Tomorrow's Medicine, Ep. 1: Oscar Uzdelewicz - Computational Neuro & BCIs

With Oscar Uzdelewicz, Computational Neuroscience Researcher. Hosted by Arseniy Arsentyev. Published April 3, 2025.

https://aarsentyev.com/podcast/ep1-oscar-uzdelewicz

Machine-transcribed and lightly edited: filler words and false starts are removed, and the wording is otherwise as spoken. Where the two differ, the recording is the record. Times are minutes into the recording. Not medical advice.

[0:00] Arseniy Arsentyev: Welcome to the first episode of Tomorrow's Medicine podcast. My name is Arseniy and today I'm going to talk with my friend Oscar Uzdelewicz. Oscar is a sophomore at NYU studying mathematics and physics. However, his true area of expertise is computational neuroscience. Oscar is an extremely intelligent and curious individual. Today he'll tell you about what computational neuroscience is, some of his research, then we'll talk about brain-computer interfaces and discuss broader picture of developments in medicine. I'll hope you enjoy this episode, so let's go straight to the conversation.

[0:47] Oscar, what exactly is computational neuroscience and how do you believe it fits into the medicine?

[0:55] Oscar Uzdelewicz: Okay, so there's two ways to interpret the term computational neuroscience. The first is using computational methods to understand the brain. So what that would mean is like using traditional data science techniques, like kind of simpler linear models, or all sorts of like inference, like statistical techniques to understand neural data and try to draw some more statistical conclusions about it. Which is more, that's more of like applying existing methods to the data we already have. And kind of more of like like the more of a progression of the existing body of knowledge in neuroscience. But I think the more interesting way to interpret computational neuroscience, and I guess this way is also called theoretical neuroscience, is using math and physics methodology.

[2:09] So this kind of bigger idea of system modeling and understanding chaotic systems, one of which is the brain, right? To think about what are the underlying principles of computation that take place in the brain? And how do those interact with the biology of the brain?

[2:40] Arseniy Arsentyev: Yeah, that's a great definition. So you're a physics and mathematics major, is it right?

[2:47] Oscar Uzdelewicz: That's right. That's correct.

[2:49] Arseniy Arsentyev: Could you tell us about why you decided to switch from being a neuroscience major to pursue physics and mathematics, and particularly like hard theory of science? Why do you believe that for people who want to build their career around neuroscience and probably around medicine, it's worth to put some effort into a theory part?

[3:13] Oscar Uzdelewicz: Yeah, that's a really important question. I'm glad you asked that. So if you're learning basic neuroscience or not, but traditional neuroscience or traditional biology, like let's say that's your major in college or that's what you're studying for your PhD, from my experience at least, you don't get a great mechanistic understanding of the systems you're studying. So what that means is you basically, you learn a lot of facts about how they operate, what the what these systems look like from our current body of scientific evidence, but you don't really get to model these systems and get a intuitive feel or a precise feel for how they work.

[4:12] And that's what studying math and kind of applying math in the form of physics really gives you, because mathematics is all about precision. So when you apply those kind of physics-based methodologies or like the first principles way of thinking and mathematical way of thinking from physics to the brain, you have a lot more power, though it's a lot more difficult because you have to be precise. You have a lot more power to actually model these systems. And I think in the long term, these physics-based methods are going to, if not completely replace, dominate the more traditional ways of understanding the brain. And though I'm not as familiar, I guess biological systems too.

[5:13] Arseniy Arsentyev: That's like how I think this field would change, that people who are less educated and do more manual and practical tasks would be more and more replaced by a specialist who will control kind of systems. And these systems will use AI and stuff. And most importantly, this system would be operated by people who are highly qualified and more theory-based. That's why I agree with you. And continuing our discussion, could you name some research that is happening right now that you're just fascinated about? Maybe you listen to some podcast or you read it somewhere. So something in the neuroscience right now that my audience would need to know in order to project some possible tracks that the future holds for us.

[6:08] Oscar Uzdelewicz: Okay. So there's a few things. I'll struggle to pick just one. Can I give you one example of something going on in industry and then another going on in academia? Is that good?

[6:22] Arseniy Arsentyev: It would be perfect. Yeah. Go ahead.

[6:24] Oscar Uzdelewicz: So starting with industry, there's a brain-computer interface startup called Synchron. I think the very recent startup, only within the last eight years, I think, they've really begun developing their product. And it's a relatively new type of device that's separate from what Musk is doing with Neuralink. But those details aren't super relevant. What's really cool is recently, I think a few months ago, they announced a plan to basically use a foundation model in the large language model sense or the artificial intelligence sense to model neural activity and serve as a default interpreter of all the neural activity in the brain.

[7:23] So why this is really important or why I find it very interesting is, first off, there's a lot of engineering problems with how do you actually train a model that can be widely applied to everyone's brain because each person's brain is unique. But also because it provides somewhat of a resolution to the challenge of calibration. Because when you have these brain-computer interfaces or any sort of neural interface, I imagine this might be somewhat similar with medical interfaces depending on their complexity.

[8:06] But when you have a neural interface, and you insert it into someone's brain after surgery, it's not like it immediately works. There's a delay where you have to actually train the brain to interact with this device properly: you need to make sure that it's reading the correct signals from this person's brain, or the patient's brain. So that's a major challenge, because that can take a long time. And you don't know if calibration will work 100% and you might get bad results.

[8:41] So this solution that they're coming up with building like a foundation AI model to be, I think their idea is, if it's not directly located within the device, it's located within the basically like the software package that the device uses to interact with the like the computer devices. So I think that has the potential to really like rapidly accelerate BCI development or the use of BCIs in normal, like day to day tasks.

[9:19] Arseniy Arsentyev: Yeah, BCIs right now are on the front pages of main newspapers, they're on podcasts and everywhere. The question that puzzles me personally is how. So is the science on the level that we can record and understand the neural information? Because from my understanding, these devices just take the, like, electrical, maybe chemical, pulses from the brain and transform it to some data that is algorithmically worked on. And you make some predictions, some models, stuff like this. My question is the science there to accurately read, the neural data or we need something more than just electrical signals from it?

[10:12] Oscar Uzdelewicz: Okay. So actually, yeah, that's very, another very good question. It also ties into the second example, I was going to mention, but first I'll talk about that. So you're completely right where current day BCIs, at least to my knowledge, they aren't, especially precise where I think even the Neuralink BCI only records from some number, like some thousands of neurons. There's, I don't know the exact scale.

[10:46] Arseniy Arsentyev: Do we have like billions of neurons or like trillions?

[10:50] Oscar Uzdelewicz: Yeah, something like that, like 10 to the, 10 to the 10 or if not a higher order of magnitude. So it's a very small subset of all the neurons in your brain, which to be fair, you don't need to record from that whole set of the billion neurons. Because there's a lot of, it's not like each neuron has, each neuron's activity is completely independent from all the other neurons. There's a lot of like inference you can do, but the current amount is quite small. And cannot completely encompass all the activity going on in the brain by any means.

[11:27] And I think other, like the Synchron, the company I just mentioned, they're not even using precise, like neural spikes. They don't have actual wires in each neuron. They use, I think it's basically like EEG signals to average activity from nearby neurons. So that's also another problem with precision. Sorry, sorry, there was a second part to your question. You actually asked about precision.

[12:03] Arseniy Arsentyev: And then yes. And the second part was is there another approach rather than just electrical or chemical pulses? And can we like look at the brain from some other direction, some other method that can also be incorporated into like BCI? That's my question. Or just electric and chemical. That's it.

[12:26] Oscar Uzdelewicz: I see. I see what you're saying. And I misunderstood at first. My bad. Is there another way in the, that's currently being developed? I don't see another easily engineered alternative. Though you mentioned chemical ways to read the brain. And I know that in many, in like traditional neuroscience studies, they'll, use all sorts of like chemical tracking devices or chemical tracking systems to serve as an analog for brain activity. So even though the chemical methodology isn't being used that much in the status quo, I'd say, maybe there's room there. But besides the electric and chemical methodology, I don't see much because think about what the brain is, right? It's a series of interacting, like molecules that basically are chemicals and produce chemicals.

[13:49] And then also, of course, there's all sorts of electrical pulses, which are the core of like what we think of as neural activity. So that's kind of the essence of the brain. So I'd have trouble of thinking of alternatives to that, but I don't know if you have any ideas.

[14:07] Arseniy Arsentyev: No, actually, I'm kind of a newbie in the field of neuroscience. I'm just exploring it from the regular person's perspective, reading stuff on the internet, watching some videos. But you mentioned some stuff and I believe my audience will find it really fascinating. But the question that your speech sparks is do you have any data or some knowledge whether this approaches electricity chemistry can help us to identify brain diseases? For example, Alzheimer's, or dementia, or even like schizophrenia and stuff.

[14:50] Oscar Uzdelewicz: So you're specifically asking about brain-computer interfaces, right?

[14:53] Arseniy Arsentyev: Yes. Could brain-computer interfaces help us in identifying the mental diseases?

[15:01] Oscar Uzdelewicz: So I don't think directly, because the main role of brain-computer interfaces, at least how they're currently being applied, and the goal of how to apply them, is to remedy issues in the brain and not necessarily diagnose them. For example, if someone has been paralyzed in a car accident and they can't use the lower half of their body, they can't walk. Then what a brain-computer interface would do is restore that ability, hopefully, through some repairing of the neural pathways or reconstruction of new neural pathways.

[15:50] But I do think there is the potential that if someone has a brain-computer interface installed, it would be significantly easier to diagnose Alzheimer's and other sorts of neurodegenerative diseases. So if you're constantly recording that neural activity, I don't remember the exact papers that I'm referencing here. But there's specific neural, like patterns in neural activity that are unique when your brain is experiencing certain neurodegenerative diseases.

[16:35] So if you're tracking someone's activity for, because you want them to be able to walk after this accident to extend the example from before, and you see these certain patterns of activity that are kind of indicative or, that basically tell you that this person is likely experiencing some form of neurodegenerative disease, then that would be a, basically the opportunity to investigate their brain further with more traditional diagnostic tools for those diseases. So to, I guess, to answer your question, like in one sentence, the answer is not directly, but if someone does have a BCI, for other purposes, it would be easier.

[17:24] Arseniy Arsentyev: So do you think, there would be BCIs that will do directly or only, as like, as a supplement for their main function as a supplement, you believe?

[17:37] Oscar Uzdelewicz: I think that depends how well BCIs are commercialized and how well, well, I guess this is, I guess this is under the label of commercialization. Basically, how, easy these companies, like Neuralink, Synchron, I guess in the future, there's probably going to be a million other BCI startups. How well these companies perfect the installation process and basically how, well, there's also the cost factor, of course. You don't want to be paying, I don't know, a hundred thousand dollars just to have a device in your brain that in the, I don't know, like the 1% chance you get Alzheimer's. It'll diagnose that. For most people, at least I doubt that.

[18:26] Arseniy Arsentyev: Yeah, that sounds fair.

[18:30] Oscar Uzdelewicz: And if there's additional risks, if these companies haven't perfected the beyond the installation process, but the actual, day-to-day effects of the BCI, to the point where there's additional medical complications. Like if there's the potential that, I don't know, there could be some further infection from, improper, maintenance or something like that of the BCI. Then I think those risks would dissuade people from having, BCI for more, I guess, tasks of convenience or from, less immediately pressing or immediately important tasks.

[19:19] Arseniy Arsentyev: Let me interrupt the episodes to say that this podcast is brought to you only by my will and desire. I put my soul into it and I want it to grow. You can do just three simple steps to offer me an incredible support and help me to further develop this project. First of all, please, subscribe and like this video. Second, is share it with your friend. And third, go and support me on my Patreon page. I truly admire your support, your positive comments and your suggestions. Please continue to listen to this episode with Oscar Uzdelewicz.

[20:11] All right. Also, Oscar, could you tell us about the interesting internship you were doing? Were you doing it this summer or the previous one?

[20:22] Oscar Uzdelewicz: The summer that just passed.

[20:25] Arseniy Arsentyev: The summer that just passed. Could you tell us more about it? What you were doing? And most importantly, with what type of people were interacting? More theory-based, more scientists, probably medical professionals. What was the context? And let's go, starting from your answer.

[20:44] Oscar Uzdelewicz: Sure. The lab I was working with was, it's a very interesting methodology or a very interesting kind of stuff they're doing, I guess, to speak it generally. But the lab was, it was at this institute called the Flatiron Institute, which is doing very, incredibly interesting work in computational neuroscience as well as kind of applying computational methods to other fields. I highly recommend any listeners of the podcast, check them out. But the lab that I was working with, which exists within the Flatiron Institute:

[21:26] Their main contribution, at least recently, to computational neuroscience has been the further investigation of this idea of neural manifolds. So, this is a relatively recent concept that basically posits that a core way to understand the brain is to look at the geometry of the neural activity. So, what that means is when you're recording, this, brain activity from a certain task and, I don't know, in most experimental settings, it would be either a mouse or a monkey, you record the quote-unquote firing rate levels of each neuron, which is basically a measure of the neural activity of the neuron.

[22:26] If you plot it out in coordinate space, let's say you're recording from a thousand neurons, you'd be plotting in dimension a thousand, so R to the thousandth. Those recordings will have distinct geometric properties, depending on what task the monkey or the mouse is doing. Or if it's exposed to different stimuli, too. So the classic example they bring up in the lab is also, so I'm talking a lot about the lab's work right now, but I'll get to the other people in the lab in a second, because that's also another really important point. But the classic example, and then I'll wrap up, is if a monkey is given a vision task, so let's say it has to basically, classify different images it sees that pop up on a computer screen.

[23:25] If it's seeing an image of a dog, this image will basically cause a certain pattern in the neural recording where, when you're actually plotting out the firing rates, the firing rate trace will exist only on this manifold in this 1,000-dimensional space. So, if you can think of it as a sphere, this firing rate trace will exist only on, and it will only move on, the sphere. Whereas if you expose the monkey to an image of a cat, this firing rate trace that you get from the cat example will also exist on an object, let's say it's also a sphere, but this sphere will be completely separate from the dog sphere over here.

[24:25] So, you basically, you can separate the two activities, based on their geometry and kind of use it as a, as a diagnostic tool. But there's also a ton of other, information theory and physics based. If you're familiar with statistical mechanics or the mean field theory, they use a lot of those techniques, and that theory to kind of get more interesting things out of this geometric idea. But that aside, the people working there were mostly from physics backgrounds, which I guess stresses what we were talking about earlier with the importance of physics for, understanding neuroscience and biology. There were some engineers, some computer scientists, and a few mathematicians. But very few people who had been educated purely in neuroscience or purely in biology.

[25:36] So, I guess that shows, that reinforces what we were talking about earlier, where you really want to learn computational, fields or be like an expert in a computational field if, and then also learn some biology or neuroscience on the side before you really engage in computational neuroscience or computational biology. But I think it also highlights how there's a lot of like, interdisciplinary activity going on. Because you had people from physics who had a certain point of view or a certain methodology for how to analyze problems. But then there were also a decent amount of people who had studied electrical engineering who might have a slightly different way to analyze whatever problem they're looking at, or someone from computer science would have a completely different way.

[26:28] So there's not, as long as you're doing something computational, I'd say, you have a pretty good chance to contribute. And you, I don't think you really need to worry exactly what path you take in that sense of like, what exactly you're majoring in, as long as you have that computational experience. And you kind of have that problem solving experience as well.

[26:56] Arseniy Arsentyev: Oscar, could you try to kind of restate, maybe simplify a little bit, this concept of geometric, stuff that was going on with neural signals, to give more people access to understanding it? Just try to rephrase it and put it into like 10 sentences.

[27:18] Oscar Uzdelewicz: Okay. It's hard to also just explain purely with words. So first you need to understand that when you're measuring neural activity, that neural activity can exist in some geometric space. Think of a three dimensional space when, I don't know, you're playing some computer game or something like that. Where each dimension of that space is X, Y, Z, in this neural space each dimension is the activity of one neuron. And in this geometric space, when an observer, in the case I mentioned it's a monkey, is exposed to some stimulus, the activity will exist at some point in this dimensional space.

[28:37] And what the work of the lab that I worked with this past summer is claiming is that the points that these stimuli form build specific structures in this space. And these specific structures can be analyzed for their properties. And you can basically get some kind of intuition of the system based on these properties. Does that make sense?

[29:23] Arseniy Arsentyev: Yeah, it does. That sounds great. And I know, I feel that people would understand. And for me, it was a new concept. So saying, thank you very much for educating me and probably educating my audience. Let's zoom out a little bit from the BCI to the more systematic. And when I say systematic, I mean like the political, economical and structural healthcare that exists right now. And my question, is the question of this whole podcast is how do you envision medicine in 10, 20, 30 years? What will change? And of course, we're talking with you about, neuroscience related. For example, if a person gets a brain damage, he goes to like to the emergency and he does his MRI scans. Then he sees the treating doctor, the doctor diagnoses him and he gets some emergency treatments.

[30:28] Do you feel that this kind of a chain would change? And if it would, in what ways?

[30:35] Oscar Uzdelewicz: Okay. So that's a pretty big question. And of course, I'm curious what you have to say on this topic more generally too, because I'm just approaching it from the neuroscience side. But I think, I don't know if in 10 years, but in 20 years, most accidental paralysis injuries, so if someone's in a car accident, like I mentioned, or if someone suffers some kind of accident and that leads to them being paralyzed, I think the effects of that will be completely gone. Like there will be a cure for all of that kind of paralysis.

[31:20] Arseniy Arsentyev: So kind of neuro regeneration. Is it true? So to regenerate the damage that was done during that.

[31:30] Oscar Uzdelewicz: I think we need to be careful when we use the word regenerate.

[31:36] Arseniy Arsentyev: Cause neurons do not regenerate for the audience.

[31:38] Oscar Uzdelewicz: Exactly.

[31:39] Arseniy Arsentyev: They just die.

[31:41] Oscar Uzdelewicz: But new pathways can be formed. Especially like with the help of BCIs, you can of course, like manually wire, new pathways throughout the body, which we don't have the current technology to do that well currently. But I think within 20 years, we will be able to, especially considering the current rate of advancement and the use of other technologies, especially AI, we will be able to solve pretty much all of those problems. I think there's still difficulty with diseases, like ALS or Alzheimer's where it's a chronic problem in the brain that neurons are degenerating. And the system as a whole is kind of crumbling apart where I don't think one device will be able to fix that. At least not a device in the form of a traditional BCI that we're looking at now.

[32:48] But I do think that especially with advancements in more of the theoretical side of computational neuroscience, we will be able to diagnose these diseases a lot better because we'll have a far greater understanding of the fundamental components and the fundamental rules, quote unquote, of the brain. But I think treatment for those diseases might still not be there within 20 years because they, unless there's some sort of like chemical or medical accident where we find some magical chemical that'll resolve those diseases for whatever reason. But because those chemical pathways in the brain are so complex, and genetics also plays a huge role, I imagine, in those diseases as well, there won't be an easy solution.

[33:59] But I do think all, especially, and not only paralysis too, but problems like vision, like kind of blindness could be repaired. The potential for a lot of like birth defects could be repaired with BCI technology. But the broader systemic issues we might have trouble with.

[34:23] Arseniy Arsentyev: Yes. And just making a step back, and looking just at the imaging part, I firmly believe, and I want to believe that we, as a humanity will create such a tool, like sci-fi a little bit, maybe a little weird. When you just like lay down and it shows all of your diseases, it's kind of a dream, but I hope something like this will happen. And right now we have brain imaging, that is not very cheap, not very efficient. And it uses X-rays or in just advanced imaging, sometimes even like radiation rays. If you have cancer or something, do you feel that the science is going into the direction of creating some alternative to it? Or is the MRI like the standard that will go through centuries?

[35:27] Oscar Uzdelewicz: To be honest, I'm not completely familiar with the advancements in MRIs. But if I, from everything I've heard in the papers I've read, it doesn't seem like many alternatives are being developed, but the MRI technology itself is becoming a lot more precise.

[35:50] Arseniy Arsentyev: Yes. That's why I'm asking you, I know it's a tough question, but me personally doing the research, I do not see any branch that is offering the alternative to the MRI. MRIs are getting more precise, but they're not getting like more compact. There are still this huge machines. That like take rooms and you need to prepare, you need to cover your body to not let X-rays to infiltrate your skin and so on. So we can transition to the next question. And this question is a little bit speculative and you need to use your whole imagination. So the question is, if you had unlimited funding, what neuroscience based healthcare project would you start today? Like if I give you a billion dollars, what will you do to maximize this Delta where Delta is the common good?

[37:02] Oscar Uzdelewicz: So it doesn't have to be necessarily healthcare related. It could just be overall societal benefit.

[37:06] Arseniy Arsentyev: It's better if we speak about healthcare. Let me take the stage for a couple of seconds to clarify. As you know, there are at least in some research and in my opinion, also, there are three main characteristics of a good healthcare system. First of all is the quality of life because some people live up to a hundred, but they're not living a disease free and like a healthy, painless lives. And the second one is longevity, when people have to live longer in order for a healthcare system to be successful. And the third one is like fertility and the percentage of children who are born with particular malfunctions should decline. So that's what I mean by good healthcare. And what would you do?

[38:03] What research project would you undertake to try to maximize the efficiency and the goodness of the healthcare in the terms that described? Would it be AI? Would you transfer our recognition to the computer just to get rid of old human beings?

[38:26] Oscar Uzdelewicz: No, that would be, that's too easy of a solution.

[38:30] Arseniy Arsentyev: It's easy for you.

[38:31] Oscar Uzdelewicz: Exactly. I think, in terms of what neuroscience can address the most, it would be the last thing you mentioned, in terms of birth defects, or basically people who are born with certain inabilities to do things that, I guess, we would normally be able to do. I think the potential for maybe more precise neurosurgery as well as I don't want to say BCIs, but kind of like neurotechnology generally to address that is kind of an untapped field.

[39:35] So if I could take my billion or trillion dollars, whatever amount you said, I guess I would split that between developing, or researching, specific methods for how do we maybe remove or, at a large scale, rewire certain parts of the brain at birth, to basically give these infants, these recently born people, better capabilities for the rest of their life. So that would be kind of like post-birth surgery or post-birth intervention. And then I would also invest in the development of neurotechnology that would kind of serve as, I don't know exactly what the right way to phrase this is, but almost like supplemental computing in the brain or adjacent to the brain, that would almost, sidestep these diseases or these conditions in the first place.

[41:07] So, basically what I would envision that looking like is you would almost have sort of a mini brain or a collection of some decently large amount of neurons that would exist in some device implanted most likely in your skull or next to your spinal cord. That would connect to your brain and basically serve as additional circuits or maybe circuits that could replace the faulty circuits responsible for certain conditions that the person might have. And therefore basically bypass the problem in the first place.

[41:54] So instead of directly intervening in someone's brain, by, I don't know, surgically modifying it, if you could wire up or create a device that could connect to the brain, but minimally, with minimal surgical interference in the actual workings of the person's brain, it would serve as basically like an alternative pathway to do whatever task the person struggles with. If that makes sense.

[42:29] Arseniy Arsentyev: Yeah, it does. If I had a million dollars to invest in some healthcare project, actually a project related to neuroscience, I would probably go with projects exploring the simulation of the brain, because in my opinion, if we'll have some kind of computational model of the brain, a model with which we can interact, it would accelerate the research, just as AI accelerated protein folding. Do you know about AlphaFold?

[43:00] Oscar Uzdelewicz: Yeah.

[43:01] Arseniy Arsentyev: Like protein folding. So it would be, it would have, the same effect on the neuroscience because right now, the percentage of people that have like their brain studied is not that big. Also people who have certain disorders, for example, rare, as you know, rare, brain disorders, like only, I believe hundreds or a thousand people around the globe are under this disease. And it's hard to explore them. If we got a simulation of the brain, it would mean a lot for medicine. It would like, it will drop the costs, accelerate the research rate.

[43:43] All right. Next thing that I want to ask you is do you have any question, any question for the next guest on my show?

[43:59] Oscar Uzdelewicz: Interesting. So it's a general question, right? Like I'm not assuming necessarily anything about.

[44:06] Arseniy Arsentyev: Yes. And you can say something, related to neuroscience or go broadly into medicine.

[44:18] Oscar Uzdelewicz: It would probably be what is, and I think this is useful for anyone who's trying to get into the computational side of medicine. What is the area of medicine in terms of like the direct biological applications and less like the bureaucratic infrastructure, like the economics. But what area of that side of medicine is least computationalized or like has the least amount of computational techniques applied to it? And how do you see, these kinds of computational approaches benefiting or tying into that area of medicine?

[45:10] Arseniy Arsentyev: Yeah. My future guests would definitely be puzzled by a question. Thank you a lot, Oscar. It was really delightful to talk to you. I know that we'll do this podcast again in some time, maybe in a year. We'll see how it all plays out. You're always welcome on the show. I wish you good luck in your studies and hope we'll hear your name more.

[45:34] Oscar Uzdelewicz: Thanks. Thank you very much, Arseniy.

[45:38] Arseniy Arsentyev: This podcast is brought to you only by my desire and will. Support me on Patreon, and the best thing you can do is to subscribe. Thanks to you all guys. See you on the next episode.