Why Stablecoins Won’t Kill Cards Ft. a16z Partner Noah Levine

[0:00] There's a bit of, I feel like a skeuomorphic point of view of moving from even e-commerce to agentic commerce, where everyone is thinking about flows where humans are in the loop. I think where adoption where is going to accelerate significantly is actually when [music] humans are not in the loop. You know, in AI today, I think the biggest bottleneck is actually human involvement. >> Welcome to Tokenized, [music] the show focused on stablecoins and the institutional adoption of tokenized real-world assets. My name is Simon Taylor, I'm your host, author of Fintech Brainfood, and head of market dev at Tempos. And back on our agentic commerce
[0:30] series Bam Azizi, who's CEO and founder of Mesh. How are you doing, Bam? >> Good, good, good. Super excited for this session. >> Yeah, this is going to be a good one. Uh joining us this week is a regular guest on the main show and somewhat of a co-host these days, helping us out when Kai or I have um super secret missions to do. Noah Levine, partner at a16z. How are you doing, Noah? >> Doing great. Excited to be on the show. >> Yeah, excited to get into your agentic commerce thesis. I know you've been writing a lot lately. Um I know our good friend Sonal speaks very highly of some
[1:00] of your work as well. So, um before we get into all of that, I need to remind you viewers and listeners that the views and opinions of our contributors today are their own and might not reflect those of companies they represent. And please don't take anything we say as tax, legal, or financial advice. This is for informational purposes only. And I'm very happy to remind you, of course, that this podcast is sponsored by our friends at Visa and our friends at Mesh. This episode, if it isn't already obvious, is sponsored by our friends at
[1:31] Visa, the leader in digital payments. Visa's tokenized asset platform, or V Tap, uses smart contracts and cryptography to help banks bring fiat currencies on chain. Whether it's launching a stablecoin or deposit tokens or something else. V Tap allows financial institutions to issue fiat-backed tokens, improving financial efficiency and enabling programmable finance. Check out the links in this episode's description to express your
[2:02] interest in VITA app. This episode is brought to you by Mesh, the first global crypto payments network. Connecting over 300 wallets, exchanges, and payments platforms, Mesh enables anyone to pay and get paid instantly anywhere in any asset. Mesh makes digital transactions seamless, secure, and universal, fueling the next era of agentic commerce. Learn more at meshpay.com/ai.
[2:34] All right, now hit me with the thesis. Where are we on agentic commerce? There's no lack of hype, but where are we now? >> Absolutely, yeah. Well, I mean, you know, definitely a lot of hype, and and I think, you know, as a starting point, I think we're still very early. You know, I think a lot of the conversations about agentic commerce have really started in the last 6 to 12 months, especially, you know, coming out of the the development of, you know, Codex from OpenAI and and Claude Code um from Anthropic. You know, and even just looking at the data, I think, you know, the work that Artemis did with Visa um coming out with some of the data
[3:05] was really good work, and it shows that, you know, when you look at the adjusted volume, there's only something like 15 million in volume, and a lot of that is concentrated amongst a a small set of wallets. I think that being said, you know, there's a tremendous amount of excitement right now. There's a lot of, you know, new protocols coming out, whether it's XPR 2 or MPP. I think there's a lot more um definition around what agentic commerce is going to look like, especially um for some of the more developer-facing use cases. Um and, you know, I think that's the the area that, you know, we've been spending a lot of time on and and I'm most
[3:36] excited about. I think if you look at a year ago, a lot of the conversation around agentic commerce was really, you know, you sort of using um whether it's Claude or whether it's ChatGPT as sort of a new discovery tool, and, you know, this is how you're going to be buying sneakers or buying uh groceries. And while I think there's opportunity there to improve how discovery works, it was kind of unclear how payments were actually going to change, and you know, even in some early, you know, testing from, you know, players like OpenAI and and and trying to embed payments within um within ChatGPT, like it was unclear that there was a significant amount of
[4:06] demand. And I think now where the direction of travel's going is is really more focused on, you know, you have this sort of two forces going to each other at once. On the one end, there's this whole new class of of developers who um, you know, maybe never written a line of code before, but now all of a sudden can build full-scale applications. They need to get access to developer tooling, and they're going to want to have to pay for that. And then I think on the other side, there's a whole new class of merchants, you know, that we're seeing that are not only building tools, but then going and looking to actually sell those, and um, you know, figuring out
[4:36] how the payments works there is is quite complex. So, I think that's the the area that we've been more excited about, and where I think there's a lot of potential. >> No, thank you for that. I I You talk about a new class of merchant and a new class of kind of buyer. I mean, the developer is fairly clear to me that this is somebody who, you know, might not started a business before, but now you're getting these, you know, single-person million-dollar ARR companies popping up, and you know, I think in Stripe's annual letter, they said that there's a category company
[5:06] formation is is through the roof, and so is this like scale of revenue growth for these single-person companies. But, what's the merchant side of that look like? >> Yeah, for sure. I mean, I think the an interesting parallel is sort of, you know, what Shopify enabled for e-commerce sellers. You know, historically, like if you wanted to start a clothing brand, it was extremely complicated. You'd have to go and buy a ton of inventory and source suppliers, and uh, you know, it it was it was you're basically building a whole operation, and you know, what Shopify did, and I think what the whole drop shipping kind of revolution did, is it
[5:36] made it such that a very small team, and if not a single person, could start a direct-to-consumer brand and get a storefront loaded and start accepting payments, and it made it very easy to do that. And I think in a similar way, what we're seeing, you know, with with these um you know, AI coding tools is it's kind of doing the exact same thing but for software. Um and so I think in a similar way, we're going to see a whole new revolution of of merchants who are creating cool and interesting tools and services and then want to go and sell it. Um and and that's I think going to be a new wave of of entrepreneurship.
[6:07] >> I got I got it. So the developer is both using AI tools and through the um code X code code, they need to buy stuff in order to do the job and build the thing. So that that's a type of new buyer. But the seller is also that same developer who's building their product and selling it with you know, it could be software, it could be AI tools. And that person didn't exist before and of course famously Stripe, Shopify made people who weren't previously sellers online or didn't take payments online be
[6:37] people who who could do exactly that. And so interested in your views in this band because uh you have this kind of global perspective given what you do at Mash. Are you seeing this type of customer base adopt uh what you do and and your thoughts on what Noah said so far? >> Yeah, yeah. Like he's a spot on. I think uh the trend that we're seeing at Mash is um this this new type of companies that they're basically building on top of Asian. They have different type of needs. Yeah, that needs from our
[7:09] perspective um doesn't fit with what TradFi or traditional payment networks can offer. For instance, we're seeing a lot of uh transactions happening on X402. We're seeing a lot of transactions happening to parse APIs, to parse websites that uh at a different scale that that that's like requires new type of payments which is global and can support micro transactions. Um that is the trend that we're seeing it on our on our end. And I
[7:40] think that's kind of where stablecoin or blockchain and AI kind of merge. Uh that that would be the crossroad and super excited for seeing more type of payment methods um or payment needs to be addressed by by stablecoin and blockchain. >> Now, you wrote a piece called uh "Agentic Commerce Won't Kill Cards, But It Will Open a Gap." And I think in in that you talked a little bit about these new um uh new merchants, and you talked about the new developers. But can can you tell me a little bit more about what your
[8:11] thesis was inside that and why agentic commerce doesn't kill cards specifically? Cuz I I do think there's a view in the stablecoin land that obviously stablecoins are better. >> Yeah, no, for sure. I mean, I think from a starting point, you know, I think the the rationale for writing the piece was I think it was right after Citrine wrote his piece and uh you know, it sent uh you know, some of the card network stocks down and saying that basically stablecoins were going to completely replace cards. And I think the the main challenge with that is it kind of ignores a lot of the benefits that cards provide. You know, for one thing, um
[8:42] basically ubiquitous acceptance that 150 million merchants globally, you know, provides things like chargebacks, um rewards I think are are a big component. And so, for a lot of the transactions that I kind of mentioned before that are sort of like proxy commerce where, you know, AI is helping with the discovery phase, but ultimately the you know, the consumer is the same and the merchant is the same, it feels like cards are actually very well positioned to continue to succeed in those types of payments. And you know, I kind of talked in the piece about how network tokenization, which is you know, been around for like 15 years and um you know, most commonly seen in things like
[9:13] in Apple Pay, you know, are actually very well suited for dealing with these types of transactions. Um but rather I where I see stablecoins having a really big opportunity is when the actual either the type of the transaction changes or you're fundamentally dealing with a new merchant that you know, it would be challenging for them to accept cards. And And again, like kind of using the historical analogs like, you know, whenever we've seen a platform shift in commerce or just a platform shift in general, there's been new payments companies or new payment methods that have come to form that have helped to,
[9:44] you know, serve serve those sort of ill-fitted transactions. You know, examples being things like PayPal with with eBay. You know, I just mentioned Shopify and Stripe. I think those are both, you know, good examples. And I think in the case of stablecoins, you know, the way I look at it is stablecoins are very much a digital cash. And so, you know, situations where cash-like payments are successful are ones where I can see this working. And so, you know, I think if you see a whole new class of merchants, these sort of headless merchants that I mentioned, proliferating, it's going to be very challenging for them to immediately
[10:15] accept payments. You know, getting under it and actually getting, you know, you know, a processor to support them, getting a merchant bank account. You know, very much similar to kind of the role that PayFac played in in e-commerce, I think is a similar role to what stablecoins can play, too. And then, I think in addition to that, you know, specifically for the types of transactions we're seeing here, which are often times, you know, very small in nature, it's very it's actually quite challenging from an economic standpoint for the card networks to support them. Now, in the piece I do argue that I think over time we will see the
[10:46] incumbents develop, you know, either new types of transactions or or or standards to help, you know, match or make these transactions work. But I think in that gap, there's a massive opportunity for stablecoins to play a role. And I think the reason why stablecoins are such a great solution is because they're programmable, they're global by nature, and I also think they they are conducive with something that's more developer ecosystem rather than traditional merchants. So, that that's that's my general perspective. >> I I think to add to that, I I don't I
[11:16] agree. I don't believe agentic commerce will kill card networks, but the stablecoin do. That's That's kind of where I can a little bit diverge from what what Noah was mentioning because if you think about it historically, we had commerce and then we had e-commerce. E-commerce didn't kill commerce, it just stacked on top of it. Then we had mobile commerce, it didn't kill e-commerce, right? Mobile commerce came on top of e-commerce. Now, agentic commerce is going to make the entire commerce 10x more valuable and then 10x
[11:47] more transactions or maybe thousands x more transactions because of the agentic nature of it and the microtransactions that we'll have. I mentioned it before that we have like 5 billion human being connected to internet, we do two transactions in average per day. You'll have hundreds of billions of agents connected to internet, they would do thousands of transactions a day. So, you can see that it's order of magnitude not like higher number of transactions. But I I think we are slowly shifting
[12:17] toward an ecosystem that the 70-year-old technology, the card network that we have and we know as of today, it just doesn't make sense anymore and it will be replaced by stable coins. I'm not saying Visa mastercard will have problems. They they will adopt, they would use that as a weapon, as a ammunition, as a technology, as a tool to address the new and the modern needs of payments, uh, but I I don't believe mainframes are going to stay around and the networks the way that we know it
[12:47] is going to be dead before we know it. >> Yeah, I think the way that we know it is doing a heavy lift in that sentence, right? Like the networks themselves sell something that I think is underpriced, which is trust and brand. So, when you see the Visa logo, you kind of have this confidence on the consumer protections that if I make this payment but the goods are never delivered, there's a rulebook that says I'm going to get made whole, you know, I have some protections. There's a whole bunch of fraud rules that sit around that. It's kind of the boring ugly stuff that you
[13:17] don't think about that like, you know, like you walk into another country with your currency, they don't necessarily accept it but they accept the Visa card. You know, accepted everywhere was always the the classic thing. And we sort of have that a little bit with stable coins. They're not necessarily accepted everywhere, and which stable coin on which network? And so, you know, you need something like a mesh um that can hopefully help make that a a little bit easier. Um but I'm really curious to drill into that a little bit more,
[13:47] Noah. Uh cuz we talked for a long time about having internet protocols for money, and we do have a couple of protocols. You mentioned it in the outset, you know, MPP and X402, but the volumes are very low. Are those protocols going to get adopted? Are they useful? I mean, I know Visa was quite involved in those, and they're they're looking to adopt them, but how does this shake out? Because I'm I'm sitting here thinking this just looks like lapping like a gentle
[14:18] commerce by the stable coin industry right now, if we're being honest. >> Yeah, it's a great question, and and frankly, I don't know the the the the true answer. I'm I'm not clairvoyant. However, I will say like I think the with any standard, the ones that are going to be the most widely adopted are the ones that are the most neutral, that are the most open, and ultimately do the best job of getting the most connectivity. I think the real opportunity is what you build on top of it, whether it's on the consumer side and in enabling you to have applications or or ways to to actually um you know, do things on these
[14:48] protocols, as well as on the other side of the acceptance and in building, you know, widespread acceptance and usage of these protocols themselves. And so, I think it right now it's it's sort of like you know, to your point, everyone is trying to create their own standard or create their own protocol and and play a role in that stack. But I think what people are going to find out is that a lot of the value, similar to what we've seen with with the internet, is not going to accrue to the whoever owns the standard. I think it's whatever gets built on top of it. Um and so, that's why, you know, for me like I I I spend a lot more of my time thinking about what what's going to actually get done on top
[15:18] of it versus actually thinking about the standard itself. And then I also think the other problem, too, is there's a bit of I feel like a skeuomorphic point of view of moving from even e-commerce to agentic commerce, where everyone is thinking about flows where humans are in the loop. I think where adoption will is going to accelerate significantly is actually when humans are not in the loop. You know, in AI today, I think the biggest bottleneck is actually human involvement. If you think about why, like going from ChatGPT to Codex was so powerful, it is because it took humans increasingly out of the loop. And even
[15:48] today, I think the biggest problem that we see and why there hasn't been a significant amount of demand for agentic commerce is people don't know what they want to buy, they don't know, you know, what they want to build. I think there's going to be a really interesting opportunity where, you know, you're going to have literally sovereign agents that are going to be either building their own companies or or doing their own tasks. Maybe they're funded from human funded by humans, but ultimately the entire decision set and the entire scope of what they do is going to be decided by them. And I think that's where it gets really interesting because you basically have something running 24/7 trying to accomplish a task and do
[16:18] something. And there's just a lot more room to actually transact and to have commerce happen versus needing a human to basically be involved in and say to do these things. And they may not know what what exactly to do. And I think this matches a lot, Simon, with, you know, you sort of your thesis of the intention economy and this idea of you, you know, coming in with an intent of what you want to do. And then ultimately the agent is the buyer. I think the the difference could be, well, what if the agent is actually not only the one that's doing the transaction transacting, but also determining what they want to build in the first place. I think that's where, you know, you're
[16:49] going to see a lot more activity amongst these protocols. >> Yeah, just to just to zoom back on that for a second. Yeah, that you're familiar with the attention economy. Most of e-commerce is based around trying to monetize the gap between you don't know what you want and we'll try and make you want this thing and then you buy this thing. Um, whereas agents don't necessarily have desires. They have a mandate. They have, you know, an intent, go buy this thing for this price or go find the cheapest price or whatever. And
[17:19] so, for them it's about collapsing the gap between their intent and the thing actually happening as as short as possible. But, their decision-making process might be quite different. You know, it might be, you know, what is the cheapest? What is the best? What do the reviews look like? And so, there's some element there of like, what does it look like when the agent is the decision-maker? And I I'm interested in knowing like, okay, so if the value is accruing above the protocols, what are you seeing above the protocols right now that's that's kind of out there along
[17:49] those lines? >> Yeah, so I think the the two biggest opportunities is one on the sort of the wallet side. I think, you know, for example, we we invested in a company Merit Systems. Um they have this product called Agent Cash. And basically what it is is an aggregation of a bunch of headless merchants uh that consumers can go and buy from. And um I think the key lock-in there is that they control the wallet which you fund and then ultimately can go and interact with these services. I think making it very simple for people to access these tools uh and then be able to pay for them is is uh makes it a lot easier to not only
[18:21] for the discovery process, but also in terms of how the actual payments get executed and um you know, the services get absorbed. And then I think the other side, too, and I and I wrote about this is it feels like, you know, the infrastructure is is largely there. I mean, I think there's still room for improvement and development, but it feels like the infrastructure is is largely there. The bigger opportunity is actually kind of creating these new merchants um and in new use cases and um sort of workflows um where they're actually, you know, useful and you know, can accomplish tasks. And while we've definitely seen a lot of, you know, new
[18:51] services come up, I think there's still probably a lot of room for improvement and um and new opportunities in the space. >> Do you think we'll have like multiple protocols being used by these services that they're sitting on top of the protocols or you see some consolidation on XPR 2 or MPP or or any other protocols? >> Yeah, I mean, I think in the fullness of time it it makes sense for them to kind of consolidate. Um it obviously doesn't make sense to have 50 50 protocols. I think that would, you know, make adoption really challenging because then
[19:21] everyone needs to go and adopt them a million different protocols. I think in terms of like which one wins, I think a lot of it comes down to adoption and and distribution. So, you know, whoever can get the most interesting um applications on on the the developer consumer side as well as being able to find, you you know, useful endpoints that they can go and transact at and I think over time there may be real network effects there. Now, those network effects may not translate into something that's very economically valuable for the protocol, but I think it will increase the utilization of it.
[19:51] Um so, you know, I think right now because it's so early everyone feels like they have a a chance of winning with their protocol, but I think eventually once they realize that there's not a tremendous amount of economic value at that layer, it'll ultimately go to whoever, you know, to to again, like I said before, whatever is the most open and um ultimately sort of trying to commoditize that part of the stack so that, you know, they can have this sort of neutral governance and be able to compete around um parts of the stack above it. >> So, I've thought it was so interesting that AWS have um built something into
[20:21] bedrock that lets AI software programs buy and sell things completely on their own. Like the hyperscalers are building this tooling and they supported both MPP and X402. So, now it seems the answer is well, we'll just support both and and see what what gets it involved. And there's a lot of tasks happening. If you think about the scaling companies in AI, it's companies like Exa and companies like Parallel who make search and organizing data just a whole lot better. And maybe it's the agent that wants to
[20:51] be the customer to that. And historically they've been selling subscriptions, but those subscriptions and um kind of card models might not make sense. Do you think this is a tipping point cuz I've seen a lot of people try and play with these companies, a lot lot people try and get it get adoption with them. or you know is is the card kind of fine for those and it really is just going to be the the net new micro merchant the net new sort of Shopify type of thing that's going to be the space and and have you seen any any counter data to
[21:22] that? >> Yeah, it's a good question. I don't have counter data to you but I think from a from a high level, you know, it feels like if what people are using these this you know sort of tooling for becomes increasingly ephemeral then the rationale for a subscription kind of goes down. So I think over time like as people become become more they they experiment more and they're they're using things on a a more short-term basis. They just want to get access to it immediately. They don't want to go through a billing team and sales team and you know go through that whole process and I think
[21:53] they want to just get access to something immediately and test it and try it and and then continuously use it over time. I also think it relates to the kind of who the buyer is to our point before. You know, if the agent is ultimately the one that is sourcing um sourcing this and and purchasing it um I think then a lot of the you know components of what a subscription is good for goes away um and over time I think that that will increasingly make um this style of of selling uh increasingly popular. >> I don't know if you saw that like um both uh I think Cosa Meta Ramp and few
[22:26] of the companies have released open router like products for for token routing as as metered billing comes to enterprise and token costs become a really issue and people are trying to get more per token more value per token. I wonder as well if there's some opportunities there as as people become more um cost pressured and you know the all you can eat subscription models start to go away. Um can the agents find you know the best value per token on on a pricing base? Like this is something
[22:56] that um again I don't know if it's if you've seen anything that looks like this now as you're looking at companies or any teams building this sort of stuff, but but interested in your views on on that whole space. >> Yeah, for sure. I mean, I think any type of orchestration business is going to be more successful when there's more fragmentation at the level of what they're trying to orchestrate. So, I think it really comes down to, you know, in the long run, you know, how many models are are relevant whether that's a performance or or even a cost basis. Um will ultimately dictate dictate that. I think the other way to look at it too is
[23:26] kind of wrapping it back to this idea of sort of headless merchants. There's also I think this interesting concept of composable commerce that could happen where it's not necessarily that they're orchestrating between multiple models, but rather if you're selling something on a per use basis, you know, it makes a lot of sense to kind of associate your costs on the same line. So, you could have a situation where somebody hits your endpoint and as soon as they hit your endpoint, it immediately hits um you know, your ability to go and get compute. And I think that that, you know, becomes a really interesting way uh especially
[23:57] with these types of businesses where, you know, it's not like they're going to buy inventory. It's like their number one cost item. I think if you can attach the end whoever the end customer is what they're doing to ultimately what your costs or the resources you need to execute it, um you can do a you can run a much more efficient business. So, I agree that I think there's a big opportunity on on compute both in terms of the model orchestration layer, but I also think there's a an interesting way to sort of have more predictable and real-time uh funding of your costs um using this as well.
[24:27] >> Yeah, it's kind of that like uh do you want kind of uh stable price of the futures market or do you want sort of um the best real-time price that's available in the market right now? Um because it seems like every day there's a new model and they, you know, sort of the level of intelligence you can get per dollar is really variable. So, you could potentially have much more efficient business by constantly routing that to whatever just came out on on
[24:57] open router. Um Bam, any thoughts on on this topic? >> Yeah, it's it's interesting. One of our customers, Kalshi, has been for that scenarios because that basically that would be the most expensive line item for a lot of these companies that they're building on top of AI. And then talking is their number one cost. So, they're they're built and announce their purpose or future for computes as a commodity that you can if if you you're spending a
[25:28] ton of money there, you can hedge against it. So, that that would that I I thought I would shout out to Kalshi for doing this interesting model. And they were like kind of the first one that they're doing it in in US. But, I I I totally agree with what Noam mentioned. What One one question that I have for him is like do you think the traditional payments can solve the need for just-in-time access to compute or we really need the stablecoin?
[26:00] Do you see like because those are not microtransactions, right? So, if you want to buy like thousands of dollars worth of compute, why should you use a stablecoin or a gentle commerce should do on top of a stablecoin versus using like let's say old-fashioned payment cards? >> Yeah, it's a good question. I mean, I think you know, the the benefit of stablecoins in this use case is the obviously the real-time settlement aspect. Yeah. >> Yeah. >> You know, I I And again, like I think another mistake that I think people who
[26:30] aren't like super deep in payments make is they kind of look at authorization and settlement as as sort of one thing and one bundled product when when in reality like they're they're two very different things. And so, you know, I I think we could get to the point where cards are still the form factor of payments and it's still the authorization mechanism, but ultimately stablecoins run on the back end and you know, you can enable things like real-time sort of streaming settlement and I know that's an area where um um the card networks are looking to advance. Obviously, you know, Visa's and Mastercard have been doing stablecoin settlement for a while. Volumes have
[27:01] increased significantly there. But a lot of the use case has been more about, you know, making it easier for issuers to settle and, you know, certain acquirers who are supporting crypto merchants. Like, I think the real unlock in that is when you can actually start to enable more real time and more streaming settlement flows. And so, yeah, I I think stablecoins are very useful now with at least in the current form for that. But again, in the long run, I think if you believe that a lot of these providers are, you know, still accept cards and most consumers and developers and businesses also still use cards,
[27:32] then I think you can end up in a situation where cards are still the preferred payment method, but ultimately the back-end settlement improves significantly and, you know, the actual way that money moves um happens much more on a real time basis. >> That 12 months ago, that back-end settlement um conversation was I think one that was uh not as mainstream, but now in my day job, I'm definitely seeing um people go, "Oh, yeah, no, this is a settlement um thing. What Visa does is authorize the movement of money and build a trust and
[28:03] reputation network around the authorization of the movement of money. And if you could stream the settlement of that, you it's kind of like when you walk into a gas station in the United States, you know, you sort of uh lift the nozzle, uh you swipe the card, and then you start pumping. What's happened there is you've opened an authorization, but they haven't captured and closed the off until you remove the nozzle from and place it back onto the tank. There's no reason why you couldn't do that in a gen tech, right? Like, if you need something faster, you swipe the card or you get started with a transaction, agents move
[28:35] at light speed, and then close the off, and then the the the payment happens and settles on a stablecoin um kind of kind of afterwards because that can happen really quickly. One One thing I want to quiz you about now is like do you see this world where these micro merchants like spin up, become wildly successful, and have that classic inventory problem that merchants had, which is you know like if I sell out of all of my widgets, but I'm not getting paid for 3 days, then I can't buy any more inventory, so
[29:06] now I can't make any more sales. And with like real-time tokens, I could sell out of inventory and not get paid for 3 days. >> Yeah, exactly. And this is kind of going back to the point I made before on this idea of composable commerce, where I again if the number one input or cost or even bottleneck is access to compute, then I feel like it makes a lot of sense for them to be purchasing or acquiring this on a real-time basis, rather than trying to forecast this out. Um especially in in I think this is true both for the really small ones as well as as they get really large. I think for
[29:37] the larger ones like maybe there's some interesting um sort of like wholesale and enterprise agreements they could have um to get access and maybe that's less of an issue. But I do think there's a huge opportunity where um if you believe that commerce is going to again become more composable where as soon as a uh an API call comes, you as the merchant are going and buying whatever, you know, service or resources you need, maybe that's compute. On the back end, I think you could have more predictable and and and much more enduring uh scale as you would if you had to go and
[30:07] forecast that out. Cuz to your point, maybe one month you have a just significant amount of demand, you didn't forecast for it, and now, you know, you're running into some cash flow issues or issues you know, you you have a bottleneck on uh the amount of compute you can get. And so I think that's also another opportunity um where this goes is it's not only is it the these merchants selling to developers, but also it's them buying the resources they need to to operationalize um what they're doing. >> Yeah, bring in the revenue in real-time by selling to developers, but in real-time go buy the resources you need to make to manufacture and distribute
[30:39] your product. It's it's kind of that streaming is such a different way of thinking about things. >> It's very much like, you know, again, I kind of similar to the example I gave earlier, it's very much like the drop shipping of uh software. You know what I mean? The exact same thing happens there where, you know, you have somebody they open a Shopify store, they're selling t-shirts, and then they have a Printify in the back end that is doing just-in-time manufacturing or just-in-time production uh for them. I think you see a very similar thing here, um you know, happening in this space as well. >> Drop shipping for software, that's a great um blog post title. You need to
[31:10] make that happen. The other one that I saw you talk about is Jevons paradox is coming for finance. And of course Jevons paradox is this idea that as things get cheaper, uh the demand actually increases, not not decreases, and and you tend to end up with with a lot more of it. Um how do you see that manifesting? >> Yeah, so I think in in that piece I was more so talking about sort of the role that stable coins are playing for for fintech more broadly, and um you know, this idea that, you know, what tokenization and and stable coins specifically are doing is it's
[31:40] fundamentally dropping the barriers to entry to to build financial products, and you know, as a result, I think you're going to see net new markets that maybe, you know, haven't been well served by incumbents because it's been so challenging to to build there, and um as a result, the overall TAM and the overall market will grow to the point where, you know, maybe some markets where historically investors have been hesitant to participate from a fintech perspective, uh suddenly it becomes a very compelling opportunity. You know, I think this a similar thing could definitely happen in in Asia and Singapore, especially from a software
[32:10] perspective, um where if, you know, you have a lot more software coming out and the cost of that is as a result going down, you know, maybe you see many more people having exposure and access to these things, and therefore, um the market becomes a lot more interesting and enlarges as a result of that. >> Yeah, yeah. I think the overall um we are like entering an environment that everything is like on a faster pace, right? Um from development of the software to shipping it, to maintaining it, to
[32:42] getting paid for it, right? So, everything is sounds like used to be like month of work to get it done. Now, everything is kind of instant. Everything is streamed. You, um, you don't have this traditional sense of like, I will like build it and then I ship it and then I collect the payment for it. And like, if you remember like we had like net 30, net 60, net 90 for for accounting. Uh, now everything is like instant, right? You build it and ship it at the same time, you get paid for it. Or you
[33:12] have a resource which is could be like for instance computer or API that you enable it and then people are using it and then they instantly they would pay for it uh, based on the amount that they're using. Uh, everything seems like in a stream uh, versus just uh, like sequence of things. Um, so it's it's a it's a very fast pace world, but also very very interesting and exciting. There is no better time for uh, entrepreneurs and builders to build. So, I'm excited to see new type of um,
[33:42] softwares and technologies to be built in this era. And and and and a stablecoin would be the of course the the winner of all because um, for every single service that you are offering, you have to collect payment and I think stablecoin is a perfect uh, form of payments uh, for global economy. >> No, I'm interested. Um, what does it take to build a durable business in this world where um, money is streaming, software is incredibly cheap and fast to build. I could build a widget tomorrow, but somebody else can build it. Um, you
[34:13] know, did did did a first principle still apply of like brand and trust and people buying from people? Uh, where are you seeing durability especially in financial services? Um, and and what are people clustering around as as you look at teams and uh, potential investments? >> Yeah, so I mean I'd say in general like as the barriers to entry get lower, I think distribution becomes increasingly important. You know, it's sort of the the LTV to CAC metric uh becomes a lot more important where, you know, either you have to find a more
[34:43] niche subset that you're going to sell to um that you can really own and and get, you know, strong um retention within them and then try and upsell around it. Um or it's just about having a really killer go-to-market strategy and doing it in a very lean way. Um and so, you know, I think historically, at least in fintech, a lot of it was around kind of who was there first and who can build for the really hard things, you know, whether it's licensing or banking integrations or whatnot. And I think now as stablecoins reduce some of some of the barrier there
[35:13] where you can just build new financial products um with a lot more speed, I think the the distribution and and who your customer is and how much you own them um becomes increasingly important. And you know, and then I think in like the in the agent-to-commerce parallel, I think it's sort of similar. It's, you know, you if you're a new merchant, I think it's you know, having a subset of developers that or even agents in this case um that have preference for you. And so, you know, maybe it's having better APIs so that it's easier to access your endpoint. Or or maybe it's, you know, offering something that um is
[35:44] novel or unique. So, you know, I I think it's hard, to your point, it's maybe harder to to build something more durable, but I also think that there's um opportunity to um go deeper amongst those types of users. >> I had a uh uh web page open for many months. I was just looking for it and I lost it, but it was like looking at the tools that Claude code chose to use because it kind of recommends something uh when it goes uh to accept payments, it 96% of the time it picks Stripe. Uh it's it's going to
[36:16] pick for sell or railway depending on what project it is that you're building for for the hosting. And how do you become top of mind for those AI agents? I think it's really interesting. And how does that evolve over time be from, you know, what was in that training data to something that's, you know, more price sensitive or or something else? So, I think that more question is super interesting. I'm also interested in like fraud and risk here because if I can build 10 duplicates of like Noah's AI agent, how do I know I'm transacting
[36:47] with the right one? And have you seen anybody playing in that sort of space? >> Yeah, it's a great question. Seen a couple teams thinking about what is fraud and chargebacks look like in this sort of new economic model. I've been a little bit resistant or hesitant about like the how much of the need there is there. And the reason why I say that is I think if you're buying something that's very high value and there's some sort of fraud that happens where you don't get the product that you want, like that's really a frustrating, really annoying. If I go and buy uh $200 pair of sneakers and they send me, you
[37:17] know, a different shoe, well, that's, you know, that's that's very unfortunate and I'm going to be upset about it. I think in this model though, if you're buying something like an API for example and it's a cent and, you know, it doesn't give you what you want or you don't like it, I think there's less of a demand or a need to charge back on a cent. Um so, I I think in a lot of these transactions, chargebacks become less of a a problem because if I go and interact with something and it doesn't work, well, then I'm just not going to interact with it again and my, you know, economic loss is uh negligible. I think
[37:47] if you start seeing transactions get larger, maybe then that, you know, increases. Um and then therefore, I think there's a lot of opportunity um maybe at a network level or or new types of companies that are coming in and, you know, build what is the sort of risk scoring and framework look like for for these types of out, you know, outcomes. I also think even at a protocol level, I think there's some interesting ideas of like what is a verified merchant look like and, you know, how do we evaluate whether um you know, is there some sort of scoring or reputation system? Almost like how many GitHub stars do they have? You know, maybe that that become the new
[38:19] um you know, mechanism for trust. Um so, I think a lot of it's unclear, but I think at least in in at least the data of the type of transactions we're seeing today, there's not as much of a problem because the nature of the transactions are super small. >> Cross that bridge when you get to it. Um, I'm interested in uh of the companies and teams you've seen, uh what what kind of categories are you looking for that you either haven't seen or that you thought you would see? Uh, you know, like what what are you if if somebody were to walk in tomorrow that does X, what is X?
[38:49] What are you what are you looking for? >> Yeah, so like I said before, I still think there's a, you know, massive opportunity within the, you know, whoever what is the like developer wallet look like and part of the reason why I think that's really exciting is because one, you control I think a lot of the discovery of where does commerce actually go to? You can sort of aggregate a lot of these merchants and then, um, you know, help to dictate which ones that the developers often go to. But then I also think too, like if you're dealing with very small transactions, there's a lot of opportunity to upcharge on price. So like for example, like if I sell you
[39:20] something for a you know, I like use the sneaker example, like if I have a $200 pair of shoes and I sell you it for $400, like mentally you're going to be like that's like way overpriced, you know, those sneakers are way overpriced. I think if you're dealing with something like an API call and it's a cent and I now charge you two cents, you know, my margins are a hundred, you know, a hundred percent and it's very hard for you to interpret whether or not that, you know, is cheap or not. And again, if you own the developer and you're the the wallet and the discovery layer, I think you have a lot of control there. Um, to the point earlier, I think what I'm really excited to see is more more
[39:51] things that are less humans in the flow, you know, more more of these like sort of sovereign agent ideas and whether that's agents forming as a independent or you know, organization or company that, um, can go and, you know, have a a very high-level task and can just go and execute on it. Um, or even workflows that are, you know, completely isolated from humans altogether. I I don't think I've seen as much of that. Most of the companies that we've seen today are either, you know, infrastructure being built on top of existing protocols or some sort of you know application where humans are very much deeply in the loop
[40:21] and I'm excited to see the kind of the crazy the crazy ideas of where agents are sort of individual economic actors and I think if that happens you're also going to see much bigger need for net new infrastructure. You know, one of the big comments of like why people were like stable coins are are better than cards for these types of payments is this whole identity notion. But realistically like if you have an agent and that agent by definition you know, the definition of agent is something that acts on behalf of someone else. Well, you can always tag that identity back to a human or back to a developer or back to a company and so I
[40:52] think it less of a need there. But when you have a an agent that is literally you know, completely self-sovereign, you know, then you start to ask questions of like well, what is what even is this and you know, how do I think about this? How do I trust it? And so I think the two will will flow together both in terms of these new agents as applications, but also in terms of that proliferation will create a new need for or a need for net new infrastructure. >> zero human companies. Uh, it'd be be interesting to see when when that's coming. That's that's certainly out there. Bam, your closing thoughts on
[41:22] this one. >> Yeah, I think I'm I'm a little bit biased, but I think the perfect tool for that sovereign AI tool that it can can basically act on on itself and take actions as a smart contracts. Again, back to back to blockchain like you can put set rules even even for chargebacks or for like fraud situation that you need to get your money back. Maybe if you're a human one cent doesn't matter. Or you just say like I won't go back to this specific service provider
[41:53] because I got burned once or twice. But if there is a smart contract in the middle, everything is rule-based and everything is decentralized. There is no human or there is no manual supervision and things will happen automatically, right? So I I could see that even for like micro transactions will have a better mechanisms and better charge backs and and fraud detections using stable coin and using smart contracts and blockchain as a
[42:23] technology. There are net new networks that are being built for these specific use cases. I can I can call Tempo as as one of them that basically is is built with the mindset of this is going to be used by machines. Going to be used by agents for the sake of payments. So there are building tools and basically embedding tools that can be used to have a better sovereign agents and have a more optimized environment
[42:53] for these sovereign agents to be able to transact on and on on on the financial side. So super exciting era that we're living in. But I I think I I think that's kind of where the crossroad of agent e-commerce or agents and stable coin and blockchain as a broader topic will merge and emerge in and I think that opens up a lot of new opportunities for a lot of new startups that they can build in that era.
[43:23] >> Now was there anything I didn't ask you that you wish I had? >> I mean I'd be curious your take, you know, kind of um especially given obviously your day job at Tempo, you know, what are you seeing in in agent e-commerce or where where are you seeing demand both from the sort of incumbents as well as net new use cases from net new actors? What can you tell me about that? >> Do you want the really quiet part out loud? I'm not. Um it's not there. It doesn't exist. It's a mirage. And I say that as somebody who really
[43:53] believes it will come. And it's a matter of when not if and I fully buy that there's a generation of companies that got big enough that they don't need this on Stripe. And that teams of humans formed companies that built products that could accept payments as classical merchants. So, I really like your idea of the headless merchant, and I don't know that the uh the coding harnesses have got embedded wallets yet, and I don't know that the
[44:23] headless merchants have really got customers yet. And until that ecosystem kind of evolves, uh your thesis remains untested. I'm much much more interested in the settlement side of the conversation because uh my suspicion is when we get uh wallets inside of coding harnesses, that wallet will mostly work with cards and use cards most of the time. But, the velocity of settlement and the amount that is settled would probably be too expensive to use the
[44:53] card rails, so we'll start to see that evolution. So, I think uh cards are actually the gateway drug to agentic commerce and stablecoins in a way that people haven't really fully priced in. Um you we would say that Ko Sheffield is our friend, but I also authentically believe that to be true. And I am seeing early shoots of developer interest. Like, the the possibly the most interesting thing is that John Collison and Patrick are hacking away on MPP at the weekend for fun and really enjoying themselves. And what developers are
[45:23] doing at the weekend is generally what everybody will be doing for their job in 10 years time. I think it was actually Chris Dixon that first um made that observation. And that is something where the timeline is just wrong here against the hype. But, the companies you're looking at now, Noah, and the the things they're building might not have their time in the sun next year, but it's it's coming. And when it does, it's going to be orders of magnitude bigger than anything we've ever seen before
[45:54] because the machine economy will dwarf the human economy. Yes, I've watched The Matrix. I know the dark side of that, but I think this is a positive future for humanity. >> Respectfully, Simon, I disagree with you. I agree with you that you're not there yet, but I don't think it takes 10 years. And I bet a dollar on this will happen next two months or so, not even next year, so. That's my take. >> I'll take that dollar. I'll take that dollar. I'm not a betting man, you won't see me on Kelshire anytime soon, but
[46:25] I I absolutely will. With that, I'm going to thank you for coming to my TED Talk. And I'm also going to ask if people are interested in talking to you and learning a little bit more about what you're doing at A16Z, what do they got to do that? >> Yeah, follow me at nlevine19 on X. >> Bam, for you and Mesh? >> Yeah, Bam as in easy Mesh or Meshpay on X. >> You'll find me at SYTaylor on all of the socials screaming into the void at fintechbrainfood.com and of course at tempo.xyz
[46:55] and you'll find a lot more of this show if you hit the subscribe button, hit likes, leave reviews. It really helps us. If you want to say thank you for this type of content, that's the best way to do it. And also, thank you for watching, thank you for listening, and we'll catch you next time.