From DNA to Drugs: How AI Is Rewriting Human Biology

Dov Gertz: You can represent as a 3,200,000,000, nucleotides, in essence, the letters that represent our genetic code. And the code is built out of four bases, a, c, g, and t. So if computer code is built out of zeros and ones, then we are built out of these four different basic building blocks.Sam Nadler: Hey, everyone, and welcome to Built This Week, the podcast where we share what we're building, how we're building it, and what it means for the world of AI and startups. I'm Sam Nabler, cofounder here at Rise Labs. And each and every week, I'm joined by my friend and business partner, cohost, Jordan Metzner. How are doing today, Jordan?Jordan Metzner: Yo, Sam. Welcome back. Another new episode. Exciting guest and, obviously, a lot going on in the AI world this week again.Sam Nadler: Yeah. Lots going on. We have a special guest this week, Dove from Converge Bio. Converge uses AI to help biotech companies discover and design new drugs faster. Dove, welcome to Built This Week.Dov Gertz: Thanks. Thanks, Sam and Jordan. Thanks for having me.Sam Nadler: Awesome. Dove, we are going to jump into what Converge does, but just to quickly go over the agenda. And before I do, actually, don't forget to like and subscribe. We have over 20,000 YouTube followers on YouTube. Hit that like button, subscribe button.We have new episodes out every Friday. This week, we've built a little game for Converge, potentially a game that they can engage with people at a conference. It is in a very, you know, life science technical space. So, hopefully, this is is a is a fun little segue into their into their business. Then we're gonna jump fully into Converge Bio.And finally, we're gonna end on some latest and greatest AI news. So with that, I'd love to demo for you what I have here. This is called the cell defense arena, and it may give you a little nostalgia of some of the games we all used to play back in the late eighties, early nineties. So the goal is the bad cells want to eat the blue core in the center, and we have some controls, and we have some special powers that we can use. We can use data.We can use some other energy boost, but the real key is not to let the energy go out. And I've played this a bunch of times. I've never gotten past one minute. So here we go. Game's going on.Oh, here's a little enemies enemy entrant. Oh. Oh, I got it. Okay. I got that one.Okay. So if you can see, I can I can use data to help me target a little bit better these guys as they're attacking my my oh, low energy? Let me see if I can use a power boost. There oh, it got me. Doing my best here, guys.Like I said, I've never gotten past one minute, so hopefully, I can beat beat the record here. Uh-oh. I'm getting a lot of uh-oh. Forty six seconds. So we didn't quite make it.Obviously, this was just something fun to kick off the conversation, but why don't you tell us what you're doing? It's much more serious than this, and and, you know, it's it's a really profound mission.Dov Gertz: First of I have to say it's like a blast from the past, this game. I'm loving it very much. We definitely have to have it in our conferences. We try to do funny things in in the conferences, so please do share. Great.So first of all, I see you're on our website. So I'll tell a little bit about the company. We were founded a year and ten months ago. We've raised $33,000,000 so far, backed by likely the best investors in the world, best summer venture partners, Vintage, TLD partners, amazing, amazing funds. We're a team of 40 folks.In essence, our mission as an organization is to empower drug developers across the world with the best generative AI solutions to accelerate discovery and development. Now let's take a step back into what's happening today in the world of AI for molecular data. So following the explosion of models around analyzing text and chatbots three years ago with ChachiPT, then the scientific community has started to harness the same types of architectures that, caused the explosion of value created, from transformers in text and image and audio today on molecular data, on data from DNA, RNA, proteins, small molecules, so molecular data at large. And as a company, we are a Frontier AI lab training foundation models directly on DNA, RNA, proteins, and small molecules and offering the models that we train through systems that integrate into workflows that our customers already have in designing the molecules and understanding molecular data better. I'll give just a few words around that.So you can imagine if we look at text, image, and audio, then each one of those kind of types of information can be represented easily by computers. So text, we know that computers can comprehend. Images are just pixels. Audio are are wavelengths that we can comprehend as a as a as a in computer science. So how do we comprehend molecular data?So in essence, if we look at, for instance, DNA, then our genome, every single person that's listening to this and in this audience currently, me, Jordan, and Sam, you can represent as a 3,200,000,000 nucleotides, in essence, the letters that represent our genetic code. And the code is built out of four bases, a, c, g, and t. So if computer code is built out of zeros and ones, then we are built out of these four different basic building blocks. So that's 3,200,000,000 DNA nucleotides in our full genome that represents us as humans. So using these language models that similar architecture is trained on text image and audio, we can now start comprehending the textual kind of representation of the genetic code.So same goes for DNA and RNA and also protein data. All can be represented as text in essence, and then we harness the foundation models that are trained on similar architectures that are trained on text image and audio for the molecular data, and then we can use them to, predict molecular, activity to generate novel molecules, generate alterations in molecules, predict given patient samples if they will respond or not respond to particular drugs. So it's just opening an explosion of value of what we can do with this data. Another interesting thing to think of is because of the drastic reduction in cost of sequencing genomes, there's 400 times more sequenced DNA on the Internet than text on the Internet. So there's a massive amounts of data for model training.But with that said, there's very, very low amount of high quality labeled data, so data that we actually know what it means. So there's a huge challenge, and we're about five to ten years behind what's happening in text image and audio. But, yeah, really pushing the frontier in terms of what's possible with these models. So very exciting. We're, like, in the open AI, Anthropic, DeepMinds kind of race just, like, seven or or five to seven years behind, and exciting times.Yeah.Jordan Metzner: Awesome. Hey, Dove. So tell us a little bit, like, who your customers are. Tell us how they use your software. Tell us, like, what they're excited about.You know? Are these pharmaceuticals? Are these bio labs? Are these universities looking for research? Kind of who's using it?How are they using it? And, you know, obviously, as you mentioned, this is kind of, you know, the early innings. So, obviously, I think everyone's excited about its potential, but tell us a little bit about how how your customers see that.Dov Gertz: Yeah. So biotech and pharma companies are our customers, and we're now opening up to academic labs as well. The way they're using it, we have three systems. You could see here on our website. You could see our solutions.So three different discrete solutions. Were always meant to really plug into, existing workflows. So here's the first solution for engineering antibodies, which is, now, in the past few years, become, the most prevalent therapeutic modality to pursue. So that's one workflow that's already existing through our system. The next one is for optimizing protein yield.So one of the biggest barriers is in manufacturing and scaling up the protein yield, and you could see it here. And the third solution that is already available to our customers is our virtual cell for discovering novel targets and optimizing clinical trials through biomarker discovery and patient stratification. That's choosing patients that are correct for a clinical trial. For and and in essence, our customers use these as workflows that are very similar to what they're used to doing experimentally. Just they convert that same kind of workflow to a AI driven computational workflow, and we're showing state of the art results across every single one of these modalities that are starting to replace existing experimental workflows.Jordan Metzner: Yeah. And so are this obviously, it's, I presume, it's saving a lot of cost, but is it also saving a lot of time in being able to run more iterations faster? Obviously, there's the offset of, you know, paying for your cost of the services and whatnot. But I just I guess if they're, you know, they're trying to look for physical patients, then able to virtualize the patient, then obviously, like, that cuts down the time. But is it really just about more iterations, or is it about lower cost of testing or kind of the combination of the two?Dov Gertz: Yes. That's a great question. I think we're we're very privileged that if you think of the three, like, core pillar pillars of value that you could provide to a customer, I like to think about it as faster, cheaper, better. And we are very fortunate that our value proposition answers all three. So it is faster, better, and cheaper.Sam Nadler: I think a lotDov Gertz: of products are either one of these or two of these. We are fortunate to to to be in a world where AI has completely changed the the the the game, and and we're we're adding value in all all three fronts.Jordan Metzner: That's awesome. And so what does this look like three to five years, seven years out as the chips get better, as the labs get, you know, faster to adopt, as, you know, your technology improves, etcetera? Like, are are we gonna see all new types of drugs? Are we gonna see new types of implementation of drugs? Are we gonna see just I don't know.I mean, I'm sure you guys think about the future probably all the time.Dov Gertz: Yeah. So I'll say first of all, I'll I'll explain the bottlenecks even before I explain the future. There are three bottlenecks. There's a data bottleneck. There's a architecture algorithm bottleneck and a compute bottleneck.So the compute bottleneck currently, we're not at a scale this industry is not at a scale that that that's the bottleneck. Text and image is in essence. But in our industry, we're still not there in terms of a bottleneck. We are in a data and, correct architecture bottleneck. So we're rigorously working on creating a massive dataset where we we pay for massive data collection, and carry data for for, creating the best dataset in the world and always innovating on the model architecture front to solve the best architecture, for for that.So we have bottlenecks currently. Given that we solve the bottlenecks and the the pace that we're moving forward, us and the industry, the bottlenecks are starting to to to to we're seeing in the foreseeable future leaps in terms of these bottlenecks being solved. The future ultimately looks like is much shorter time for drugs to reach patients, much cheaper cost for drugs for patients because there's just gonna be more and more and more competition for for drugs because it's gonna be easier to deploy them and much more personalized. Because now, Evan, that's the last that's gonna come, the the true personalization. But we right now, drugs are, in essence, sometimes called personalized, but they're not.We are because of the way FDA approvals are are built, we have to have, like, very, very robust inclusion exclusion criterias, and they they aren't, in in essence, very personalized. But as the time to market will and and cost to market will reduce, we will see more and more personalization. That's my prediction. Obviously, these are predictions, and I'm a very optimistic person. But I I I do believe that we're going towards there.I also believe that that's the number one thing that we should be harnessing our capital and brainpower to solve. I mean, chatbots are really, really nice, and it's really cool that you can create a cell game in the spirit of the nineties within a minute. But can you actually improve your quality of life and length of life? I think those are the real questions that we should be asking ourselves. I was just gonna ask, is thereSam Nadler: maybe all potential illnesses and conditions will benefit from from this technology, but is there a particular illness that we'd be familiar with that, you know, is in the crosshairs on the horizon that, you know, the patient population will immediately you know, once used properly, will see immediate impacts?Dov Gertz: Yeah. So that's a very good question. I think here, I can only hypothesize, but my hypothesis is based on data. So we do have the most data on cancer, and that's where I believe we'll be seeing the most value. Also, we see the the just in terms of the market potential diabetes and and and is is a massive market and infectious bowel disease related issues and autoimmune disease in essence.I think neurodegenerative will be relatively later, and there's a lot of molecular barriers there and and whatnot. But these are all predictions. It's very, very hard to tell at this point, and it's it's a balance of complexity of the disease and how much data we have available to model it.Jordan Metzner: Dove, you know, you mentioned earlier a little bit about kind of the shortening the timelines and, obviously, like, regulations specifically in The United States and the FDA is obviously, you know, I guess, like, one barrier to shortening the timelines. Have you started to see the government start to pick up your tools or look at your tools, and, you know, what kind of change do you think we'll see from, like, a regulatory perspective?Dov Gertz: So first of all, from a regulatory from a regulatory perspective, you could think of two areas. One is the area of discovery, which is kind of maybe the first five years, like target discovery, drug discovery, preclinical studies, whatnot, is is not where there's heavy regulation. Where the regulation starts is when you start clinical trials, and that you're talking at a a bulk of about another five years, give or take, depending on the details. So the the the the lowest hanging fruit in terms of shortening timelines is in discovery where the the the the drugs are not treated on humans. So, therefore, there's much less regulatory burden.So that's where the lowest hanging fruit in terms of shortening timelines. I'm not even sure that in a decade from now, anything will change with clinical trials. I assume not. So I assume the five year bottleneck of clinical trials will remain for the foreseeable future once the model become extremely good at modeling human interactions with novel molecules, then perhaps the regulator will shorten that slightly, but that's very, very far from from from from real from reality at the moment. Yeah.So that's kind of my two cents there.Jordan Metzner: Okay. So, I mean, I guess to synthesize that, you think, like, we'll see a lot more early stage drug discovery, but we're still gonna have to go through these clinical trials and still see it. So instead of a ten year time frame, we might start to see six year, seven year, eight year time frames for getting drugs from discovery out to market.Dov Gertz: Correct. That's in the foreseeable future. In the very long term future, then also, I think clinical trials will be shortened. But I think that, I predict that clinical trials will be more successful in terms of success rates because we're gonna get better molecules into clinical trials. And with products like ours, predict the patient response better and therefore choose the better population, and and it it's good for everyone.Like, fewer people are gonna get drugs that won't help them in clinical trials, and more people will get drugs that will help them. So even in the value of, like, folks getting drugs in that horrible time where they're just looking for a novel drug and they're, enrolled into a clinical trial, will hopefully, if the technology pans out as we all hope it will, will enjoy better success rates, for for those novel drugs.Jordan Metzner: Awesome. And just one last question here. Not to know too much about how how these foundation models work, but are these all running on top of NVIDIA? Is, you know, is NVIDIA's innovation kind of a downstream impact to you guys as as new NVIDIA chips come to market? It allows you step function improvements.Is it is it NVIDIA chips are not optimal for this type of this type of learning and and transformation? Or just tell us a little bit about how it how it may differ or be the same as kind of how the LLM market is.Dov Gertz: Exactly the same. No difference at all. I'll just explain that within this the the foreseeable future, that isn't the bottleneck for this industry, but it might tip very, very quickly. So this is an extremely dynamic market. We're one breakthrough away for it to become compute bottleneck and and see the scaling laws that we see in in text image and audio.But we're we're in in my opinion, we're not there, and no no no modeling organization is there yet, but we're a moment from there.Jordan Metzner: Yeah. I mean, I know for us I mean, sometimes it's hard for us to get GPUs to run trainings and things like that as the market is, like, on satiable demand. I can imagine as as these as these pharmaceutical companies start to really hit the GPU consumption, start really eating a lot of tokens, you can see kind of an additional, you know, kind of player taking chips out of the market as they're using them as well. Right? So more people all with the same, you know, supply chain, I guess, so that'll kinda increase demand and decrease supply maybe.Dov Gertz: Yeah. True. We're talking about one of the largest industries in the world, and this is what we believe to be the largest financial opportunity in the history of life science. So once I mean, we're already seeing this in consumption by pharma of textual and image models within their context. Okay?So for filing better regulatory affairs and analyzing data and and whatnot and designing clinical trials, a lot of that is through textual and image related models. So we're seeing a massive increase from budgeting in in a on a token level from pharma companies. But I'm talking about the molecular level, that's gonna explode, and we're gonna see a whole new, demand emerging from from pharma there.Jordan Metzner: Yeah. Totally. It just seems like there's not enough chips to go around, or at least there won't be for a while.Dov Gertz: I I don't think that's gonna be the the massive bottleneck going forward. But but yeah. I I think it's obviously a bottleneck, butJordan Metzner: yeah. Yeah. Well, the data centers will get built. The, you know, the power will come. All those things will happen.So, yeah, I'm sure that it'll it'll fill the market's needs as they come along.Dov Gertz: I think we're gonna we're gonna we're gonna have to solve this as a as a as a species because it's just so so necessary for for our day to day currently. We need to build those fun and computer games in a minute. So yeah.Jordan Metzner: Awesome. Alright, Devil. That was a great great history and great story about your company. I think we can kinda jump into the news this week. But anything else that we didn't cover, me?Sam Nadler: No. Let's jump into the news. Hard pivot. Thank you, Dove. That was super inspiring.I'm super excited to to continue to follow your story. Hard pivot here to OpenClaw, just a, you know, fun AI product that's taken the AI world, I guess, the last several weeks by storm. You know, I've seen a couple I haven't set up my own OpenCLaw. I know Jordan has, and I've toyed around with with his a little bit. But I've seen a couple podcasts with Peter.I know in the last couple days, he he decided to join OpenAI. But, Jordan, I know you've you've built your own OpenCLaw. What's your what's your thoughts, and what's your thoughts on him joining OpenAI?Jordan Metzner: Yeah. So I I guess I'm pretty early adopter. I'm deep on OpenCLaw. I installed it when it was still two names previous at Claude Bot. You know, it's fun and interesting, but the value is still pretty low.And maybe it's my lack of creativity or ingenuity, but I've spent a lot of time watching YouTube videos, following Twitter, following other users, and most of the things I've seen has have been tasks that I would normally use, like, an LLM for or I would normally do as one off tasks that could kind of more be automated. So I found it to be interesting in the sense that it's very much like cron job type or task based focus. I haven't been able to, like, get the best out of it yet, but I think what makes it so interesting is that when you talk to OpenCLAR or Cloudbot, it it doesn't need a lot of rules to understand what you're talking about. It doesn't need a lot of framework. So you can say something like, you know, hey.I wanna post on Reddit about, you know, this this article or something like that. Where are the best subjects? You know, go where are the best, you know, subreddits? Go research it, and it'll just go figure that out. And you probably could do the same thing with a ChatGPT or whatnot.And so you start to see that, like, maybe they're very much replaceable, but I think what starts to get interesting, and I think, you know, we're all seeing this now, is just the the launch of kind of, you know, these types of teams or agents that are managing teams and kind of becoming an agent manager. And so I think it'll be interesting, you know, how how OpenAI integrates kind of OpenCLAR or these kind of agentic types of teams. Obviously, we know they have their new from tier tool. I've been using Claude teams, which is like a multi agent kind of where they run together and they come back and talk to each other, but it seems like, you know, the the paradox seems like they can kind of go infinitely scaled up. Like, you know, if one person can manage, you know, two agents, can they manage 10 agents?Can they manage a thousand agents? Can they manage a million agents? And, you know, how how deep can that can that possibly go? And then, you know, you bring it back to Dove Software, and you think, wow. I mean, oh my gosh.How many how many, you know, aged scientists can be go doing, you know, drug discoveries? And, you know, I I think back to your you know, how many how much research is not done, Dove, because, you know, it's believed that it might fail or it's too risky, or it's too dangerous, or, you know, kind of what other things are people not doing that, you know, that by lowering the cost of testing and inevitability that that people will start to do?Dov Gertz: Yeah. I know. It's an exciting time. I I I think I think the agentic workflows are definitely gonna open up, at this level, areas of value in our in our space, in general. I have to ask a question.Just, I saw that, like, a lot of people are buying MacBook Pro MacBook Airs now because they don't wanna run it on their own, computer, OpenClaw. Did you do it on your own computer, Jordan? OrJordan Metzner: No. I just put it, like, on a I just hosted it on, like, DigitalOcean or something. It took me, like, fiveDov Gertz: minutes install.Jordan Metzner: I don't really understand why everybody thinks that they need, like, their own piece of hardware to run this thing that yeah. I just, like, ran a terminal command, and it was hosted on DigitalOcean in, like, five minutes. And I've updated it multiple times without a problem. So and I think it costs, like, 6 ball $6 a month. So I haven't found the need why you'd need a an on premise device.And I know, you know, a few weeks ago, heard Chamath on All In Podcast say on prem is the future, then kind of it's coming back. And I actually think the opposite. I think, like, in the past few months, we've seen, at least for myself, like, my MacBook is getting burnt. Right? I'm just, like, maxing out the hard drive.I'm maxing out the RAM. I'm maxing out its energy consumption. And to me, it seems like it would be much better to have more of a cloud desktop. And the more and more I get, I actually want a Mac that has almost nothing on it, and I wanna, you know, c SSH into my cloud desktop or multiple cloud desktops or multiple agents and and try to make my machine as dumb as possible. So while he says, like or he believes maybe that, you know, kind of removing more and more to on premise, I actually think the opposite where, you know, running these models is gonna be almost impossible to do yourself.You're going to need a large management of cluster management in order to do that. It's never going to be as efficient as sharing them with foundational labs. And I think that over time, these computers are not gonna be able to keep up with how big surfers are going to get into the cloud. And so I think you're actually gonna get a machine that's very lightweight, that's very cheap to produce, that doesn't have a lot of RAM or memory locally, but is able to produce is able to communicate with s h SSH through your server. So I actually see us, like, kind of going away from on prem and moving more and more to cloud and even away from the desktop itself.But, yeah, curious how you guys see kind of even just, like, internally how your engineers are developing software.Dov Gertz: Yeah. I think very similarly, like, there's no question that I I I believe very similarly to you in terms of on prem versus cloud. Like, I I I I think we're at cloud vector definitely. I mean, some of our computers have a 128 gigabytes RAM, like, to run, like, locally. I'm not sure I agree with the with the logic behind that because, like, some folks rather sometimes run things locally, but the raw the the the the general majority is is all computers on cloud.And, yeah, there's there's but sometimes little experiments people run on their desktops. But yeah.Jordan Metzner: Yeah. I remember when I first joined Amazon, I think it was, like, 2016. I was in the personalization department, and they just have, like, a box in the corner of some, like, NVIDIA NVIDIA gaming cards. That's what they were using. You know?So I I definitely see it for, like, this playtime, but, I mean, imagine, you know, every every engineer or company needing, you know, 256 gigs of RAM on their computer in order to operate. I mean, it just seems like a non scalable model versus, you know, kind of cloud compute. But I guess we'll we'll see what, you know, Apple launches re new, you know, and which way the market goes. But, you know, again and as well, like, as the foundation models go as well. Right?So whether they even allow you those abilities to run them yourself feasibly locally as well.Dov Gertz: So do you think, Jordan, the hardware providers like Mac will be inclined to opening their own clouds?Jordan Metzner: The fact that Apple hasn't launched its own cloud has, like, still shocking to me. The fact that Facebook hasn't launched its own cloud is, like, still shocking to me. And it just seems like all of them have some sort of compute and opportunity to sell that downstream. I mean, Amazon's been monetizing Mac computers and AWS for years doing, you know, device farming and things like that. You know, right now, obviously, we're in a shortage of RAM and, you know, hardware hardware prices are are going through the roof.But, you know, right now, if you want a Apple computer with more than one terabyte, an Apple laptop with more than one terabyte, it it requires a custom order. They're not available in any store. And a terabyte I mean, my computer hard drive has a terabyte, and I'm full. So, you know, I I think, like, definitely, the hardware manufacturers are far away from being able to keep up with, like, the amount of compute power we see on, you know, either, you know, GPU based or even CPU based servers. I've even seen some people say that they're they're buying, like, old HP servers, you know, and, like, sticking them in their office because they have a 120 gigs of RAM on them already, and they're pretty much, like, free.So but, yeah, I I I think, like, cloud in general to me has seemed like it was a good solution back then, and it kinda solved a bunch of problems. And it doesn't seem like it's a logical thing to go buy everyone a $10,000 computer. It seems like it's much more likely to buy a thousand dollar computers and and put that $100,000 into the into the cloud. But, again, I I don't know. I I the only thing I can say for sure is that there's going to be a lot of change.And, you know, I I tell all my developers, like, if you're not willing to change the way you work almost on a weekly basis, like, then you're not ready for this world. Because, like, if a new model comes out next week that we see a step function improvement, then, like, we, you know, we immediately move, and and that actually is a bigger dictator for us in how we work. Right? So, you know, if OpenAI launches, you know, Codex 5.4 or 5.5 and it's only available via APIs in the cloud, then, like, we're never gonna go on prem because they're not gonna allow us to. Know?You If Anthropic does the same thing, we're never gonna go on prem. Now if we start to see some of these open source models actually catch up or even, you know, surpass the closed source models, that might change the behavior. But I think then we'll just see the cloud providers providing these at, you know, at the lowest cost possible. So I think every turn, it just goes back to, like, it's better to have a professional run a cloud, and you worry about your computation locally, but, you know, it's it's ever changing. And, honestly, like, who knows what's these agents are going to invent and invent and invent and, you know, maybe we can become a lot more efficient, and maybe, you know, a 100 gigs of RAM isn't necessary because we can run things with, you know, 36 gigs or something, you know, even going back to 16 gigs or something that.So but it's super exciting and and a lot of fun. I think, like I mean, I know myself and, like, most of my developers haven't been sleeping the past few weeks just because, like, you know, as soon as as soon as the job ends, you wanna, like, tell it to do one more thing. You know? So it's probably like a weird vicious cycle, but it's been like, it's crazy how, you know, the the virality of of building has been able to expand beyond developers to kind of other people inside of the the R enterprise, and then their excitement about building stuff when previously they were never allowed to. And, you know, Sam mentioned something to me the other day, kind of you know, engineers were always kind of siloed in the sense that, like, their brains were focused exclusively on writing syntax based code and kind of, you know, don't disrupt the engineers that are writing this code, and, you know, you kind of do everything you can to prepare a PRD and a document and all these things so that when their time comes, they'll be able to pick it up.And now just the democratization of everybody we'd be able to build kinda just opens that up to the point that, you know, there there is, like, not that type of silo anymore, and and it's become kind ofSam Nadler: Great episode, Dove. Thank you so much for joining. Anything to wrap up, or otherwise, we'll we'll call it the the end of the episode.Jordan Metzner: Yeah. Dub, tell us whereDov Gertz: Sorry, Jarrod.Jordan Metzner: Okay. Tell us where people can find out about you.Dov Gertz: Oh, so our website is converge-bio.com. Please do that or follow our page on LinkedIn. We share a lot of interesting scientific content. Yeah. And this has been really good.I need the I need the game. Please send me the the cell game.Sam Nadler: I will.Dov Gertz: In ourSam Nadler: next conference.Dov Gertz: Great.Jordan Metzner: Okay. Awesome, Duff. Well, thank you so much. Thanks for joining us. Thanks, Sam.This was a great episode. Thanks, everyone, and hope everyone will see you soon.