🪄 Introducing Modulus: Bring AI On-Chain
“We are the music makers and we are the dreamers of dreams”
We’re ecstatic to announce our Seed round today, marking the formal debut of Modulus: the Accountable Magic Company.
We’re helping dApp developers access tamper-proof AI— at massive scale — with the world’s only specialized ZK prover for AI. This means that smart contracts can now enjoy powerful AI features without breaking the trustless creed.
In other words — Accountable AI for Magical dApps — the biggest upgrade ever to smart contracts.
We’re backed by inspired investors like Variant and 1kx, as well as Floodgate, Inflection.xyz, Alliance, Bankless, Stanford, GCR and others. Additionally, we’re joined by mighty angels from the Ethereum Foundation, Worldcoin, Polygon, Celestia, Solana, Microsoft, Gensyn.ai, and many more.
Stay tuned over the next few months as we:
- Debut the largest AI applications ever implemented on Ethereum via partnerships with Upshot, Worldcoin, Ion Protocol and others
- Open-source our prover, “Remainder” — the world’s only specialized ZK prover purpose-built for AI, and
- Roll-out the Modulus API, giving every dApp access to powerful, expressive AI with ZK accountability
In fact, we’re opening the waitlist to our “On-Chain AI” API today, which will start with support for powerful decision forests.
In the meantime, follow us on Twitter, check out our swanky new website, and read on for the juicy story of how we ended up here.
TL;DR of Today’s Announcement
- đź”® The Problem: today, for dApps to access powerful AI features, they need to abandon decentralized security and assume centralization risk
- 🔮 The Solution: Modulus uses zero-knowledge proofs to efficiently check that AI providers didn’t manipulate their algorithms — on-chain
- đź”® The Insight: specialized ZK proving purpose-built for AI delivers wildly powerful AI features for dApps at a fraction of the cost
- đź”® The Result: dApps enhanced by tamper-proof AI leapfrog their peers in UX without betraying the trustless creed of the chain
- đź”® The Vision: as advanced AI gets introduced to our legal, financial, medical, security and education sectors, cryptographic accountability will safeguard the backbone of our computational future.
“AI Forces a Faustian Bargain”
AI is pretty damn magical.
Each time we pester our AI copilot for advice, queue up our latest credit score, or refresh our social media feed of choice, a supercomputer somewhere runs through enough mathematical operations to make Turing himself dance in glee.
AI is going to radically transform our society. And yet, this modern day superpower is entirely off-limits to anyone building on blockchain technology. The reason is simple: AI operations are too computationally expensive to run directly on-chain. This means that to bring AI features into the smart contract, we’d need to trust AI operators to always run models without manipulation.
Let’s just say that, in the context of crypto, trust is nearly always a very bad idea.
đź”® The Problem: for dApps to access powerful AI features, they need to abandon decentralized security and assume centralization risk
“ZK is Accountability End-Game”
Thankfully, we needn’t despair! To bring the magic of AI to the blockchain without throwing away decentralized security, we rely only on the verifiable magic of ZK.
As applied to AI, zero-knowledge cryptography allows us to verify that an AI model was executed “correctly”, without revealing the secret sauce behind the model. In other words, we use math to check that AI results were not secretly manipulated. By verifying this “proof of correctness”, on-chain, we get blockchain-equivalent security at a fraction of the cost.
Thanks to math, we can now know — with cryptographic certainty — that your AI copilot didn’t supply secretly doctored code, that your credit score wasn’t nudged by a prejudiced bank, and that your social media feed hasn’t taken on a political bias. We call this marriage Accountable AI.
🔮 The Solution: Modulus uses zero-knowledge proofs to efficiently check that AI providers didn’t manipulate their algorithms
“Specialization Always Wins”
That’s all fine and dandy, but bleeding-edge ZK cryptography doesn’t come for free. In fact, even though it saves money by naive blockchain standards, generalist ZK provers still add >1,000x more costs to AI compute (over just trusting our service providers to do the right thing).
In fact, in the past year of building the world’s 1st ZKML applications, we’ve continued to bump into the same cost+performance ceiling. There’s simply no way around it: ZK + ML == economic violence.
That is, until we noticed that modern AI is wholly predicated on the advance of highly parallel processors (read: GPUs). If we create the same step change in ZK proving, we’d be massively opening up the design space for smart contracts.
Or, to put it another way, instead of counting on a faster all-terrain bus, ride our maglev bullet train (codename: “Remainder”) — we have champagne on board ;)
đź”® The Insight: specialized ZK proving purpose-built for AI delivers wildly powerful AI features for dApps at a fraction of the cost
“Accountable AI is a Superpower”
Empowered by “Remainder,” we’ve already started to build out a community of developers and partners obsessed with the next generation of AI-enhanced dApps (wanna join the fun?). In particular, we’ve been consumed with the task of bringing powerful AI features to concrete use-cases — at massive scale.
To help us achieve that vision, we’ll be debuting multiple high-octane ZKML applications over the next few months, starting with a collaboration with NFT appraisal leader — Upshot.
Upshot uses AI to deliver extremely accurate, timely NFT prices (upwards of 100+ millions appraisals per hour at 3–10% MAPE). This means the opportunity to open new financial markets for long-tail assets: awesome! And yet, despite this exciting potential, subscribing to Upshot’s price feed today also means trusting Upshot to never tamper with their pipeline.
The holy grail, of course, would be to deliver Upshot’s astonishing AI prices to the chain without giving up on blockchain security. Which is precisely what we’ll be providing by integrating their machine learning oracle into the Modulus API.
The result is an NFT price feed forged with AI superpowers AND blockchain security, a marriage uniquely enabled by Modulus’ specialized ZK proving.
From tamper-proof AI price feeds to authentic machine artists, dynamically intelligent DeFi to trustworthy game economies — there is so much to come. Not to mention the forthcoming paper, the open-sourcing of our codebase, as well as a couple surprises along the way ;)
đź”® The Result: dApps enhanced by tamper-proof AI leapfrog their peers in UX without betraying the trustless creed of the chain
“The Secret Sauce is the Team”
Coming from our experience in academia, we knew Modulus needed to be the kind of company where integrity of character is never compromised. And from the folks we call “Moodlers” here in the serene Stanford hills to our investors, clients, and partners — I think we’ve knocked it out of the park.
In particular, leading our Seed round: Jesse, Alana, and Geoff at Variant, Dmitriy, as well as Justin and Wei at 1kx. Thank you for helping us focus on true value creation, and the good work of impact-oriented technology. Additionally, massive shoutout to Ann Miura-Ko from Floodgate and Tyler Whittle from the Taproot Wizards for being our earliest supporters. And rounding out our superhero roster are the spectacularly talented investors of Inflection.xyz, Alliance, Stanford, and GCR.
As well, we have been the fortunate beneficiary of Zhang Zhenfei, Riad Wahby, and Rand Hindi’s galactic-class advisorship. Seriously, they are the best in the biz, and we’re pinching ourselves over just how lucky we are to be in their orbit.
Finally, joining our cap table, a truly mighty roster of angels. Including… Daniel Lubarov, John Adler, Harry Grieve, Ben Fielding, Remco Bloemen, dcbuilder, Georgios Vlachos, Nick Emmons, Brandon Da Silva, Wei Xie, Will Papper, Nicola Greco, Robert Drost, Cathie So, Don Ho, Akilesh Potti, Casey Caruso, Stuti Pandey, Spencer Noon, James Wu, Vassilis Tziokas, Santiago Santos and more.
đź”® The Who: Integrity of character was the single most important factor we prioritized in how we built Modulus.
“Do It with Joy”
When Ryan, Nick and I started Modulus just over a year ago, we had a genius plan:
“Build something really, really hard”
The heights of our strategic sophistication is nothing short of awe-inspiring! But there is a method to our madness.
The thinking was that if we worked on something people thought was nearly impossible, but do so with absolute sincerity and earnestness — we’d attract the most talented, nicest people in the world to work along with. All they’d ask in turn, we knew, would be to build the kind of company that was sober, spirited, and most importantly, joyous.
Good news: so far so good.
Better news: it might just change the world.
đź”® The Vision: as advanced AI gets introduced to our legal, financial, medical, security and education sectors, cryptographic accountability will safeguard the backbone of our computational future.
Interested in adding AI to your dApp? Fill out this snazzy form and join our community! And hey, for more ZKML projects, papers, and announcements, you’re already in the right place.
”Welcome to a world of pure imagination…
