imagine that in the near future trillions of AI agents will deal with each other doing odd tasks like you want to book a trip and these agents they do everything that is needed that you come from here to there and have a good holiday.
But how does it work? How does those agents deal with each other? Because if one agent uses another agent, they have to deal somehow. And what comes in here is AI commerce. So this is about how to manage those agents in an economic situation. And that's exactly what we talk about today. Our guest today is Don Gossen. He is CEO and founder of Nevermind. And they are the
papal of AI commerce. Yeah, welcome to Beginner's Guide to AI, another interview edition. I'm your host, Deepmar, and let's go.
So, but before I talk too long about Don Gospel myself, let's go to him. So Don, what is it about you? What got you into AI, actually? Well, first up, thanks for having me on the podcast. Appreciate it. What got me into AI? So I've...
From a career point of view, I've basically always been in it. I've been in the machine learning space for 20 years. I studied computer engineering.
at university, and then went into commodities trading on the credit risk side. So basically internally modeling our clients risk scores augmenting what we were getting from external counterparties like Experian Moody's TransUnion. This is in North America. And then modeling out our own scores, they're doing
statistical modeling. What's the likelihood that this client of ours is a deadbeat and isn't going to pay the credit or hedged position back on this natural gas or our energy trade? Isn't as exciting as it sounds.
So I parlayed that into a career in IT consulting for the better part of a decade and a half. And it's a subject matter expert in data and analytics. And that took me all over the world, building large scale data states for machine learning purposes for some of the biggest companies on the planet, HSBC, L'Oreal, Sharp, Mizzou-Ho, AXA, et cetera, moved.
from Canada to the US to Los Angeles, and then from LA to Tokyo, and then from Tokyo to London. And did that for a while. And then jumped into the entrepreneurship game in 2017, started a company that was focused almost entirely on AI.
data provenance, rewards and payments for algorithmic models and data sets. And just have kind of been in this space my entire career and really like hyper focused on AI for the last going on decade now. Yeah, that's, that's long. That's, that's much longer than touchy pities around. Yeah, way longer than it's been cool. Nice.
Yeah, exactly. I think I just remembered, was it Alberto, were you studied? Yeah, so like, this is a fun fact. It's like anecdotal, right? So my Alma matter reminds me of this fun fact. But yeah, my Alma matter is the University of Alberta.
Within Canada, it's, you know, a well-known, you know, highly regarded institution, but globally, you know, it's, it's, it, it doesn't really
hit a beat, but it's the number two AI school in North America. It has been since I was there and prior to that. So it's got a long pedigree in artificial intelligence. If you're in the AI space in North America, you will know the school. Google and DeepMind have put in a bunch of investment in the 2010s.
Whenever, you know, they hit me up for fundraising, it's always, you know, front and center. Hey, remember, we're the number two AI school. Let's keep it that way. So give us some more money. Yeah. So, you know, even before I started an industry,
And my career trajectory, it was already heavily influenced by AI, right? Back then, I was in the computer engineering side of things. Obviously, we were influenced by this. And I remember taking some AI courses because they were like mandated as part of our curriculum and just thinking like, nobody's ever going to use this range science crap. A lot. And lo and behold, I've spent 20 years focused on this stuff.
It's been a very interesting journey and I've gotten incredible exposure across a broad set of industries and use cases and also geographies. Now as we move into the AI agent landscape, understanding how
different agents, whether they're human carbon-based agents or AI, silicon-based agents. I think I've got a strong understanding of how these agents work in these systems, both from a socioeconomic point of view.
regional and cultural influences impact like the economic and commercial side of these agents, right? Anyway, yeah, it's been a very interesting career.
And actually, that's what you do actually now. You do something with agents. So actually, agents, I think the idea that we, as humans, are also agents, but like, carbon-based. And then you have digital agents that is a great explainable agent, but an agent does. But now you have a firm never-minded. Actually, I didn't ask you before, the name never-minded. Where does it come from? Well, so technically, it's pronounced never-mind.
It kind of dates me and my co-founder, right? As children of the 80s and 90s, you know, Arikins back to sort of like the grunge area and one of the preeminent albums that was released by Nirvana back in 92. You know, so there's an affinity towards the name.
But the idea from this, for sort of the entomology of our company name.
is predicated on this belief that there should be self ownership or self sovereignty, regardless of the agent type, that obviously as carbon-based agents, humans, we strive for independence and self sovereignty and our own agency.
And we kind of look at the world as, all right, these AI agents possibly in the future could be seen as equivalents. And so there's this sort of undertone of like, how do we create a scenario where you and I are mine?
but also these agents aren't mined, right? And that we take sovereign ownership of our assets, right? Our data sets, our models, our algorithms, our workflows, the way that we process things, that sort of thing. And so it also alludes to some of the enderpinnings from a technological point of view and our overall ethos
you know, being open source and having a Web3 element to what we're doing, etc. So the independence factor, let's start with that. Yeah. What's the worst case scenario in this? Because do we have independence of them? Not the worst case, but like what's the possible future where we don't have those independent agents?
Yeah, I think we already see it in a contemporary sense, right? Like, we are already heavily influenced by algorithms on a date of date basis, whether that's our buy and spend.
you know, the characteristics or the way that we engage with and on social media. These are already influenced algorithmically. I mean, I think you can historically, over the last 25 years, there's been a predilection by big tech.
to minus, in particular, for data and information, contextual assets, contextual data sets that are relevant to training their algorithms and those businesses profiting. You know, Musk asking the ex people for sending their health data to them. Yeah, exactly.
Yeah, that's it. That's exactly it. And I think what we're seeing sort of play out in real time now with more people that are sort of conscious about this, because let's be honest, you know, the early to mid 2000s and into the 2010s, most people weren't cognizant of this sort of dynamic, right? If the product's free or the product or, you know, that
you're the one that's being used. People just didn't, they weren't really conscious or cognizant of this fact. And now more people are. We see this in the blowback that X receives.
We also see it early on with the release of these transformer models and in particular, you know, chat GPT and some perplexity, et cetera, that are scraping the internet for context, right? And then relaying that information back to users without arguably proper attribution to the creators, the source of that material.
And so it becomes a question of how do we not prevent this from happening, but how do we prescribe proper attribution up and down these value chains? And there's a recognition at least from our side. What we've kind of predicted is that
these AIs, not only is this going to be desirable, right? I think from our point of view, human influence point of view, this is sort of a desired outcome.
I think from an artificial intelligence agent point of view, it's a mandate, right? You and I can establish things like reputation just by having this conversation, right? But questions around how does an agentic AI that is, you know,
Semi are even fully autonomous. How does it establish reputation? Well, it's likely going to establish that reputation through what it provides to an ecosystem and record of what it consumes, right? And it's the record of its actions. Was it a good actor or a malicious actor, right? And that's how it's going to build up this reputational element.
And so if that ability to generate reputation through recording of actions is either mitigated or centralized and obfuscated, then an agent's ability to actually build out its agency mechanism is limited.
So I think where we're heading, what we need to have is the ability for more transparency, finer grains, attribution, and that stems from the ability to really establish a providential integrity across the board.
No, you never mind. You do stuff with those agents. You said they have more agency, but what is it that you help than the agents do? So we are hyper focused on getting agents paid and enabling their ability to pay.
What we've recognized in particular over the last 24 months is that as these models and subsequently these agents become more sophisticated and we say as human operators engage with these things and
Request, you know, more and more tasks be performed by these autonomous or semi-autonomous agents. There's going to need to be an ability on that recognition front to pay or get paid amongst these agents or reward and get rewarded, right? We kind of see these two things similarly.
But it's this concept of attribution, right? If I make a request to you, are you going to charge me? What are you going to charge me? And actually, how do I?
How do I pay you? And what we've in particular seen is just a very discreet need for payments rails for AI agents as this technology evolves and scales and becomes more productive in nature, right? You know, it started initially with stability AI releasing the diffusion models. So we saw the uptick
in commercializable work from these models initially on the AI art side of things. And then obviously, so that was like in the summer of 2022.
And then the following fall in November, open AI release, chat UBT, and that really kind of unshackled the technology and brought this groundswell of interest and with that developer capability to the table. And subsequent to that, we've just seen this massive movement of capable engineering
capacity move into the AI space, because before that, quite frankly, it was quite fringe. I've been doing this for a long time on the fringe, and people have looked at us like, what the hell are you working on?
And so what we really what? Statal diffusion from stability and and shatchy bt from open AI didn't just catalyze the understanding of what's possible and that led to the development first of more productive like fine-tuned or rag-tuned models and that has that is evolved into
the creation of like productive AI agents, right? And these are capabilities that you can task with certain objectives. And these agents are a composition of a bunch of different AI tools, usually more than one model, right? That
the solution, the agent itself can select based on the type of request that's being made to it at any given point in time, right? You have an example, what would be... For example, just from a very rudimentary standpoint, in terms of engagement,
The shift is like, if you used chat GPT in the early days, you were manually selecting which model you wanted to use as they were releasing more models, right? At first, it started with GPT-3, and then they released GPT-3.5, and then 3.5 Turbo and 4, right? But you were manually selecting through the interface which model you wanted to engage with. So that was sort of the first generation, right?
The second generation is 4.0 or 1.0, right? Where you submit a request and you're no longer having to manually select which model capability you want to invoke. What's happening is on the back end, your request is getting decomposed by that AI service.
and optimally rooted to the model that's going to sufficiently respond to your request. So a simple example is like, I provide a simple request to
this agent and it's going to root if it's, you know, if it's Sporo or if it's based on the GPT series of models, it's going to root the request to GPT-3 for that response, right? But if I submit a multi-modal request, that request is going to get decomposed and rooted to the model that can handle the complexity of that multi-modality, which is obviously going to go to GPT-4.
Now, herein lies the complexity because the cost of invoking GPT-3 versus GPT-4 can be like an order of magnitude and difference, right?
And existing payment systems are very static in nature. They are not set up to handle the variability in that cost mechanism, right? The way I describe it, Stripe is set up to sell t-shirts on the internet. It's a skew-based architecture. Each product has a skew, right? Small t-shirt.
medium t-shirt, extra large t-shirt, and what you do within that payments construct is set a static price per skew. So when I go to the web page and want to buy that t-shirt, it's 10 bucks. When you go, it's the same price, right? But in this scenario with an AI agent, I can make the simple requests to this agent.
And you can make a complex request to this agent. That agent needs to be able to handle the dynamic nature of that input. It does that by invoking variable services on the back end. What never-minded enables is accounting for that variability, by taking the metered cost of that service running and translating it into a settlement cost for the user. Yeah.
So if I have a simple, like, what's the capital of Kenya, then I don't need a complex model. Right. And it's basically cheaper to get an answer. And then it would be like a waste of resources to get a big model answering that. It could answer that, but it doesn't need to. So I get rooted through a smaller one and pay less. Yeah. Yeah.
Yeah, and then you start to issue complex requests, right? Like what's the capital of Kenya and book me a flight and accommodations and show me pictures of where I wanna go and things like that.
Right? So then you start to line up this series of tasks that are likely handled by discrete agents in and of themselves, right?
like a like a travel agent coordination agent and then a sub agent within that travel coordination function, which is like flight discovery, flight booking, flight check-in. You've got the, you know, the hotel discovery booking and check-in function, right? And each one of these can be compartmentalized.
into very unique or discrete tasks that one agent performs. And then the question becomes, how do you provide this sort of recognition of what each of those agents is doing? And in a simple economic environment, it's pure payment, right? So this travel agent, this travel booking agent, right, this coordinator,
is imbued with some funds and it needs to pay each of the underlying agents that are, it's gonna call upon to do some set of work, right?
This is where we step in, in recognition that there is a bunch of variability that goes along with determining what flight you should take, right? And how you're going to book these different services. There's decision making and logic and tool selection that's involved in all of this, and those have different cost structures. So enabling that
payments mechanism to be imbued within these individual agents that are performing these discrete functions and doing that in a dynamic and variable fashion. And because of this, because of the dynamic nature,
and the variability of these tools, existing payment rails are insufficient. So we need to accommodate for all of this in the sort of digital, emergent landscape with these silicon-based agents that are gonna perform this work for us. Yeah, it's different than buying a T-shirt, definitely, yeah. And it's Mark into like just saying, hey, internet, buy me some clothes that are gonna make me look good, right?
as opposed to me doing the coordination of trying to discover what I need to buy in order to make myself look good. Yeah. And yeah, this is basically the payment. I'm an economist, so I find it really interesting. And then the payment, you can basically have an agent and give it the payment function of Nevermind.
Yes, yeah, that is the role that we play, right, to become this dynamic payment mechanism for these AI agents. And really where we are setting our sites is in the future state, right?
We have a very, I think, a very sort of eclectic view on the world. It's becoming more emergent at this point. You know, six months ago, there weren't a lot of people talking about agentic AI. We're still talking about models and then this overarching concept of like artificial general intelligence or artificial super intelligence. But our focus has been on, you know, for the last couple of years,
specifically on this concept of AI agents, right? And what we believe we're witnessing right now is the rise of a new consumer that's going to manifest as trillions of AI agents.
and that these agents are going to change commerce forever. In our opinion, we're in a category defining stage, this category being AI commerce. So it's sort of an extension or an evolution of e-commerce where the actors that are engaged are no longer human carbon-based actors or solely carbon-based actors.
we're now introducing these silicon-based actors, these AI agents, that are going to not only perform tasks for you and I, but they're going to do work amongst themselves and collaborate with one another directly. It's this view of the world where you have trillions of these agents basically
creating their own economic universe amongst themselves, that we may or may not be largely ambivalent to, right? We'll see how this plays out. But yeah, that's sort of the end state that we are defining towards and recognizing the fact that like, you know, right now it's taking place, you have AI agent builders, they're going to productionize, industrialize, commercialize their propositions,
And what we see happening is this, you know, the kind of hit a wall when it comes to the payments piece. We see a lot of like, hey, this one size fits all subscription type model, right? And then for more sophisticated teams, they have to build this infrastructure, the system of unit accounting that does what we do.
right? That takes the metered costs and translates that into a settlement cost. So effectively what you have right now are all of these agent builders building their own payment rails, right? So imagine a world where you have trillions of agents all trying to transact with one another, negotiate, etc. And not only are they trying to negotiate on the price of the service or the asset that they're trying to buy or sell, they also have to negotiate which payment mechanism
is going to get used because Agent 1 has its own point of sales and Agent 2 has its own point of sales, right?
It's not very efficient. It's not the way things work. It's prime landscape for standardization or protocolization. Yeah. Everybody knows this. If you go to a foreign country and somehow the things you have don't work, your money doesn't work. If it's really bad, your credit account doesn't work. You go to Cuba, the US credit account doesn't work there. Exactly. Or another contemporary example, right?
You go to the UK, which is now cashless. And you don't know that. And maybe you don't have a debit card or a credit card. You exchanged whatever your local currency was for pounds before you got there. You can't actually interact within that economic or that financial landscape.
You're effectively an outsider, right? So it's this type of like paradigm that we're attempting to account for, right? That's the goal of what we're working on. To handle these different permutations of how each agent or set of agents is gonna transact amongst themselves, and then, you know, sets of agents transacting with other, or networks of agents transacting with other networks, that sort of thing.
I totally like it. There's this economic theorem, the coast theorem. It's about transaction costs. And if you reduce transaction costs, so much more gets possible. And that's actually a thing. If you reduce transaction costs, more things are possible because this is basically useless costs. Yeah. Yeah. Yeah. Like driving down the cost of engagement, right?
to as close to negligible as possible, but to a state where the network can still function, right? Because there's probably still going to be some form of taxation within these systems in order to maintain
the rails that all of these transactions take place on. But really what you're aiming for is to minimize those costs so that the money can be spent in making products better, right? Making services better. That sort of thing. I mean, this is where we see massive inefficiencies, right? So just for sake of example, Stripe released an integration with the Gentic, with AI frameworks.
the week before last, right? It's interesting. I mean, we have a stripe integration ourselves. So we have firsthand knowledge of how this is gonna work, right? They're basically replicating some of the technology that we've already built. But agents are, go back to this use case where you have like this travel coordination agent.
that's then distributing tasks amongst the network of agents, right? Any one of these agent services, whether, like, let's just say it's discovery of a set of hotels or flights that could be options for booking, right? That is, that function is not
very heavy, unlike the energy consumption side. It's not an expensive task, right? So the question then becomes, how do you account for that? And the direct answer is it's a micro pain.
Right? But then you look at what a stripe does from a business point of view, right? Their business is predicated on a commission, a transactional fee to the tune of over $5.
Depending on the region, but like kind of you know a blanket this is like 2.9% plus 32 cents a transaction, right? They try and facilitate micro payments under five dollars and five percent a transaction plus five cents, right? But that sets the threshold to what you can charge For these like what an agent can charge for its services to at a minimum of five cents and then it makes nothing
if you're using Stripe, because Stripe's gonna take that entire revenue line in its 5% condition. So as these agents manifest and scale and become more ubiquitous, micro payments are going to become the norm and existing payment rails are not set up to handle that type of transaction.
It's impossible based on the current lines of business that they operate and the margins that they extract. So either they have to cannibalize their business models and move into, you know, this micro payments, pure micro, like true micro payment space for
they're going to continue to operate in their existing domain, but they're gonna miss out on this new emergent economic ecosystem. And I tend to think it's probably gonna be the latter, right? I think they're likely gonna miss out. Some will gain entry, but most will be too slow to do it.
If you, if you see like just to remind the listeners that if there are trillions of agents, if everybody, every transaction between, between them has to cost at least five cents. That is, yeah, it's, it's untenable, right? Like just this, just the service and network were into tens of billions of dollars. And that doesn't seem to make a, I mean,
Look, from a business operator point of view, that'd be great to take that transactional story. But the reality is I don't think that's going, I don't think that's realistic. And the reality is also there's technology that exists that pushes that cost closer to the nominal value of zero, right? That technology being watching technology, right?
Perfect. That was the next thing I wanted to ask you about. How was the technical background of that? Yeah. So this is a conscious decision that we made initially less in part, less from a
Economic or, you know, a finance point of view, the payments, the innate payments of a blockchain or settlement capability wasn't the primary driver. Initially, the primary driver was, you know, the high fidelity providence.
that can lead to fine grain attribution, right? Blockchains are very elegant solutions when it comes to providential integrity of all of the assets and contributions made towards a specific outcome, right? So you make a contribution, I make a contribution, a thousand other contributions are made. We have, there's a record of all of those contributions and when that output is commercialized,
you can take the commercial value and potentially redistribute it, re-attribute it, back up that value chain in a corresponding, maybe preset royalties or residual structure or what have you, right? So that was the initial driver to leverage this stuff, just because if you are familiar with NL,
analytical systems, AI now, specifically, you know, these agents work not just because there is an algorithm there that's sophisticated. It's the algorithm plus the contextual data set, right? First, contextual or topical information that's used for training purposes, and then contextual information that goes into these models, these trained models for inference purposes, right? So you have to bring
the algorithm and its computational environment which forms the model plus the contextual data set together to get like hyper relevant context and therefore, you know, accurate inference.
And so we were looking at this type of amalgamation, these the combinatorial structures where you have to bring these things together and how do you influence
the reward system or the payment system behind that. And what we realized is in actual fact, the most critical piece to start with is enabling the payment site, right? If you can capture sort of the, it's that last mile. If you can, if you can capture the value of that output, right, and you provide the proper infrastructure to capture that end state value, then you can propagate that
value back up that work stream. And so that's the underlying technology that we're using, namely, you know, blockchain infrastructure in order to accomplish those goals, right? And then it just so happens that, you know, we prescribe to open source sort of ethos and philosophy
which is, you know, one of the primary drivers of the decentralized landscape and making these systems open and ubiquitous, you know, where you kind of see
our underlying drivers and alignment to the decentralized space is in kind of our outlook on things. So imagining this world where we have trillions of AI agents transacting with one another and potentially with you and I, we look at, we believe that payments are the singular choke point.
in this environment, in this landscape, right, for these agents. If like an open AI Microsoft, or a Google, or a Facebook, or Bay, monopolize that payments mechanism, if they control it, right, their ability to shut off one agent becomes an existential threat to all agents, right? Microsoft says, ah, this agent competes with one of my lines. Business, booth, I'm gonna deplatform it, de-bank you.
I mean, it doesn't matter how that agent is constructed, whether it's using centralized or decentralized technology, right? If it lacks the ability to pay and get paid, at least from an economic point of view, it might as well not exist, right? So providing optionality to these AI agents to always have an option to pay and get paid, that's the motivation.
Yeah. Totally makes sense. Yeah. I think economists are like, yes, yes, yes. Exactly. That is like, there's certain other motivations of humans, but basically we work for money and that's why, yeah, we could have done different things today, but no, we go to work because the more we have to pay things, that's, yeah, yeah, definitely. We work for some sort of acknowledgement that that work has been performed at a satisfactory level, right? You know, our
Our construct for that is like salary. But at the end of the day, it's the fundamental ability to capture that reward and then leverage it. And if that capability is centralized,
then if someone or something doesn't like you for some reason, they can effectively take that reputational value, right? It's a discrimination. Exactly. And we see that.
You see it in social media, and I think a lot of what's going on right now, you know, across X and other platforms is this sort of kind of digital tug of war is taking place kind of under the, at least in North America, like the guys of free speech, but really it's like,
reputational capture and, you know, what is, what is kind of arbitrarily defined as good and bad, right? And, you know, there's lots of implications to all of this and I'm not going to sit here and, you know, it's all talk about it like I know what the Alan talking about, but, you know, I do see like it is, it is kind of, you know,
It's representative of sort of like where we're heading, right?
It's a microcosm, a sort of discussion. With AI agents that should do work, but get discriminated. Yeah, exactly. I want to jump in. One of the things we talked quickly about, blockchain, and I wanted to just say that is not Bitcoin, and I think we have to make a point. No, no, that's exactly right, right? Like this is...
Um, a mechanism to capture that productive value that an agent is producing and then enabling like the settlement based on that production, right?
So it's a mechanism for recording the value creation and then recognizing the attribution of that value creation, right? This isn't a store of value per say in like the classical Bitcoin or like stablecoin type argument. What we're representing
is the value that this agent is creating, right? So the token or the coin or what we call it, we call it credit, is a representation of the productive value of that agent.
or set of agents, because multiple agents can basically choose to use the same set of credits to do this system of unit accounting. Yeah. So it's simply just a payment unit and not a value unit. So I think that's important. It's not about speculation here. It's just real. They have to pay. And basically, if there's no inflation or whatever, it stays the same.
this is exactly it's it's it's effectively capturing yeah it's the proxy for that value creation right represented on chain and so it's you know it's it has fidelity to it right it could also represent the margin that's baked into that service but it's transparent right
and therefore it is by extension a function of that agent, right? It's not a function necessarily of a broader ecosystem. It could be if that entire ecosystem decides to use one set of credits, but you know, that's not, we don't believe that that's how this thing will evolve. What we think is gonna happen is that these agents in the future will basically, trillions of them will exist and appear to appear bartering.
environment, right? Where each agent can self-discover other agents and services and assets that it wants to either consume or sell to, right? And then directly or indirectly by spinning up replicas of itself or what have you, reach out and negotiate with these other agents, right? And instead of paying some proxy,
currency, for example, it just trades directly in those assets, right? So, you know, D bar, I want, you know, your asset A, I have asset B, I'm going to give you two of my asset Bs for one of your A, right?
And it's that type and it's just this direct bartering trade that we make. That's our future vision. And so now we're in a scenario where we need the scenario plan back to where we are today. And we look at things like, okay, if this is the future state, each of these agents is discrete in nature. It offers a unique service because it's coupled directly to that unique or discrete agent. And therefore we can replicate that now with unique
tokens as a proxy for that agent and its asset or service or generally its productivity. Oh, to take this the whole thing down again, because at a certain point we have to. I could, this is because it's going in the economic thing. I could ask so many questions about it, but I wanted to know a thing about you don't. How do you personally in daily work, live private, live use AI?
Yeah, so I mean, I use it to like expedite my output, right? So pumping stuff into chat GPT. Asking perplexity about stuff to get responses that are, you know, more aligned.
or arguably more useful than getting a list of websites back and then having to go through and discern which content is relevant. It's competition to Google that basically you type in a question, you put your prompt, and it gives you a more precise
answer through summarization of different sites and source material. So you can use it for search across websites, if you're looking for a particular product, but you can also use it to summarize news and things like that. So it reduces the surface area of the internet.
and requires less engagement from the requester, right? Or less manual intervention and provides, you know, I think better feedback from this massive engine that is the Internet.
And that's what these tools are generally doing. The other thing that we're actively pursuing is leveraging these agentic workers.
service response agents. So they're like call calling agents, for example. There's a number of these that are popping up where you can provide a bunch of company information and it will rag tune. So provide that context to the agent. It'll learn your business and then it'll identify from its list of tens of thousands, hundreds of thousands, maybe millions of potential clients.
which of those clients you should actually reach out to, and then draft emails, even send those emails out, and now you're getting agents that can make phone calls as well.
Oh, I got a really bad one just funny from Portugal. It was like this record advice. And it was like, there's so much more that you can do to stay on the call and tell you something. There are there are some relatively sophisticated one. Like you can tell there's a buffer flag, right? But.
These things are very conversational, quite topical in context. Anyway, from our side, this type of engagement is a bit more exploratory.
at this point, just determining really like what works for us, but also what generally works within the domain space of AI agent workers, right? Because that's going to inform us kind of the next position for our organization in terms of what types of agents are actually useful.
Because that will inform us which ones are going to get paid, right? Yeah. That leaves just one thing. If the agents take over power, so my last question is the Terminator or Matrix scenario, how do you do you think about it? Is a problem for us, humans? I mean, maybe we're already in the Matrix. I don't see this like...
war of the world's scenario taking place. If for anything, it's like, it's a horrible use of energy resources. But yeah, so my kind of
almost malistic cake on this, there's a high likelihood that we get domesticated. But I'm also gonna argue that that process has already started. You know, going back to what we were talking about before, with algorithms for social media and commerce that already exists, like this isn't
a domestication function, right? I think that's likely going to persist. But I think it will also persist largely for our overall benefit. I think what it's going to do, like I do think there is a major threat to, in particular, like white collar knowledge work.
So I think that that's going to, there's going to be like a real shift in across industries. But I'm hyper optimistic that that's actually going to unshackle a lot of like human creative endeavor, right?
Um, either on the artistic side or on the scientific side, right? It's just going to free us up to more exploration. Um, and I, and I'm very confident that in our, in humans ability to adapt, right? I think that's our sort of our, our, uh, best and most fundamental trait.
is this ability to adapt to circumstances and do so within that time continuum of the universe very quickly. So I'm very optimistic about this. I think it'll be like other revolutions, right? Industrial revolution, digital revolution. Everybody bemoaned the, oh, we're gonna lose our jobs. Yeah, well, it opens up avenues for all kinds of other
creative input and outputs. And yeah, I think it'll allow us to explore ourselves in the universe, probably significantly more.
That's a positive note to end the podcast with because I really like that that we will get freed and can do what we like and the works done somewhere else. But tell the people where can they find, never mind, and where can they find you? I will put even the show notes for the people who want me to write. And if you want to track me down and docs me, I'm in Lisbon.
But yeah, you can find us. The best place to go to start is probably our website. Never mind N-E-V-E-R-M-I-N-E-D dot I-O. From there, you can jump into our Discord. You can follow us on X. Yeah, we're, you know,
We've been doing this for a while. We're in this for the long haul. If you're interested in, in particular, the commercial aspect of AI agents, we've been thinking about this probably longer than almost anybody else. So we'd love to have you in the community and talk more about this.
Perfect. And thanks for the great interview. I learned a lot. And yeah, then we'll see how the future of AI commerce is and how many agents will have and when this future will happen. Thank you. Thank you.
Yeah, thank you, Don. That was a great interview. As an economist, I really loved it because I think that the economy of agents is something that is still following the same economic rules. And yeah, I learned a lot about how to do that. Yeah, it was deep. But I think we all got a lot of information from that. So I'd say, look, what comes there with the AI agent? Don't miss that train. It's really important. There's a huge development coming.
AI agents are the next big thing, in my opinion. And also, don't forget to follow the podcast or go to the newsletter. It's argobelline.com slash newsletter. You can get all the interviews and some AI information from there. Thank you for listening to the podcast. And until next time, signing off stigma from argo.belline.
