Signals Through the Noise: People Not AI Agents

[00:00:04] This is the Crypto Hipster Podcast. This is not a traditional interview show. These are perspective-driven conversations with founders, builders, and independent creators shaping what comes next.

[00:00:26] We go beyond headlines, beyond hype, and beyond price to explore ownership, freedom, and opportunity in the digital economy where builders talk freedom, not price.

[00:00:47] The crypto industry spent years making things more complicated than they needed to be. Now, everyone wants to simplify everything, but systems and systems issues don't disappear just because the interface got cleaner. So, the question becomes, what are we hiding from users when we simplify infrastructure?

[00:01:16] Today I have Kareem Abdul-Rakman with Simplify Labs. Kareem, I want to find out first, what parts of crypto infrastructure should never feel simple? I would say that if we're talking about what should feel simple or not,

[00:01:36] everything should feel seamless in terms of the user experience or the experience that you have or rely for the end user. So, I would say everything should be simplified in terms of the end user experience. However, when it comes to a security level, this is something that should not be simplified for sure in terms of your company or the company that you're working with.

[00:02:05] So, you said exploits. So, where do founders accidentally create fragility by abstracting away too much? The crypto itself is not the hardest thing in the world or not the hardest technology that is existing in the world.

[00:02:26] That is for sure. If you dive deeper into it, you do understand that it's not that hard. It's not that complex. It's pretty simple to develop or to use. So, the end users always should be simplified as much as possible because the retail clients, the end users, they got used to the products that are existing on the market for tens of the years, for the decades.

[00:02:49] And that is why it is important to make them feel simple. It is new, but it is simple.

[00:02:58] But when it comes to the founders, you need to pay attention to the possible pain points that might be as a critical risk in your businesses as a vulnerability of the smart contracts, the liquidity pool that you have inside or the provider that you're using as a liquidity provider. For example, the system as a whole, the data storage of your users, etc., etc., etc.

[00:03:27] What you're saying is that on the founder side, complexity is necessary, but on the user side, simplicity is necessary. Yes, yes. You cannot name yourself with all the respect, of course, but you cannot name yourself the founder without totally understanding where you are, what you're doing and how complex your product is because you should understand your product in and out.

[00:03:52] In order to make sure that, yes, the funds of your clients, the volumes that you are processing are 100% in safe for your users because they are relying on you. When users are doing some sort of a financial transactions, they are relying on your system, they are relying on your company, they are relying on the people that are working with you, right? So, yes, you should understand all in and outs and that's complex. That's when it becomes complex.

[00:04:19] Got it. So, have we trained users to confuse convenience with safety? We go in there and that is why on one hand, we have a lot of talkings and actions that are connected to a licensing on different type of markets like Mika, on the territory of the European Union, Vara, etc., etc.

[00:04:42] Yes, that is all about the safety. That is all about the safety starting from the point where you as a user cannot trade more than you particularly could be trading as an employee and ending a different type of regulations in terms of the layer,

[00:05:07] how you store your funds, how you store the data of your users, how you manage the funds of the users, etc., etc., etc. That is lying to the non-chain staking, for example, that was really, really famous so-called back in the days. So, it's not the on-chain staking. So, you actually do not delegate the funds to the validators in the network.

[00:05:33] So, yeah, I think that the regulation is the big step to the knowledge and learning of the end users and how to be safe in the new type of the finances that we are in right now. Stacking of all, there are a lot of valuable resources where you actually read or watch about what was exploited recently,

[00:06:04] how users are acting with their own funds, how you should double-check the wallets that you're sending the funds to, how you can be hacked, and so on is a call. So, yeah, I think that we are still on the very early stages of the knowledge base or education level of the end user and how to keep safe on this market.

[00:06:27] But, yeah, we definitely go in there with all of this type of the travel rule where it will be more similar to the classic banking system when you send the funds to, for example, I would be sending funds to Jamil. Now, I'm not going to be only sending the funds to this particular crypto wallet, but I'm going to know who's behind this crypto wallet. So, I'm sure that, yes, there is Jamil behind this crypto wallet and I'm safe in that space.

[00:06:54] You know, trying to expose system tension, what do regulators misunderstand about how systems actually operate? We actually was, when I was talking with our legal team in regards to all possible legal frameworks in regards to the MECA particularly, I said one actual point that the regulators are trying to mimicate the banking system to the crypto niche,

[00:07:22] which means that they want to create a new banking system and they want to apply those rules that are actually existing for hundreds of the years. So, in the banking system, in the banking system, through the crypto niche. That's a bit not correct way from my point of view with all the respect to the authorities on how this should work because that's a new era of the e-money.

[00:07:47] And there are new rules on how to operate with them. So, you still need to be more flexible in terms of how the technology was developed, what it was developed for, and how users are obtaining this technology, like for which particular reasons.

[00:08:08] So, yeah, that actually makes hard to interact with all of the rules and requirements that we see from the FCA, for example, that are requiring those type of regulations from the MECA compliance companies. Yeah. Got it.

[00:08:33] So, that's not a one-way, so this is not, so that it's not one way and it's reciprocal. Where do builders misunderstand compliance? Well, the builders want to simplify everything, right? So, they are about completing the task. Let's put it this way. Compliance is all about constant and permanent risk that should be avoided in any type of the operations.

[00:09:02] And that fighting creates a lot of misunderstanding between both parties. So, the compliance wants the system to be absolutely reliable to all of the existing rules that are there from any regulation, because you might enter the new market tomorrow, right? So, you need to follow also these regulations.

[00:09:24] And the developers are trying to simplify the system itself and not to make it overcomplicated in terms of the usage as well. Because you still cannot or it is hard, at least, to build an absolutely hard system, like complex system,

[00:09:48] without making hard the life of the end user. Because you still need to click a lot of buttons, you still need to check a lot of check boxes, fill a lot of forms. And that's all that you have in terms of the requirements from the compliance team. So, yeah, you need to find a common sense in it. You need to find a middle spot between those two fighters, so-called.

[00:10:16] And to still not to leave the idea of having a great and simple to use. Product for the end user. So, even though you have a lot of regulations around it. Got it. So, in your view, can infrastructure still remain permissionless once compliance gets embedded deeply enough?

[00:10:45] Yes. Yes. Utilizing all the existing tools, like, for example, Notabana with its transaction monitoring, KYC layers, AML layers, plus a little bit of the AI on the top. Yes, you can make it. You can actually make it seamless. I've seen one company that was founded in Finland by two founders.

[00:11:14] They've made this happen. Yeah. They have a couple of compliance officers for the whole spectrum of the clients that they have. KYB, both corporate clients, individual clients, because they have a lot of external tools that are plugged in into the place. Plus, a little bit of the AI that helps them to rely some of the work, like, basic work job on the AI.

[00:11:42] And the compliance officers are only triggered when some sort of alerts coming up. Yeah. That is possible. Got it. So, let's talk about the user's perception of your system, right? So, invisible infrastructure causes, you know, consumers, like, you know, I'm wondering what part of the invisible infrastructure stack fails the most? And the thing with the blockchain is that blockchain is up and running 24-7.

[00:12:12] So, some client may send the funds by the, I don't know, New York time when you are already asleep somewhere in Europe, right? So, that's when you need to make sure that you synchronized all the nodes, that you catched all of the transactions that were paused, so-called, during the downtime. And you credit or execute all of the transactions that were under the downtime.

[00:12:41] So, you mentioned earlier, you mentioned AI. And it has an impact on the ability to simplify things. So, does AI simplify infrastructure or just hide operational risk? Yeah. Yeah. I would say that AI is more about operations. Yes, AI is pretty decent in terms of the development, empowering, but it's not about the simplification. No.

[00:13:11] I mean, like, we still have engineers. We still have bright minds that are doing the simplification for the clients that we have. AI is just a tool that helps you to achieve some certain goals or simplify a certain tasks that you have. Like, it's just speeding up your routine. I think we've seen what happens when people interact with systems that they don't understand.

[00:13:38] But what happens when AI agents do it automatically? Oh, no, no, no, no, no, no. I think that we are not on that level of... Like, I would not trust AI agents when it comes to a fintech business. That is for sure.

[00:13:59] Like, they may interact with some data that is related to some sort of analytics or data set that you have back there. But they cannot make decisions on themselves. That is 100% red flag. Like, I haven't seen that yet implemented.

[00:14:22] Honestly, I haven't seen any company yet implement an AI agent that would actually take a responsibility and make some actions. Especially when it comes to the financial operations. But it's a red flag. It even sounds like a red flag because you never know. So, could automation increase system fragility instead of reducing it?

[00:14:53] Well, it depends on the automation itself. Like, what are we trying to achieve with automation? Where we implement the automation? Like, I guess you cannot automate everything. Like, it's just really impossible because there are actually some cases where you need the involvement of the real person. Real human being. Like, you can automate a KYC process because you have all the necessary plugins, tools that help you to, I don't know,

[00:15:20] search the data of the user in terms of all the databases that you have. Like, government IDs, right? Check whether this client is AML or not, PEP or not, etc. But when it comes to the, I don't know, red flags that might appear in the dashboard for your compliance officer, that's when human interaction comes in place. Like, you cannot leave it on the system. There are some cases.

[00:15:48] When, for example, the limit is increased. So, your monthly spend limit is increased. How? Yes, you do know that what type of the documents you need to request from the user in order to increase this limit or review this limit, whether he's applicable for it or not. But it's still on the human to check it. Got it.

[00:16:13] So, one thing that I keep thinking about, you know, in infrastructure was that simplification doesn't eliminate complexity, right? So, it just moves it somewhere else. Right? So, I'm just wondering, as a builder, you know, what part of the infrastructure work for you do you feel is the least appreciated in what you do? I would say back office. Yeah. Yeah. That's like, that's when you take care of all the information about the, I don't know, client profile, his financial activity,

[00:16:43] any sort of data that is reliant to the client and his activities on the platform. There is for sure the most underappreciated part of the system that we have in terms of the client, of course. But our ops are loving that we do take care of it because without that, it makes you, it gives you a lot of a headache to manage the user, his activity, some sort of a data that you have in terms of that user. Yeah.

[00:17:12] That's for sure should be a core part in your, I don't know, puzzle of the systems that you have inside. But a small percentage of the people are thinking about it. That's for sure. Yeah. Sounds to me like that may be where all the hidden complexity lives. Yeah. Yeah.

[00:17:35] I mean, like, well, it's a bit tricky here because ops team would always want more and more functionality in the back office. And that's when you even need to have maybe a product manager who is going to be managing only the simplification of the back office even. Because they would send you tons of requests on the back office. And they would have a new type of data visualization in the back office that they want to have.

[00:18:03] But if you implement it just the way they like, it would not be possible to read or use. So, yeah. You still need to tell them, hey, guys, let me help you. Of course. Yeah. Cool. You want to have this and that. But let's put it this way. You already have this information, but in a bit different way over here. And we can put this information here.

[00:18:28] And we can just expand the data set of the user profile here. So, we're helping them simplifying that as well. That is for sure. But, yes, it is easy to go deep into the rabbit hole when you're trying to get more and more into the back office features. Yeah. Yeah. So, a lot of people, a lot of users see things in a system. They're like, they think it's permissionless. It's automatic.

[00:18:57] It just happens, right? But how much of that infrastructure today still depends on humans quietly, silently fixing things behind the scenes. Like all parts of the system. Look, let's take a look on the big fintech companies, right? Like big tech, like, I don't know, Revolut. There is, everybody's screaming about the Revolut. Recently, they are huge.

[00:19:24] They have over, what, 10,000, 4,000 employees all around the world. And the majority of those employees aren't based in their head office in the European Union. All the companies are still relying on the employees. So, employees is the main and major artifact of any company that wants to succeed on the current market.

[00:19:52] You, all those stories, again, with all the respect, I'm not trying to be skeptical. I'm not the, I see this happening. But let's face truth. We cannot see yet a company with one founder and a lot of AI agents that make a big business. That is for sure, especially when it comes to, I don't know, transactional business. Like, you cannot do it.

[00:20:19] The smallest company I've seen that still had 15 people on board. Of course, that's a transactional business. They had 10 to 15 people on board. But that's still people, right? And their whole system is plug and play. So, they are relying on the majority of the big players.

[00:20:39] So, if you count, there's still a lot of human interaction with the systems throughout the product itself. Our example. You still need DevOps.

[00:20:52] You still need developers even to maintain the system, to make sure it works, to run auto tests, to develop some sort of features that would allow the consumer to use more and more your product, to bring more on the table, more features, more technology. Because we are not stopping here, right? The world is evolving. The market is evolving.

[00:21:22] And you need to keep up with this pace. You need sales. You still need customer success. People prefer people. Like, we, I think that the portion of our success, current success and current growth lays in the ground. And the fact that we talk a lot with the clients person to person, just casual talks, try and find out what do they need? How are they going with their business? Even if it's not about our services.

[00:21:51] I just want to find out how we can help them, how we can be useful for them. And AI cannot do it in the manner that our human doing that. So, support as well. Yes, you may have knowledge base. You may have AI automated support agent that is replying to your, to 9 out of 10 requests that you have.

[00:22:16] But you still have this 10% of the requests that should be done and processed manually by the client because it's out of the scope, right? It's unusual case. And that happens. The more clients you have, then you will figure it out that the more edge cases you have. Funny story. We recently released a new product devoted to the cards ecosystem and the software.

[00:22:44] And we were doing internal tasks. The software itself was working perfectly. Auto tasks, perfect. When we released it to the big player on the market, damn, the first 24 hours, it was a nightmare for us. Because the number of the edge cases that we've seen from their clients is amazing. And I love this because you cannot make it in the lab.

[00:23:14] You cannot make it in the lab. You can feel it only in the real life. So, yeah. AI is good. Once again, AI is a tool set to empower your business, empower your daily tasks. It's on all the layers. But business is still about the people. Let's talk about those edge cases. Because I see a risk here, right?

[00:23:42] Say you simplify things so much where you reach a point where nobody fully understands the stack anymore. And then you have edge cases. What are the risks there? It's impossible. It is simply impossible because you still have an ownership on all the layers.

[00:24:05] If your system is built up on top, it consists of a lot of microservices. For example, you have internally the people that are managing those microservices. So, you have an ownership, first of all. And it's not a simplification in terms of writing a code the way that nobody would understand it. Simplification is about making your daily routine easier. Just let me give you an example.

[00:24:34] So, we talked about simplification on the consumer level. And that's understandable probably for all. But when it comes to the themes, like internal simplification, that is when the magic comes in. So, yes, you can use AI tools like cloud to code, for example, for the coding skills. Simplification comes in place for the development team, for example, when you need to create some sort of use cases for the new feature that you are developing.

[00:25:03] So, you simply upload those lines of the code that you are developing. You're asking AI to analyze it and create all possible use cases that might happen in terms of that feature. Once he's completing that, you're saying, okay, now you have some sort of... And this AI agent eventually has the background of the products that we're working with.

[00:25:33] And now you're saying, hey, now I want to implement this feature in this part of the system. Now, create me a full flow of the user plus the use cases. Take your time. It's important to tell him. Take your time. You can think as many minutes as you want. But give me a complete and extended list of the use cases and the flow scenarios when this feature is now implemented into the product that I have currently.

[00:26:03] So, that is simplification. Back in the days before the AI, that would take you... So, you know this probably that you may develop the feature for three to four days, but then you may test it for seven to eight days. So, testing period is always bigger than the development period itself. So, now we're coming at even.

[00:26:25] So, now the testing period is sometimes shorter than the development period just thanks to the AI tools that we have that are helping you with all the type of the use cases. You don't need to make it manually. Then you have automation. You have automated QA that is just a run tests that are doing the mimic tests on all of the features that you have.

[00:26:50] And only then you have human interaction as a final point of making sure that, yes, everything is flawless. So, yeah, that is why I think that it's just simply impossible to simplify it the way that we would forget about it. So, again, you should use the simplification in the way where you now understand, okay, last time I had to spend eight hours on it.

[00:27:19] Now, I'm spending 7.5 hours on it. That's a simplification when it comes to the internal policies. Got it. So, there are thoughts. The more you simplify things, the more you take, the more you hide complexities, people will misconstrue that as the system is now centralized instead of decentralized. How do you bridge that misconception?

[00:27:48] Well, centralization and decentralization is totally two different types of businesses, I would say. And you have two different types of consumers for this type. So, for the decentralization, there is still simplification because decentralized system in the crypto is not for the newbie, I would say.

[00:28:11] And even there, like when you, again, look back in the days using decentralized systems, systems would be much more harder than using centralized. Example, Metamask. Metamask initially decentralized wallet that is now having a lot of the knowledge base and the tools and the features for the consumer.

[00:28:33] That's when you see that decentralized system simplifies the way so the end users, the consumer that previously was the client of the Revolut, for example, is now using Metamask with its card, with its simplest flow of the exchange, deposit and withdrawal, storage of funds, and etc.

[00:28:55] Centralized is not about the ownership, so-called, or not about the simplification, but it's about the business model, let's say. Centralized business is when all the private keys, let's say, are storing at one hand and you use this service as a banking system.

[00:29:17] Because, again, banks do have an access to your funds technically, but they don't do it without your permission. The same with the centralized fintech. Yes, in the centralized fintech, the private keys are stored in the one single place that is absolutely safe,

[00:29:38] but the system itself cannot execute any type of the actions without your permission. Moreover, centralized is much more simple in use for the end user still, for the consumer, rather than decentralized. Because once you lose, why there is a centralized and decentralized?

[00:30:06] When it's centralized, the only thing that you think about or that you need to remember is your pin code or your password from your account. When it comes to the decentralized, and you still can restore your access when you have your email, right? When you have an access to your email, you can still message the support, you can provide your personal documents saying, hey, I don't know, my computer was hacked. My email now is hacked.

[00:30:34] I don't have an access to any previous personal information that I have, but I have my passport. I can send you the photo of me and my passport to you, which gives you an idea that it's me, and I'm just trying to access my, just restore my access. With the decentralized systems, it is simply impossible.

[00:30:55] Yes, there are MPC wallets that allow you to break the private key on three, five parts, but still the majority of the decentralized systems is one way. If you lose it, if you lose your private key, you're done. Like, there's no way to restore it. And when it comes to the big amounts, it makes sense. So, yeah.

[00:31:20] But for the centralized exchanges, there is also a big risk about that because they store your funds at one place. So you can imagine one non-custodial wallet storing funds of all of the clients that are there as their consumers. So it's only about your own choice for the safety of your funds. Just a cold wallet.

[00:31:46] You just store it there, and you just make it live only when you plug it into your PC for the quick transactions, for the day-to-day transactions. But for the time, it gets the more progressive systems. You know, yeah. What's one thing that you think that the industry keeps trying to simplify that actually requires people to think harder instead?

[00:32:16] Hmm. People, you mean consumers? Yeah. Instead.

[00:32:31] I would say that it would sound maybe strange, but it's all those requirements that you need to follow in terms of the regulation.

[00:32:47] Because there are so many steps that you need to complete before doing any actual target action on the platform that you really start to think about, like, why do I even need to mention this information or to do this step? Like, for example, you simply want to receive a deposit from the reputable platform. And that's an example of the Revolut, for example, or the OKEx.

[00:33:17] They have so-called Satoshi test. If you are new to the crypto niche, you would face some hard points when you complete the Satoshi test, honestly. Because, like, yes, someone might find, like, first of all, first distraction. I receive funds from my wife, for example. Just an example. Can you imagine? I'm a newbie in the crypto. I'm just trying to test it out.

[00:33:45] I'm like, hey, Jamil, send me some funds. I don't know. I've heard about, like, the USDT, USDC. I want to buy some coin right now. It's hyping up. Like, I really want to go for it. You say, yeah, no problems. I'm giving you some cash. We're doing some sort of OTC behind the desk. And you're sending me $1,000. And now I'm registering the Revolute because Revolute is a banking app. Like, everything is convenient for me, right?

[00:34:14] I'm not talking about the OKEx right now because OKEx is more for the crypto geeks. But we're talking about, like, the newbie. So, yeah, it's the Revolute. Now I'm receiving the crypto. So, first of all, I need to click on a checkbox. So, I need to go past KYC, blah, blah, blah. That's done. Now I need to go specifically to the section where it's about the crypto.

[00:34:39] I need to read all the disclaimers about all of those risks and everything that I'm taking on my end when interacting with the cryptocurrency, which makes me feel that it's safe. OK, I click on a checkbox. Yes, I do agree with all of those risks and disclaimers. I have read and accepted terms and conditions. Now I do want to receive this $1,000 USDC from you.

[00:35:07] And my incoming deposit is frozen. Why? I have no idea. We need to make additional checks. Oh, my God. $1,000 is not that much. And you still need to make those checks. Fine.

[00:35:21] Then Revolut asks you to complete the Satoshi test, which forces you to make exact same transaction in the exact same network with the exact same asset, which can be USDC on Solana, for example. But for the small amount, let's say 0.01 USDC. And I'm talking to you like, Jamil, I don't know.

[00:35:49] I have this notification that I need to perform some sort of a Satoshi test. I don't know what is it, but they're asking me to send 0.01 USDC or one USDC to this particular wallet once again. Or to a different wallet. Usually they're asking you to perform this transaction, like Satoshi test, to a different wallet. Not your deposit wallet, but different wallet that they have specifically for this type of the test. And now you're saying to me, dream, but I cannot do it.

[00:36:18] I have centralized system.

[00:36:23] I have a centralized exchange.

[00:36:51] And the centralized exchange that I was using to send those funds from was not allowing me to send 1.3 USDT that they were asking me to send. So I ended up speaking to the support 72 hours to resolve this issue.

[00:37:15] That's when the distraction, that's when you try to make everything more safe and simplify. But you eventually making it much more harder for the end user to live with that. 72 hours to resolve $1.30. Wow. That doesn't make things simple at all. The amount that was frozen eventually was something around $2,000, let's say.

[00:37:44] But I could not perform the Satoshi test. That is an issue. But can you imagine if those $2,000, I would need urgently? Like, I don't know, to pay for the bill, to send to my mom, for example, etc. Pay for the school of my kid.

[00:38:02] So it took them 72 hours, a bit more, 76, like three plus days to finally credit those funds to my wallet. But you got your $2,000 eventually. Sorry? You got your $2,000 eventually. Yeah, yeah, yeah, yeah. I do have it because I'm a fighter. I'm a fighter. Got it.

[00:38:32] There's a difference between making systems easier to use and making people understand them less. And sometimes those two things move together. The modern internet normalized invisible systems. Most people don't know how banking works. Most people don't know how cloud infrastructure works.

[00:38:59] Most people don't know how algorithms shape what they see. Crypto originally promised something different. It promised transparency, participation, and verifiability. But if every layer becomes abstracted away behind clean interfaces, AI agents, automated compliance, and invisible infrastructure,

[00:39:27] then eventually users stop interacting with systems consciously at all. And at that point, convenience may have won. But understanding may have lost. An error, we've seen the way around using the camera, so that we are a bit more than a ultimate effect. And at that point, we will have a good additional update.