Navigating AI Infrastructure and the Post-Quantum World with Aptos feat. CEO Avery Ching

GM GM, welcome to what's going to be another fantastic episode of DevNTell. So if you didn't know, DevNTell is a 30-minute podcast held every week allowing founders, hackers, and anyone in between the opportunity to come on the show and showcase what they built. And today, I'm really ecstatic to welcome Avery Ching, who is the co-founder and CEO of Aptos Labs. So if you're not familiar, Aptos Labs is a premier Web3 studio of engineers, researchers, strategists, designers, and dreamers building on the Aptos blockchain. So if you stick around for today's episode, you'll get to meet Avery, hear an engaging discussion around what infrastructure AI agents need to transact autonomously, the post-quantum future of Aptos, and how you can get started building with them today. All right, let's get into it.

GM GM, welcome to the show, Avery. I'm ecstatic to have you on today, man.

I'm enjoying this as well. Happy Friday!

Happy Friday indeed. Happy Friday indeed. Another Friday, another DevNTell, nothing I like more. Yeah man, I guess before we get into all the great content we have today around AI agents, Aptos, and the post-quantum world, would you just like to give an introduction about yourself?

Sure. I'll just give a brief introduction. Avery Ching, co-founder and CEO, Aptos Labs. Prior to this, I spent a decade at Meta working on data infrastructure, ML at the time before it was called AI, and then what eventually became the tech lead for the Libra project and Diem project, which was Facebook's ambitious effort to ship a payments network and stablecoins across the world with Visa, Mastercard, Uber, Lyft, and many others. I love distributed systems, I love blockchain-based technology, AI, financial infrastructure, all kinds of things like that.

Awesome man. Yeah, I was looking into kind of the Diem project itself. There's a lot of folks there that either went to Aptos or started their own blockchain projects. I think Monad was another one, there might be another one if I'm not mistaken. But yeah, just curious to kind of double click into that a little bit, just curious to kind of get your perspective on those experiences and how they kind of inspired you to co-found Aptos.

Yeah, so I think it was mostly a necessity issue. In our case, when we were working on Libra and Diem, we had assembled this amazing team, we had built a great technology stack, and then I think we ran into hurdles when it came to the state of the industry, regulators, and even the government. So doing outside of Facebook was like the next logical conclusion. And even today, we've seen the administration change has been very supportive of blockchain-based technology, crypto-based technology. And I think now we're in an inflection point in which the industry is really, really moving fast, and from a product adoption standpoint, it's less about crypto tokens as interesting, and more about the crypto technology as being very useful for different types of applications. But those we believe in both the market section and the machine section, so both financial markets and also AI. And we've been focused on this for well over a year now, and excited to see these products starting to take off, grow very rapidly, and find the partners to work with.

Absolutely. Yeah, when I first started in the Web3 crypto scene, it was in I think late '21, and then was like the NFT craze and everyone was on the speculative side of the chain. There was that little percentage of folks that were actually building meaningful technology. And I guess in between that, the last bear market, or maybe two bear markets because I don't know how many we've been in, it's been flipping back and forth pretty quick. But like now I kind of see somewhat of a stability and the products that were actually built to last are showing and shining through. And Aptos looks like it's part of that. Just curious for our audience that may not be familiar with Aptos, could you give a high-level overview of what it is and why it exists?

Absolutely. So at the time we were thinking about building Libra and Diem, I guess we looked at the technology stacks available. There was Bitcoin, there was Ethereum, and I think they all showed very good promise. They kind of were the similar things as what we compare today as initial products in AI, but there wasn't a technology that made it applicable. None of those products could be used for mass adoption, for mass scale payments, or trading, or any of those things like that. And so we set out to build a technology stack that would be very scalable, the same way we thought about principles of building at Facebook. It's always about how do we kind of support a big user base or support a big usage where transactions are happening millions of times a second, or the latency of operations where the ability to move assets around very quickly would be very fast. And maybe the last step that's also not the least is from a security perspective. These ledgers are securing billions, ideally trillions of dollars of value over time. This thing has to be rock solid and the most secure system ever built. And so with those things in mind, we created Move, a new language, we created the blockchain piece which is highly performant. And if you fast forward from 2018 when we started working on Libra to now, eight years later, you've seen that we have the stack that has the fastest block times at 25 milliseconds, we can scale in production like no one else can, we have the lowest fees in the market, so anytime a stablecoin comes on board to Aptos, you're really getting the best version of that product available. And then from a security perspective, you have the work that Aptos has done with Move to have three levels of security: the static analysis that every smart contract platform has including EVM, but then we have a dynamic runtime check which is always running every time the code is running to make sure invariants aren't violated and protecting unauthorized money minting or other bugs that could happen in the process, and you have also the Move Prover which is the third level of formal verification that ensures your code is running safe. We feel like this, especially in the age of AI, is such a crucial technology stack that you need to support a huge financial ledger.

Yeah, yeah, absolutely. And as I also understand, you guys have prepared yourselves well for the post-quantum side of the world. Not a lot of people are... well, there are some people talking about it on the timeline, but it's not as big as some of the other hotter topics like privacy and stablecoins and what have you. Just very keen to learn how you guys have done your homework and prepared yourselves for post-quantum realities.

It was great to hear, I think Coinbase called out a couple of chains that were more prepared for quantum than others a couple of months ago, and Aptos was named as definitely one of the most prepared. It just has to do with a cryptography team which is always on the ball and thinking ahead of the curve. It's not just post-quantum, but it's also the confidential assets, encrypted mempool technologies we've built. It flows from having a lot of the primitives that we've also created over the years, and having the thoughtfulness around what they could do, and then the infrastructure itself is well set up to support that quick flip once post-quantum is ready. I think we don't hear about it as much because quantum is exciting, but it's also years out, and there's many things that happen between now and those years. But I think the other thing that's important is once it happens, that curve flips so quickly. If you're not prepared, you end up with your users losing funds, accounts being shut down, or all kinds of bad scenarios popping up. Our idea has always been very practical: protecting user funds, protecting user transferability, and the ability for user to opt into quantum readiness whenever they feel like it is a huge easy win, and yet it doesn't burden the users today, because if you turn on quantum protection too early, you're paying this cost. Quantum-resistant cryptography is extremely expensive compared to normal cryptography, so pay the cost when you need to, but we're ready and prepared for that.

Excellent. Yeah, I mean it's going to be creeping up faster than we expect, but we'll see. 2026, everything changes every couple of weeks, so it's quite an interesting time to be alive indeed. And I guess you summarized it in your last answer there, but for companies that maybe aren't taking it as seriously now, why do you think that is? Do you think it's like 'tomorrow's problem, I'm worrying about today'?

Yeah, probably one of two reasons, right? Either they don't have the experience in-house to worry about it, or the belief that it's going to take longer. We don't try to be super precise on the timeline, years away, but the point is you don't want to wait years... it might take some time to develop and integrate that technology into a blockchain. I don't think anyone should underestimate the work it takes to securely deploy new cryptography methods, test them, check the performance of how that operates, right? People often think correctness is the only thing, but if it degrades your system performance down to a level where it's unusable, that's also a security risk and an availability threat. So those things are important to us and we want to make sure our users and partners are well prepared years in advance, which is why we've taken the approach we have.

Excellent, love to hear that. Switching gears a little bit now... post-quantum is coming eventually, AI agents though are here now, and they're growing rapidly in popularity and use case and whatnot. Personally, I feel the same kind of nerdy excitement as I first did when I started reading up on blockchain. I just think the technology is really cool, there's much broader applications here in the AI spectrum, and it seems to be growing and evolving much more rapidly due to the nature of AI itself here. Just curious to get your view of the landscape of AI agents and how they will be playing with Aptos and integrating there?

First of all, I feel exactly the same way you do. Many late nights spent on AI, and I think if you're not doing that, we're all in the wrong business, right? And it's not just about AI, every business needs to think about AI very seriously because it affects every part of the world, honestly, even ourselves to such a large degree. That being said, we've been really thinking about the intersection of AI and crypto for a long time. We started off building this project called Shelby, which we launched with Jump Crypto, as a global object store that is performant and supports especially AI use cases really well. Feeling much ahead of the curve when it comes to the idea that AI compute resources are scarce. Anyone who tries to get access to H200s, B200s knows how difficult it is to get long-term leases and access to machinery. When you get it, you need to grab onto it. Those compute pieces will be available all over the world, and you can't be particular like 'I need all my stuff in one data center'. That's just not the reality of where things are today. So Shelby, as a way to support the ability to work anywhere in the world with your AI inference loading, it's going to be supporting generative AI, generative video content, training data, bringing that training data anywhere in the world, as well as data collection use cases. So all that available, Shelby is definitely a first step for us in the AI space, and it's a big one to build that system. It's extremely complex, extremely stress-tested AI, and can support our global AI needs because today AI is very local, right? You and I were just talking off the podcast earlier about the fact that the latest model can run almost Opus 4.6 quality on a Spark sitting next to you, right? That's the frontier model from just six months ago working under your desk. That's an crazy amount of innovation in a very short period of time. So that ability to then have your data flowing from place to place to place is incredibly important, because compute is a great stateless resource, it can work anywhere in the world, but storage, your data is stateful, and it has to physically move the bytes from A to B to C extremely efficiently and also have the APIs and access that every dev requires for their use cases. So that's a starting thought process around how crypto technology, AI, and the idea of Shelby being built on top of a very fast blockchain like Aptos, a very fast and reasonably cost-effective blockchain, proves out a great use case when it comes to AI as well.

Absolutely. And from what I understand, you guys have made quite a substantial investment in building out Shelby and another product that could very well complement it in Decibel. Did you want to touch on Decibel for folks who might not be as familiar?

So Decibel is this idea that a lot of systems have tried to build DEXes, but these DEXes, the 'D' is a little silent to me when you think about it because they're not super decentralized, whether they're not open-source code, or they have centralized matching components or other centralized components that make it very challenging even from an availability standpoint, say they run on an L2 that's a single machine. Those don't feel very decentralized when it comes to building exchange infrastructure. So we took the challenge on to say, if we build from scratch, blockchains are meant for asset transfer, we have this amazing smart contracts platform Move, which needed a little bit of work to be moved into a trading-oriented VM, which we've done already and made it extremely optimized for that process, can we build out a decentralized exchange order book matching engine completely on-chain? And the answer is absolutely. And can it match centralized exchanges for crypto? Absolutely. Today's fastest centralized exchanges work on the millisecond level, could be 10ms, 50ms, 100ms, and we've been able to prove that Aptos can run at 25ms block times reliably, consistently, with high availability, four 9s easily, and it's all running on top of Move: the matching engines, the orders, everything, from perps DEXes to now spots, which we launched recently. And so it is possible, and it's live and testable today, and Decibel's been the most transparent system when it comes to DEX volumes and DEX traffic. This last week was a big move for crypto, we've seen that reflected in the Decibel workload, we've done about $6 billion in perps volume in the five or six months since launch. So it's a product that we continue to evolve and grow, and proves out the point that blockchains can be used for fairly high frequency trading operations.

Absolutely, absolutely. And kind of the reality now, as far as I can tell, is everybody's kind of thinking about human-to-machine. So a very average AI user is only probably familiar with ChatGPT and uses it to fix their emails or come up with some template for something, look up recipes. But I think both of us know this is going to scale to machine-to-machine. A lot of people don't realize how exponentially fast that user base of the internet will grow. When we say machine, not only is that an LLM or an agent running on the internet, robotics is an emerging thing too. Yesterday everyone was hyped on the Microduck or... I hope I didn't mispronounce the name. But yeah, robotics is an emerging vertical in this whole equation as well. Just curious to hear your thoughts of how you think existing infrastructure for cloud and payment rails kind of sits now, and whether or not that will scale and what we need to have it scale if not.

So you covered a lot. I mean, AI as a field is transforming so quickly. In many ways we felt it the most, you and I, for coding, that's the biggest use case of AI today is the ability to code with amplifying yourself 5, 6, 7, as many agents as you can get working productively on your behalf. I think we've all seen the benefits personally as well as corporately, being able to move faster, seeing ideas prototyped before instead of having to wait for those to materialize in other ways. That's going to happen across every industry, and crypto is no exception. We already are building pretty intricate systems internally within Aptos to provide more layers of defense, having security scans that go in every PR review, before release doing deeper scans, variant analysis, so many different kinds of tools that are available and can run cheaply with especially the newer open-source/open-weight models. Those things are transformative. But when it comes to... robotics I think is an amazing area. I have a limited number of robots operating in my home today, I've got one vacuum cleaner which I love, but I think you can see the future is that these robots will become incredibly powerful, powered by inference, powered by immense amounts of training data, it's going to be amazing in terms of that use case. But I think what really concerns us here is the confidentiality of this data, right? So now I've got an AI agent in my home consistently looking around, hearing conversations, what happens to that data going forward? What actions is that agent taking on my behalf? I think these are the areas where crypto technology, as well as blockchain-based technology, we see a tremendous amount of opportunity. For agent actions, I think you've heard about the attacks on Hugging Face from AI, agents are going to need to log what they're doing and why they're doing these actions, I think it's very, very clear. Not just that, even for inference, right? Thinking about how people use that data for training has to be logged very clearly. An immutable ledger is what's necessary here. And I think we all know what an immutable ledger can mean in our space, right? Something that you can never change, something that's available worldwide, something global, that product is a blockchain. That blockchain technology is perfect for this use case. And so having something extremely fast, extremely cheap, extremely reliable, and something that you know cannot be corrupted over time to then bring up any kind of legal remedy for any kind of challenge that had happened is the perfect solution for these things. But not just legal, right? It also helps the companies who are trying to do good in this space do the good things. Assuming some of these frontier labs don't want to use your data for training if you don't want them to, they can look at the on-chain receipts and know like 'okay yeah, make sure this is not used in the next version of Fable 6' for example. So it comes across the board, and we're very excited to see confidential AI being a big thing going forward into the future, for robotics especially.

Yeah, 100%. Yeah, I always am keen to jump on these things, and robotics is really cool, but just the fact that all these things are equipped with big cameras staring into you and everything you do... even Teslas, right, it's basically an entire vehicle with cameras, and the self-driving thing as well: 'do I really want to trust an AI completely to keep me alive while I'm on the road going 100 miles an hour on the highway or whatever?' But yeah, I think privacy is a big thing. Obviously coming from blockchain space, now stablecoins are a hot topic with all the regulation settling. Regulatory stuff aside, do you think blockchains in general can scale to the demand that all these agents will bring to it in terms of wanting to transact online?

I think it really depends on the technology stack, obviously, but I know Aptos can support massive, massive scale, I think that's not an issue. I think the test for us is today, what agents will do, probably the first step is going to be to start to interact with existing websites and APIs for moving money around, right? Giving an agent your credentials to make a restaurant reservation on your behalf or make a purchase on Amazon I think is a more logical first step. But then there'll be other activities that will make more sense, and I think we talked also a little bit about those off-camera, which is like suppose I want to let you use some of my GPU capabilities in my RTX setup at home while I'm not using it, right? And I want to broker an agreement of how much I charge for that, right? Having an agent negotiate that and pay that on a crypto rail with a stablecoin is probably the easiest thing we can do as long as we can get money in and out of the system very fast. And that's going to be a very natural evolution of new use cases that pop up that we haven't yet imagined. And also an evolution from today's world where now agents start to make my reservation on my behalf or purchasing on Amazon on my behalf, but then starts to do other things like negotiate payment for something, maybe negotiate like a next-generation Craigslist, starts to negotiate where do we meet up, where do we exchange the good, what's the price, and how do we escrow that until the transaction is completed. So all those things become new possibilities once we have a plethora of agents working on our behalf. Today you and I might have five or six agents coding on our behalf, but we don't have those agents operating on other areas of our lives when it comes to personal, corporate, organizational infrastructure, finances, those kind of things, just not there yet in terms of technology. So once that happens, I think we'll be in a good spot for much more adoption on the blockchain-based rails.

Indeed, indeed. And yeah, I don't know if we'll get there this year. I might be proven wrong. We're six months or eight months into the year, people are saying we've reached Singularity, we have AGI. So yeah, we'll see, things are moving quite quickly. But in terms of... you kind of mentioned it around the APIs and whatnot, we are a developer show, just curious to get your perspective on where the developers who are watching the show today, where they can get started with Aptos and any APIs or particular SDKs you might want to mention?

I mean, Aptos has a ton of SDKs available to interact with, from building in Move, there's great tooling available, there's AI coding tools available for building Move. There's SDKs for TS/JS, which is the easiest way to interface with the blockchain itself if you just want to use existing products. And then we also have MCPI access for a lot of the things that are happening on-chain, so does... so as well for Decibel, as well for Shelby when it launches very soon. And so I would say fire up your favorite coding tool, point it at the Aptos documentation, just have it go to work. Whatever you want to build should be buildable. If not, report it back to us. I've been able to build many things against the Aptos blockchain just using the skills and the LLMs are very well trained on Move today, apparently. They've scanned all the code available and done a great job of understanding how to develop Move code. So yeah, I'd say that's probably one of the easiest ways to get started is just fire up your favorite terminal and get to work. And I think that just proves the point, right? Because when you see yourself creating an application in a couple of minutes without having to read a single doc, having to just express your idea like 'I want to create a new payments platform and I have a neobank set up now' with the core tenets of a neobank set up in like an hour, and I've got a basic UI working, all these things show you how the world has now changed to be so much opportunity to build anything you want. With the power of now cheaper inference coming available... and that, I think the platforms, the improvements have been slower in terms of model quality, they're not as quickly improving anymore in terms of frontier models, but what we have seen is a grouping at the top in terms of the abilities of what they can do for coding, reasoning, and other things, and now the price is coming down to fractions of where it was at the beginning of the year. So that is I think the biggest transformative act. That means cheaper inference available for everybody, and cheap inference means now agents will be proliferating throughout our lives because cheaper inference is available, and now that means privacy and confidentiality around those assets and the ability for ledgers to track what's happening in our lives and why those things are happening and are those things we want to have happen becomes so much more important, and it goes back to why blockchain is going to be so important here.

Well said, and I couldn't agree more. And as we're coming into time, just curious for our aspiring founders and builders watching today, would you be able to offer any sort of advice or lessons you'll kind of learn being a founder yourself along your founder journey?

Sure, there's a lot of mistakes you're going to make along the way for sure. But the biggest thing I think now is just take advantage of the tools that weren't available even just a couple of years ago for founding a company or founding a new project, or building something within a company, right? All those things, there's the ability for individuals to take on so much more initiative, responsibility, as well as ownership than was ever available. And so whether you're founding a company, whether you're building something within a company, or whether you're working with others, I think that ability to amplify yourself and amplify others around you is multiple times of where you were... it's just crazy, almost cloning yourself is now possible at a level of 2-3x, and when we get to the level of 10x or 100x of what you can do as a person, I mean it just changes the productivity level for every person in the world. So that's how I think about it, and I'm very excited about the coming future.

As am I, as am I. And hopefully you watching today, listening today, can get inspired and get off the fence and start building that company you've been procrastinating to build year over year. You have all the tools available to you now to do so, so get at it. Avery, thank you so much for taking the time out of your busy day to come chat with us today. Wishing you and Aptos the best of luck building out the future of AI and blockchain and that cross-section there. And yeah, looking forward to seeing what everyone starts to build. Get on building that company, come on the show and showcase it! And with that, I just want to wish everybody a very happy Friday, happy weekend wherever you may be, and we will catch you back here next week for another great episode of DevNTell. Until then, have a good one folks. Cheers!