Jolt: An open source zkVM by a16z crypto research and engineering - Berkeley 2024

[Music] uh thanks everybody for coming today I'll be talking about jolt and open source project by the a16z crypto research and Engineering teams built with the intention of unblocking innovation in decentralized systems I hope to use jolt as a tangible example of how we a16z crypto research and Engineering think about our role in an innovation ecosystem open source software more generally and future models for funding development decentralized systems are one that push the authority and validation Logics to the edges the most common examples of these are blockchain networks such as Bitcoin ethereum and salana uh as a sweeping statement of the resulting properties you get computers or networks which can make commitments no one entity has power over the other participants in the network now or at any point in the future this means new businesses and networks can be built at top these networks without concern that the rules will change uh when it Best Suits an intermediary so for a while since 2016 snarks have been touted as the scaling solution for blockchain networks these are succinct non-interactive arguments of knowledge I'm going to go slowly through this because uh you kind of need to understand this context to understand any of the rest of the stuff we're talking about um so a snark allows a prover to run a program and create a proof that it ran that that program executed it correctly and they then create a receipt or a proof that they can give to a verifier and the verifier can check that that proof is correct uh in significantly less work than running the original program um in some sense you can see this as compressing the original program um so they they execute a program they compress it and they can provide it to a verifier who can check that in a lot less work than originally um in the case of a blockchain network this allows the prover to compress that workload before it reaches the edges of the network um where validators check all of the work of the intermediaries okay so finally I can talk about jolt so in terms of jolts potential um first you remember the issue esoteric dsls we have to program in that crazy programming language uh jolt is a zkm that means you take a normal programming language like rust you target risk five which is a assembly instruction set and then you can run it on jolt the second issue was that there was this 10 to the8 prover overhead um jolt brings this down to 10 to the 6 with a potential upside of 10 to the 5 that makes it a lot more practical to run this thing it's it's not very very expensive to create these uh succinct proofs one of the biggest issues is that these things are incomprehensible you write these crazy arithmetic circuits and then no one really knows if they're right um jolt has a much simpler programming model where you write basically two functions and you get all the programmability uh we've found that people are better at reading these things um but perhaps most importantly this is a fully new snark approach um this means that has potentially more room to grow than the existing approaches um and more juice to be s squeeze so the left side here is these Starks that I was talking about this is where much of the R&D has gone historically um which means that there's been minimal R&D and minimal iterative efforts to improve these things unfortunately um it's very expensive for a startup to try to go to a different land the multivariant land um based on some Theory cooked up by four academics there's no experimental section in these papers um because it's very hard to build this stuff and because it goes back backwards in the lineage much of the research and Engineering work would have to be redone from scratch you can't just pick the stuff up off the shelf so for a startup or business entity there's uh to maybe too many risks to to start from scratch um there there are engineering research product and Personnel challenges that might be too much to overcome but we as a venture firm uh see these kinds of asymmetric opportunities all the time and we're very comfortable with them the theory went that if jolt works it has massive potential to accelerate all decentralized systems jolt is fully open source uh it always has been and always will be uh many in the permissionless um Network ecosystem are thinking about new architectures to be built a top zkv systems like jolt and we hope many more will come um we hope many of those people that build a top jolt uh take some time to learn jolt and give back to the code base itself so that everybody can grow um okay so zooming out a little bit um I wanted to give like a little bit of uh intuition on where jolt or zkv Ms like it should be applied um the first is when the prover overhead is less than or equal to the number of validators so in the case of ethereum we have 600,000 nodes that run the exact same computation um if you can bring the prover overhead below 600,000 X then all of a sudden it makes sense for somebody to pre- compress this work before it's sent to 600,000 people uh another case is where two parties that want to have a trustless relationship in other words they check all of each other's work um have asymmetric compute costs so a common case for this is when you want to uh verify something on chain running things on chain is extremely expensive if anybody has done a blockchain transaction you know you pay a lot for that very small amount of compute and so you can think of onchain as a compute environment with incredibly High compute costs but you you can do some work offchain in a data center where it's incredibly cheap so very asymmetric compute costs succinct VMS make sense so in our eyes um there was obvious demand for ZK VMS that is block space but we're also willing to bet that ZK VMS will be applied more broadly uh ZK is the closest thing I've seen to algorithmic Magic in my career uh in the early 2000s TLS or transport layer security https or just the standard for encryption on the internet uh brought the internet to a new era era where writing sensitive details back to servers was possible for normal application developers with little cryptography experience we take this for granted today but uh web 2 entities didn't big Tech entities didn't they use the last 20 years to become the most powerful entities in the modern world they are fundamental to Modern democracies in Broad Strokes Society trusts the web to monopolies but there's reasons to believe that these entities shouldn't have a strangle hold on Innovation We Believe tools like jolt will increase the practicality of decentralized Alternatives so we take bets on technology which can lighten the burdens ship the burden of Entrepreneurship ship We Believe deeply that decentralized and trustless systems will rearchitecturing value to their users this should be the bread and butter of any company jolt is an idealized example of what we targeted a16z crypto engineering core tooling which has leverage over the space and on its own is not a business Innovation needs this tooling but it's unclear if there's an incentive compatible party to build it for us a rising Tide Rises All Ships we are the incentive compatible party for Innovation okay cool thank you appreciate you guys listening