STARKs: Facts, Not Fiction

Transcript

okay uh hello everyone uh i am ellie ben sasson i am co-founder and president of starkware and i want to talk to you over the next 10 minutes about facts not fiction so we all know that crypto twitter is very good at being enthusiastic about certain things for instance something like lightning network and it is hailed as being the solution that's going to solve all of blockchain scale problems and this is just one example of an article that came out back in the day um but then what happened within half a year until this day well you know actually scaling didn't wasn't actually solved by lightning in spite of what crypto twitter has said so

there's a lot of enthusiasm about certain things and there's a lot of um you know misperceptions and what i want to do in these 10 minutes is talk a little bit about some of the perceptions that are not quite right about uh our core technology which is uh zk starks and about what it can do for you today so let's start and um you know so here's a piece of fiction um a statement that says that zk stark technology well the prover is too slow in reality it's the opposite the zk stark prover is the fastest approver of all proof systems out there it is 20x faster or more than all other proof systems out there when you measure head to head here's another fiction the zk

stock verification time is too large and that's why you can't use it well in reality the zk start verifier is just as fast or actually faster than all of the other alternatives so even when you go to very large computations the amount of time needed to verify a proof is on the order of you know three to four milliseconds and this is actually without any deep optimizations here's another piece of fiction well verifying a stark is just too expensive in gas and this is partially correct you should know that among zero knowledge proof systems starks are the only ones that actually don't have any dedicated you know pre-compile or subsidy on

ethereum while other proof systems do have it so a single start indeed in gas cost is roughly 10x more expensive but however if you look at amortized gas cost per transaction and you measure it over things that are running in production then you will see that actually um the amortized gas cost which is all that matters is lower significantly lower than all the other competitors out there and what you have here is the gas cost not in thousands of gas not in kilo gas but in actual gas and for instance in production this is the amortized gas cost across all of our systems we have four systems uh already running in production and for validium which means

that the data is off chain you pay less than 1300 gas per transaction this is in production including all the auxiliary data it's not just the the gas that's related to the proof loop ring which uses a roth 16 snark is on the same order of magnitude and the lowest amortized gas that we actually reached in production was 17 gas 17 gas per transaction this was last year as part of the reddit bake off so in contrast to what the fiction says the stark amortized uh gas cost is the lowest across all systems in production now for you know optimism arbitrary planck systems i don't know the exact numbers of total amortized gas costs but

i don't think they're going to be any better they're probably going to be worse than what you get with starks okay now this is this was the state of affairs up to you know roughly one week ago and now we launched another great star core innovation sharp which is shared proving which means that we take all of our applications and put a single proof that advances all of them so what you're gonna get is even lower amortized gas costs and this is just now started so we'll see the gas cost going down even further for a systems in production and uh here's another example of something that happened so about half a year ago sorer which is one of our uh partners uh before it launched with us it was paying 472 eth over a week that amounted at

that time to 1.5 million dollars a week in gas alone and now when it's running with us it is paying 70th a week so down from 1.5 million dollars spent on gas for a system in production with our system it has gone down to seven ether a week one ether per day so these are systems in production all of them using our technology another example about the scaling of stark and rl2 no other scaling solution reaches these things and these are all metrics that have been you know that are operating out there in production we already settled by now it's close to 12 billion dollars of cumulative trading we

settled over 10 million transactions over multiple different applications we serviced more than a hundred thousand users we minted seven million nfts the largest mint we did was 600 000 nfts in production this was done for immutable x uh done a few months ago inside a single proof and we have four customers all of them operating and we've launched already one year ago now here's another piece of fiction zika starks rely on new unsafe assumptions you better worry you if you want to use them in reality it is the exactly the opposite the snark systems be they graph 16 or plonk they are the ones that are uh susceptible to quantum computers they

require trusted setup or toxic waste and they rely on new assumptions that are pretty new and very little um economic uh activity has been based on them so they're not as safe as the stark that is post quantum secure has is transparent that's the t in the acronym stark no trusted set up no toxic waste and the security assumption essentially is the existence of a hash function which goes back to the 70s and all of e-commerce is also based on this one assumption so the point here is if you're building infrastructure for trillions of dollars of value and you want it to be used for decades henceforth you want it to be future proof so you're better off

working with the stark than with other zero knowledge proof technologies another piece of fiction i can't build a start-based system it's moon math you know only the wizards at starkware or some other teams might do it well again this is not the reality we have extensive tooling so you can scan this qr code right now and go into cairoland.org start downloading there's a playground there's a bunch of tutorials this is a programming language in which all of our systems were written the ones servicing you know billions of dollars of volume they were all written in this language and you can use it today it's securing complete language extremely efficient and you just go there and start programming it comes with extensive tooling there's a

compiler virtual machine id extensions tracers application code so you can use all of that today it's and all you need to know is programming that's it here's another piece of fiction starks are the long-term solution we acknowledge that but there's still three to five years out well i just showed you that the most volume and number of transactions across all layer twos right now in production are actually done over our technology so are they are the most this technology is the most ripe today but more than that starknet is already out there in alpha mode it already supports general computation l1 to l2 interaction permissionless deployment of

transactions and smart contracts which you can do today and there's another qr code that you can go there and start working with it and composability will likely arrive this week so you will also have that and it will reach mainnet before the end of this year and it will be much more performant and much earlier to market than all the other l2s that are being worked on so another thing is that this is an ecosystem that is being built not just by us but by a large and growing uh team of other parties we have the turbo guest team with alexey kunov and others building the full node we have a project um building a compiler from ethereum

virtual machine we have block explorers by nethermine there are functional language compilers and contracts standardization by open zeppelin and this is just the tasting so this is now something you can join you are invited to join it and basically i hope that all of this sort of facts will convince you that if you're considering scaling and you want a future proof system that is operational now and will save your users millions of dollars you want to explore and start using stark net and the stark based systems so that's basically my very short speech thank you very much