AI, Biomarkers and the Future of Longevity Medicine, with Elio Verhoef, Co-Founder of LongevAI

[MUSIC] Welcome to Beyond Long Jeviti, the podcast that explores not just how we age, but how we can build a longer, healthier future for ourselves. [MUSIC] My guest today on Beyond Long Jeviti is Elio Firth, co-founder of Long Jev AI, a company building AI tools for longevity clinics to help clinicians analyze biomarker data, streamline documentation, and create personalized client plans more efficiently. Elio comes from a computer science background, and from a young age, he was deeply focused on health, energy performance, and the question of how we can use better tools and better thinking to improve long-term well-being. That eventually led him into the longevity field, where he began exploring how AI could be applied in a way that is actually useful in clinical practice. In this conversation, we discuss what AI in longevity medicine can and cannot do today, where it can help clinicians, where the risks still are, and why human oversight remains essential. We also discuss where Elio believes the future of AI and longevity may be heading over the next few years. [MUSIC] Hi, Elio. Welcome to Beyond Long Jeviti. It's really nice to have you on. You are co-founder of Long Jev AI, an app, a technology that helps clinicians in their daily life. Why don't you introduce yourself a little bit to our listeners and tell us a bit more about why you founded Long Jev AI and what it does? Sure. Thank you for having me. I am a longevity enthusiast, and I'm also very much into AI learning as much about as possible, and applying it to my daily life and in the business as well. I'm Adio Verhoof from Midlands. I'm Hof Austrian, and I've studied the double bachelor of computer science and information science, and ever since the larger AI models have been rolled out, I started to get into that, like GPT-3, etc. At the same time for the entirety of my life, I've basically been focusing on optimizing almost everything I can see, optimizing the computer, with also optimizing myself as a person in terms of personal development, in terms of optimizing my energy, health, especially end-of-all-being. From the start, I've avoided sugars and all these kinds of things, and later on I got into longevity, even though I didn't really know the term yet, like finding ways to optimize how I behave, how I feel, my health in general, and because I was into this, and at the same time, because I didn't see a lot of other people similar to me or around me who had the same passion, I started looking online for communities that I could join that were similar to this. Then I found a community in Madeline that organizes monthly meetups in longevity, and that is where I met Cosmina. Every month I've been joining these meetups, and then December 2024, after getting to know Cosmina for quite a while already, I found it's longevity AI together with her. And that's because we knew each other, we saw how well we could work together, she had extensive consultancy experience, and I had the AI experience that we were both dedicated to longevity. So then in December she asked me, "Well, do you want to start a company together focused on longevity by leveraging AI?" and I obviously said yes. Fantastic, sounds really insightful your lifestyle, and at such a young age, so you're ahead of the curve, I think. Do tell us a bit more about Long Jeff AI, what exactly is it, what does it do, who is it for? Long Jeff AI started initially as a longevity AI company to help longevity companies by leveraging AI. So it was a consultancy company to apply and build custom software for longevity clinics, where that custom software leverages AI as effectively as possible to help them automate operations, be more effective in what they do, and we also started working with researchers this way, and this is still partly our business model, but now we're shifting more into software as a surface because we've seen that Long Jeff AI clinics specifically had quite a few of the similar issues across all of them, and we can build a platform that can cover most of them in a very configurable, customizable way, and that's what we did. So now this is a more scalable way to help way more longevity clinics help their people improve their health, and that's what we started doing while still working to better with researchers and other companies on the site. You mentioned that you have custom software to automate the work, what exactly does that mean? In this specific case, it's a software for the clinicians of these longevity clinics, and they do have quite a few manual operations with regards to the data of their clients, and will be interpreting the blood work, creating client reports, creating notes during conversations with clients, and quite a few other things like creating an action plan for the clients, all this documentation, note-taking, and creating of shareable documents for the client in a branded clinic style can take many hours, actually, for every single client if he don't automate it in some way, especially if you want to go a bit more comprehensively. So if you have these conversations, which can sometimes take three or more hours, and then you have to have somebody in that conversation to take notes during the call, and then process it afterwards. That just is a lot of overhead for these clinicians who would rather spend time on their clients. So we can think about how to utilize AI in a smart way in each of these steps without taking over the job of the clinician. So the clinician remaining in control while AI guiding them on the steps where automation can be present, and that is mostly in creating initial versions of the documents that the clinician would otherwise make himself. And also, of course, to just record the conversation like right now. We're being recorded, we can create a transcript of that, and we can use the transcripts to convert it into a meeting note summary or a client summary for the longevity clinicians, that we can then share with them. And the same goes for bio marker interpretation, instead of having to read the lab report, you can let AI read it, extract the biomarkers, link them to the reference ranges that your clinic uses, all automatically, and then review those instead of having to do that from the lab report itself. There's many different steps. Also, the action plan can be initially generated by using the methodology of the clinic. So first, if you just once captured the methodology, your clinic uses for generating actions for any kind of client in any kind of situation, after that AI can do that to some degree. It can create a initial version of an action plan for you to review and edit. And then once it's approved by you as a doctor, the client will also be able to see that. Sounds amazing. And it sounds like a real help for the clinicians. So just to briefly sum it up, what you've said, and correct me if I misunderstood. But on one hand, it's saving the clinician's time with the bureaucracy of things, getting the actual notes down facts, but it also helps the clinician interpret the biomarker data. For interpreting biomarkers, why would a clinic use an AI tool? Is it because of speed or because of accuracy or both? Mostly for speed, but also for accuracy in some regard, because in the end, with these kind of things, always the clinician should remain on control and be able to see what is being shown to the client in the end. And perhaps also why that would be shown to the client. So when interpreting biomarkers, usually the clinicians have some kind of intuition and knowledge and framework on doing so. However, everybody has a limited memory and everybody has a limited capacity. AI models are trained on trillions of documents across the internet to have a broad knowledge on all kinds of topics. And if you can prompt an AI model well, then you'll be able to extract to some degree, what other clinicians might have extracted from those biomarkers. And if you also connected to your own methodology, what the AI model can do is it can generate for each of the help domains that you might adopt as a clinic a description of what is the case for this client, what has been found for the clients. And that description, because it uses both the general knowledge of the AI model and your clinician's knowledge, is usually initially to some degree not what you would have done yourself as a clinician. And in some cases, it can be better or more comprehensive or accurate than what you would have come up with a clinician. And then by combining that initial great input from the AI and your own knowledge and expertise as a clinician, you come to create more appropriate description of what is going on for any client. And this way, you also improve as a clinician, because you'll see what the AI comes up with with regards to some help domain. And then you can verify that against literature, against your own knowledge to check if there is anything that you don't agree with. And if there's something that's new to you, you can learn from that. And the other way around as well, after such a description is generated by the AI model as a clinician, you can give feedback to the AI model and the AI model take that feedback into account in all future generations of the description of such a help domain to improve its own output over time as well. It says, "Synbiosis is a goal working of the human and AI to speed up the process and improve accuracy." How accurate is the AI at the moment? That's very difficult to say. I don't know. I know that AI still makes mistakes. The AI models I've seen right now are incredible and I would trust them personally more than I would trust a longevity condition. However, if it doesn't have the access to latest research and I would, if I wouldn't have the knowledge I have with regards to longevity itself, then I would be more skeptical because sometimes it does come up with overutilized conclusions or opinions in the longevity space that are incorrect. And this is because it has been the main way of reasoning, for example with Alzheimer's and tauplex, etc. There's many different partly correct interpretations of these subjects. The AI model is trained on all of this old data and therefore it can conclude old incorrect things. So I'm not sure how accurate it is, but I do know that for every answer I get out of the AI model, I will of course validate it against my own knowledge, but if it's something that I don't know a lot about, I will definitely also use AI to perform deep research on the web with regards to the latest research on that topic to get a better understanding of that topic so that I can improve my decision making for myself. So if you use it correctly, I see it can be a great benefit to the clinicians and to the patients, you know, and everyone involved, but aren't you afraid that it can sort of lead to the clinicians not double checking what AI tells them because you previously said the clinician should be in control. So how can I as a patient be sure that the clinician I'm seeing knowing he's using AI be made comfortable and be reassured that the clinician is not just using straight up the AI interpretation of my results, but is actually applying his own knowledge to it. When you are going to any kind of condition, there is a chance that even though they don't tell you, they are using AI to interpret your data and the chance is increasing because AI is becoming more prevalent and more used. So it's always the responsibility of the clinician to do this and there's not really anything we can do too much about that to always 100% enforce it. However, like in the system we built, there's always the need for the clinician to first approve before any information is shown to the clients. So it always needs to be approved. It's always the responsibility of the clinician. How does long JVI deal with privacy and security issues? Because those are quite prevalent and relevant in today's world. Well, we prevent them. We try to prevent them as best as possible, making sure that only the clinician and the administrator of the clinic have access to the data of the clients and also the solution is GDPR-proof. So all data is hosted in Europe, but 100% guarantee it depends on the clinician not sharing any data. And for the AI models, we are ensuring that all the data sent to the AI models is first anonymized to not share any of the clinics information within the AI model, even though the AI models are already saying that they don't train on that data. So the companies behind the AI models say like, hey, we don't train on the data you sent, the models are hosted in Europe, but just to be sure we also anonymized the data before sending it. So for the non-tech people in sort of layman's terms, how does the AI learn if it doesn't use the data from the patients that it's being fed? They I learn from the feedback from the doctor. So if there's a description of a certain health domain like for example, Cardio Vascular Health, they I will have initially generated a description of what my current Cardio Vascular Health is based on my LDL, HDL, APOB, LP, little A, all these biomarkers, split biomarkers. And well, the clinician can come back to that description of the health domain and say, well, actually, this is a bit too technical. So could you make the description a bit less technical? And then for all future generations, the AI model will be able to adjust its description accordingly to be less technical. However, it can be content related as well. Like for example, if the doctor shares, well, this is a good analysis, but you didn't take the LP little A values into account correctly. You have to do this isn't this. Then we can save that feedback together with the description, which doesn't contain the name of the client to in the future ensure that the AI model will not miss out on that interpretation of LP little A again. What do you see as the limits of AI? As in data gap bias or lack of longitudinal data, which is data that's collected from the same group of people repeatedly over a period of time, rather than snapshot data. You know, me going to the doctor, one every three years having a blood test done once and that's it. One major limitation is the amount of data that's a AI model can process at any given point. It's expanding, but it is still not that large and to be able to give an accurate analysis of any given person, you do need either a very large amount of data or a selection of the most critical data and that sub selection can be relatively difficult to get. You need expertise in the longevity fields for that and you need to spend some time to create that and a sub selection could be some general information like age, weight, length, gender and the core biomarkers over the past five years as well as how I'm feeling over the past five years and a general description of my life like the most important things that can still be provided in a relatively short amount of text. If you're able to provide that well, then data model can perform really well. However, with lots of this longevity related data, it can become a lot quite quickly and some of it can be somewhat relevant to a certain degree. So the difficulty here is before applying any analysis before sending the data to an AI model to first select the most relevant data and exclude all the potentially relevant data because it's the more data you send to an AI model, the worst its output. This is because just like a human, it's kind of has to pay attention to all kinds of different pieces of information, different words and the more that becomes the lower quality the outputs become and at some point it cannot even handle more than a certain amount of data and it will just not be able to produce any outputs. I have a return question. What exactly do you mean by the data cap? What I meant was the AI in order to produce its model, it needs to have comparative values or consecutive values or continuous values. If you just feed it one off data in the abstract, is that a limitation of AI that it doesn't have the continuity? Well, that's not necessarily a limitation of AI itself. It's more about the software built around the platforms because this is about providing context rights, providing context about you as a person, what you have been through so far in your life. Yes. It's up to you to provide that information to the AI model or we need to have a software that keeps track of this. So this is also where chat GPT or clause memory comes in, where across conversations it will look for relevant pieces of information and store those to bring them up in the future again to have some context about you. How do you prevent AI making up its own mind as to what goes into the gap? So, you know, we all know when we use chat GPT and sometimes the AI doesn't really know what to answer, it just creates a fake answer that sounds very real, but is absolutely not. How do you prevent that from happening? Or can you prevent that from happening? Up until now, there is not as far as I'm aware a 100% foolproof way to prevent this from happening. There might be some models out there that are not necessarily as smart, but very secure in this area where they will only respond to information that is indeed correct or indeed provided. However, how it basically works is, LLMs are prediction machines. They predict the next word in a given text. So if an LLMs trained on always seeing the words Singapore after beautiful country, then it will produce that word in almost all cases unless there is a significant difference in other input tokens. So even if you explicitly tell it not to generate a word Singapore in that sentence, it will still do it because that's what it has been rewarded to do in a training procedure. Models have become a lot more a lot better at this prevention of hallucination over time. There's also things you can do to prevent it yourself. For example, explicitly instructing it to not provide any information that's not explicitly present in the output or deductible from the input and to reduce the creativity of the model. There's a parameter that you can set 40 models when you use it in code. And it is called the temperature of the model. If the temperature is high, it will be more creative. It will allow for more variations of words as the prediction. If it's lower, it will be more consistent and robust. So lowering the aperture will prevent that more from happening if the instruction is also to note includes those random. Do always use the input. But surely clinicians are not tech people, in most cases, and they surely don't have the time to train the system. So what security do you build in for that not to happen? Because that's quite a scary thought as a patient. And I guess also as a responsible clinician that you just have AI creating facts rather than relying on real facts. So the system is already including all the instructions under the hood to prevent that from happening as best as possible by indeed instructing the AI model to only use the data that is provided about the specific client, as well as the expertise of the clinic that is provided explicitly to only use that for its output and nothing else. That will help a lot, as well as reducing the creativity of the model will also help a lot to provide better outputs. And of course, yeah, just always using the latest models that are the best in overall model of the longevity fields, as well as the lowest rates of hallucination. But in the end, there's always the need for this clinician to review the output before editing or updating it. Do you train the clinicians that use your model? Yes, we'll just give instructions on how to use the software. We don't have that yet, but we will also generate a full documentation system on how to use it with short videos so that it's easy to use. But we've built it in such a way that you have to make it as intuitive as possible and as similar as possible to other tools out there, other platforms, ebg's to make it easy to use for the clinicians. Why do you think buildings, AI system, custom solution is very important for clinicians, especially clinicians, sort of, in the longevity field? Well, because it doesn't exist yet. To build something that takes into a county entire client journey over a long period of time, with all their data from perhaps variables as well, but also from their biomarkers, perhaps from their DNA tests, and then automatically every step along the way, as best as possible, while taking into account the feedback from the clinician, it's worth to have that in one platform. And that one platform did not exist yet, so we decided to build it. So that's why this custom platform is important to have. You've mentioned variables. How important do you think is the integration of existing systems and how important is compliance overall? How important is compliance? I think it's perhaps the most important because if you're not compliance, you will get fines. (laughter) So as a clinic, you need to be compliant and wearable integration. I'm not sure. I think you can get very far as a clinic with just blood biomarkers. Perhaps also, of course, consults to figure out what's going on and to track those over time. But wearables can add an additional layer of information, especially with regards to how well some patients sleep or clients sleep, because that's not always easy to guess for themselves and to share that way with the clinic. For your personal health, it's worth to have a wearable like a whoop to see how well you sleep, how you respond to certain stimuli or foods or supplements to improve your health yourself as well. So let me ask you a personal question now. I'm sure you've put your own data into Long Jav OS and if you don't mind, do tell us about the results and there were any surprises. So I have indeed put my blood biomarkers into the system. It was initially, mostly, to test the entire technical side of things to see are the biomarkers extracted correctly, which needs to work for any kind of lapper boards, right? And indeed, also what's description is just a generator for me and what advice does it give for me specifically. The tricky part here is that my blood work is already quite good. So there was not a lot to improve upon, honestly. So I didn't get a lot of interesting feedback from the system because, well, my biomarkers are already quite good and they already went over it manually with AI before. But I can go into how I go about that. What I do is I first, of course, get my PDF lapper boards and this is something important, I think, to come back to the context aspect. AI models work best if you give them the relevant information and the relevant information only. If you give more information than necessary, it will perform worse in general. So what I did is not directly upload my PDF lapper boards to the AI model. For analysis, initially, I uploaded the lapper boards to an AI model to just extract the biomarkers from it. Of course, I check if that's correct. I extracted the biomarkers and the reference ranges for me specifically and, of course, the units. And then I used that text, excluding the logo of the company, excluding all the images and text about when the lapper board took place and what the name of the lab is and all the extra details, like the reference number. It was overmoved. And instead, I could just use this text as input for the AI model for my analysis. Then I will have text about myself ready to give some context. And this is about who I am, what I do, what kind of exercise I do, what kind of diet I follow, what kind of supplements I take, perhaps medical history as well. And I use that combined with this list of biomarkers to say, well, this is important as well, to provide a role for the AI model. Please act as my longevity doctor. Here is my lapper board and here is some information about myself. So this would be, I'm a 23 year old meal. I run a lot, I go to gym, I try to sleep as best as possible and generally get good sleep. My resting heart rate is this, my HIV is this, my VU2 max is this, just as context. And I'm generally feeling good. Here are my biomarkers, including reference ranges from the lab used. What do you see? What do you notice here? So what catches your attention? And what advice would you give me as my longevity doctor? And then also, why would you give me this advice? Then I will get an output, I'll look into it, I'll see if all of this makes sense to me. And for the parts that I don't fully get, or that I perhaps even doubt, I will ask questions back today, I'm allowed to clarify, or for specific things that are more in depth, I will go into a separate AI chat and go specifically into that direction. For example, it might give me some feedback that's, well, I need to do something about my liver biomarkers, but I know that my liver is, and then I can look into a separate AI chat into the intervention that it suggests for that, by using a deep research functionality. And in that case, the AI model can perform deep research into the data's literature by searching the web, searching PubMed using a connection to PubMed to directly look into papers and provide me answers based on that. And this is also important, I guess, for people at home to look into these integrations with sources like PubMed, for example, in clots.ai, you can just go to the connection step and search for PubMed and turn it on. And then clots can, on its own, look for these sources in PubMed with regards to your questions. That's very interesting. And then, does it tell you what you should do or shouldn't be doing, or does it merely interpret your genome or biomarkers? If you ask for advice, it will provide advice. If you ask for merely interpretation and things to look out for, it will provide that. My example I asked for both, so I will also get some concrete advice. And it will be relatively high level initially, but if I provide some more context or some things that I would like to prioritize in terms of type of interventions, then it can go into that as well. Going back to investors that are not in the AI field or the medical fields as such, they just want to invest in the longevity markets. What do non-tech decision makers most often struggle to grasp when it comes to AI in healthcare? That's a difficult question. I wouldn't necessarily know too much what they would struggle to grasp. But I do think there might be some outdated views or perhaps incorrect views from the past as well, because some examples I've seen are that people who are not up to date with AI can have one bad experience with AI models in general and then overgeneralize that to mean that also AI models of today behave in the same way. For example, this can be hallucination, right? They've seen perhaps links to sources being generated that do not exist and this can be very troublesome. Nowadays with the latest models that always never happens anymore. And for the rest, I think that many people underestimate what the AI is capable of, especially models like cloths, opus 4.6 or AI like Gemini 3. These models are just very capable and they do indeed still have biases. They do indeed still, they are indeed still capable of hallucination. However, if you prompt it in such a way that you say Here's my data and for everything that you're not sure of, please search for extra confirmation either in PubMed or the web and based on that give me like the interventions or suggestions to improve my health. For every intervention or interpretation day I model makes of your health data to ask, please explain your reasoning behind why you interpret it in this way and why you suggest this intervention for me specifically. And of course if it doesn't do this adequately initially you can always ask a follow-up question to do it anyways. And this is also what we do in Long Jeff OS. So in the platform for all the suggested interventions or actions the doctor can see with by offering over an eye circle what's the reasoning was for the AI model to select that intervention for the specific client. And this is very important to have because then you can understand and yeah what is going on why it was selected and how to improve your library of actions as a clinic to adjust your selection for specific clients in future. You mentioned to AI models that already exist you know Genesis now but can you just explain a little bit more to our non AI specialists who are listening to the podcast what exactly they do. What they do. Let's start with Claude because I like Claude a lot myself. It is a large language model which is able to predict text in a sense that if I say hello I am the next word will probably be a because usually in loud words come next and the AI model can do that and it can do that continuously over a large sequence of text. It is also trained on this role play of its being the AI and you being the human. So when it receives a message from the human it will predict the next token as the AI model trying to generate a response. If we would not add this layer of training then it would just continue with whatever you were saying as if it were you. So that's basically what it tries to do. It tries to predict the next words in a response to you and the next word can be any word in English or any other language usually. That's just the basis of it and it will use its internal knowledge of being trained on basically the entire internet. As effectively as possible to create a good answer for you that has been rated as good during the training phase by other humans and sometimes also as an other computer. And what it can also do these days is tool calling and tool calling basically means that it can use external resources to inform itself and create a better output and to perform actions in the real world like sending an email, booking a call or creating a slide deck or running a piece of code. It can do all of these but it can also do a tool call to just search the web or get information from PubMed by searching for any term like diabetes. So in that way AI models now can first use its tool calling mechanism to get information from the web or PubMed or other sources and then perhaps execute some actions and then let you know, hey this is what I searched for, this is what I did. Here's my answer to your question. Gemini does a similar thing? Yes it's very similar in most aspects however the connections to external sources like perhaps applications like Monday or Asana or others is still lagging behind clouds. Claude is capable of integrating with almost any system relatively easily from their website directly. Let me you know go back to sort of the field of longevity and AI. Where do you currently see its limitations and in two years in five years, where do you see AI and longevity's heading? Good questions. I think we went into a lot of limitations already like the bias from the training data, the occasional hallucination, the not having the full context of US person at all times if you don't explicitly provide it and also the limitation of not being able to process huge amounts of data at once since it will only be able to process a search in amount of tokens or parts of words at one moment of time. So yes this is currently limitation. It has been limitation for quite a while and there are techniques to go about this effectively. What I see AI being for the coming years, it's difficult to predict but it has been getting better and better and there are very many different ways in which you can creatively solve many of these problems I just talked about and it's just a matter of time for those to be implemented. So I do see the possibility of AI becoming way way smarter and kind of super intelligence being achieved because in some domains this is already very close to that. I would trust AI more in many aspects than many of the people I know and of course there are still some things where people yeah have the upper hand and that would still be in my opinion emotion related cases or judgments or interpretation of more complex cases that have a lot of nuance. However for the technical cases AI is already exceptional. You're pretty young so with extended help or without you still have quite some years ahead of you. Where do you personally see the longevity field going and where do you see the AI connection taking it? I haven't yet the personal experience to be able to come up with my own grounded conclusion but I do see other researchers who have their interpretations and what I see also looking at it logically is that gene therapies might be a very good opportunity for AI models to look for new gene therapies and look for the implications of those in the human body and I also see the possibility of modeling a full human cell including all its organelles and other aspects over time to then be able to model organs and the human body more effectively to then look for what kind of interventions if combined and structurally applied over time to such a model of a human would lead to which outcomes. This is still incredibly complex because you would also kind of need to simulate the brain and human decision making but if possible it would at least even if imperfect lead to a lot of breakthroughs in the longevity field a lot of molecular interventions being generated and also perhaps gene therapies being generated that can bring us closer to living forever which is one of my goals I want and would like to live forever and I think that AI could help us in that I'm not sure if I'll reach it in my lifetime but I'll try my best to do so and yeah this phenomenon of technology becoming more capable and finding ways to lower biological age more quickly than humans naturally as far as in their biological age over time is called longevity escape velocity and I think we can achieve it. Well look I don't think I'll see it in my lifetime but you're young enough to stand a very good chance that you will but it's very interesting to have someone as young as you take such an interest in the developing field of longevity and what can be done and I find it very encouraging because I've been speaking to a lot of people in their 50s, 60s, 70s about longevity and they all feel that the hardest thing is to motivate the younger generation to look after themselves and take an interest in longevity but it certainly seems you have your private life as well as in your business life do you see that around you too from people in your age group that it's a concern or an issue of interest? I do and I feel like there are many factors at play here. I also see quite a few young people who do turn to the longevity movement and to our interest in improving their health but in terms of human nature many people start addressing your problem once they see it and if you're young you don't see the problems yet so I can understand that point of view and what I also notice is that it's perhaps to some degree a lack of self-awareness perhaps where people are not necessarily aware or perhaps even they don't have the information about what the implications are down the road if they continue in the same way as they have been living which is perhaps not as healthy as they could be living. There's not yet that many examples around of people consciously focusing on their health and spending a lot of effort on that and showing what the benefits are not in the future but also in the moment and I feel like this is coming up a bit more and if this is coming up a lot more then surely people even younger ages will start to replicate that. Another factor is probably all the different things these days that are attracting attention of humans in terms of social media mostly I see or entertainment services can grab a lot of attention from people and if you grab the attention of people well they don't have to attention anymore or the time and energy to look into longevity or spend time looking for ways to improve themselves. Thank you so much Elio with super enlightening to have you on the podcast today I usually ask five rapid fire questions to Thank you. guess in the end. So if you're okay with it, here we go. The first question might not be so applicable to you, but let's try it. What's the single best advice you would give your younger self? It would probably be to be aware that whatever I say or do, it doesn't actually matter as much as I think, and that I shouldn't care at all about what people think of me and just have them. I have to say it's amazing that at 23 you've already come up with that. Normally, takes another 23 years at least for people to realize that. So good for you. The next question. Name one habit everyone should adopt for a longer, healthier life. Meditation. It allows you to be more self-aware, to think better, think clear, and to make better decisions for your future self. And that way you get to spend more time looking into the right practices for own Jeff D, learning and improving your health across many, many years and to keep improving it over those years. If you weren't in longevity science, longevity tech, what career would you have chosen? Hey, I like, hey, I development going into achieving super intelligence basically. Why I want to live forever is because I enjoy life incredibly much and I also enjoy learning and all the various experiences I can have throughout my lifetime. And I feel like those are so much more diverse than what I've been exposed to in the boss 23 years that I want to keep experiencing more things, but also especially learning more things I want to learn well, practically all languages if possible. Initially the most spoken ones, Finnish, Chinese, Hebrew, Arabic, Japanese. I already speak German, Dutch, English, and some French. But yeah, like just learning languages, learning about cultures, learning about AI and learning about how to improve my thinking. All these things, it's incredibly interesting and fascinating to me. And I would like to see whatever is possible in the films I heard to achieve that. Amazing, it sounds great. It's a good reason to live forever. Exactly. What microdose habit, 5 minute routines yield outside longevity benefits. Performing coherence, breeding for 5 minutes before going to sleep, a one-time action could be to place some books below your pillow to elevate your head to improve the clearance of waste products in your brain during sleep. And in the morning to go outside for 5 minutes and get as much sunlight as early as possible. What's the craziest longevity myth you've encountered? And is there a new truth to it? I don't know that many too crazy ones, but I mean X just eating X is being bad for your health because they contain cholesterol. That's just an old myth that has been debunked but still some people do believe it. Dietary cholesterol is taken up for around 11% from foods. And cholesterol from foods does not imply increased LDL cholesterol in blood biomarkers. Correct, like you said. It's been proven for quite a while that that's a fact. Well, Elyah, thank you so much for coming on Beyond Drone Gevity. It was really insightful and amazing to believe that you're only 23 years old with such a vision, such a creativity, and such a thirst for living forever. So thank you so much. It was a very fun podcast to be in and I had a great time. Great, thank you so much. Thank you. Adios. That was Elio Verho, co-founder of Longgevei. Elio and I not only talked about the technology itself, but the way Elio thinks about AI as a tool to support clinicians rather than replace them. There is real potential here when it comes to saving time, improving workflow and helping doctors make better use of complex health data. But there are also clear limits and a real need for caution. Overside and responsibility. We also touched on something broader. The fact that a younger generation is starting to engage seriously with longevity. Not just as an abstract idea, but as something worth building on. And whether or not we ever get close to the kind of future Elio emergence, it is clear that AI is going to play an increasingly important role in how Longgevity Medicine develops. Thank you very much for listening to Beyond Longgevity. Please subscribe, rate and review.