tested, decentralized networks like Ethereum.
This will allow us to merge the power of verifiable computation and decentralization, defining a new era of innovation in distributed systems.
Stay tuned for my @gevulot_network deep dive later today! 👀
3 images on X
What came next · since 23 May 2024
Since then
Zero has written 1 more piece on Gevulot since this one.