This Week in Startups - David Friedberg on AI-First Startups & the Future of Biology, Business & Creativity | AI Basics with Google Cloud

Episode Date: May 22, 2025

In this episode, Jason chats with David Friedberg—CEO of Ohalo Genetics and co-host of the All-In Podcast—about how AI is transforming agriculture and startups. David introduces Ohalo’s "Boosted Breeding" technology, which enables plants to inherit 100% of genes from both parents, potentially doubling crop yields. They also discuss building AI-first companies, genome language models, and the future of creativity in an AI-driven world.*Timestamps:(0:00) David Friedberg joins Jason to discuss AI Basics.(1:44) How AI leveled up hiring and operations at startups(5:46) AI and economic opportunities, complex problem-solving, and leadership's role(11:48) How to build an AI-first company culture(16:48) AI's transformative impact on biology and DNA sequencing(21:36) The “GLM” - GPT for DNA Is already in production(23:17) Biology meets AI: designing perfect plants with CRISPR and genome models(32:08) Is the “Age of Abundance” around the corner?(35:56) Democratization of creativity through AI: personalized Star Wars musicals & the future of media*Uncover more valuable insights from AI leaders in Google Cloud's 'Future of AI: Perspectives for Startups' report. Discover what 23 AI industry leaders think about the future of AI—and how it impacts your business. Read their perspectives here: https://goo.gle/futureofai*Check out all of the Startup Basics episodes here: https://thisweekinstartups.com/basicsCheck out Google Cloud: https://cloud.google.com/Check out Ohalo: https://www.ohalo.com/*Follow David:X: https://x.com/friedbergLinkedIn: https://www.linkedin.com/in/davidfriedberg/*Follow Jason:X: https://twitter.com/JasonLinkedIn: https://www.linkedin.com/in/jasoncalacanis*Follow TWiST:Twitter: https://twitter.com/TWiStartupsYouTube: https://www.youtube.com/thisweekinInstagram: https://www.instagram.com/thisweekinstartupsTikTok: https://www.tiktok.com/@thisweekinstartupsSubstack: https://twistartups.substack.com

All right, everybody, welcome back to this week in startups. We like to do our basic series. What's our basic series? Well, entrepreneurs ask us all the time. How do I get this blocking and tackling these tactics done? Well, one of my besties from the Allent podcast, David Freeberg is with us today. He's run multiple companies. He's invested in dozens of companies. And he is an expert on all things. Technology science. And of course, AI. He's running O'Hollohel Genetics. He founded it in 2019 when he was the CEO of the production board. That was his incubation. startup lab, David, welcome to the program. That was your most successful of the cohort. You had like two or three really good ones come out of the studio, and O'Halo was the one that objectively you felt most passion towards personally, and that had the best future chance. I mean, I don't want you to pick your favorite baby here, but... Yeah, I mean, we had a couple projects that we shut down.

We have a few that are still going. And then O'Hollo is just a very special company that had these very important breakthroughs where we changed how plants can be used for, plant breeding using CRISPR-type tools. So that's created a new system for plant breeding, unlocked tremendous yield potential, which is the key driver for agriculture, and unlocked the ability to make seed in crops where you can't make seed. And so the general kind of storyline is this becomes a huge game changer for the entire tea of agriculture. And so it was so important where we had these breakthroughs. I said, we can't mess this up. I'm really committed to

making sure this company succeeds. I have to make sure it reaches a potential. So I stepped in in November of 23 as the full-time CEO. Love it. And of course, today's AI basics episode is brought to you in partnership with our friends at Google Cloud. They produced an incredible report. It's called The Future of AI Perspectives for Startups and it features 23 leading experts, including my bestie here, Sultan of Science, David Freiburg. And let's get into it. I was thinking about our discussion today because we talk about AI in our group chat constantly. We've talked about it for three years. here on this week in startups and all in,

and basically taken over every conversation we have. So I want to talk to you first, in terms of running a startup again, right? You did the production board, which is arguably a startup, making startups, and then, you know,

your climate.com company before that, and of course you were at Google originally going full circle here. How has changing a company changed over, you know, those two decades you've had your sleeves rolled up because of AI specifically. Like, how is building a company, I mean?

Yeah, the company side of building, you know. Well, it's changing every week. As anyone who's building a business today can tell you, it is every job, every function, every workflow, can be improved, levered, scaled up using AI. And so, like, I'll give you an example. My chief of staff, Laura, who you know. She's great.

We posted up a job rec for like a partnership manager, for supply chain role. So I put it on LinkedIn. I have a lot of followers on LinkedIn now. Get a bunch of resumes come in. So hundreds of resumes come in. Like, how do we filter through all these resumes? So Laura asked AI to write a Python script for her to download all the resumes from greenhouse, scan them, score them, and give her results. She did this all in a couple of hours. So she's not an engineer. She used AI as a tool to create a tool to give herself an incredible amount of leverage and very quickly screening through these resumes, prioritizing them. That's a feature that should exist in one of these apps, for some reason. It doesn't.

and she was able to cobble it together in literally hours and get leverage out of it. It was a great example. We're creating a new system of plant breeding, and we wanted to do some experiments using controlled environments and LED lighting. And so we had AI, rather than have a project manager put together the whole project plan, which would take weeks or months, we had AI put it together by feeding the AI all these research papers, giving it some guidance. And I did it all myself.

Within a few hours, I got this output. And here's the project plan. And now the team is executing on it. there are so many examples of that where what would otherwise or historically have been large kind of project or human intensive workflows, you can simply create the script

or create a plan using AI in minutes. And so just the people that know how to use the tools are just getting incredible leverage. And that gives you faster turnaround. You can do more with more nimble teams. It's just an incredible set of tools. It's amazing when you think about it, the amount of time we're spent making that script

was, you know, a couple of hours. She gets all that leverage because we're taking 15 minutes, 20 minutes per resume to go through it. So now you're looking at whatever, 35 hours of time, at least. And the time it would have cost to find somebody and hire somebody would have been weeks to find an HR consultant, you know, to do this function. And you probably got better output is the punchline of all this. You probably actually did a better job. And I had a similar thing happened recently. We had a, you know, now that the wrath of Lena Khan is over, we're starting to see MNA transactions. I had an MNA transaction come in. You know what that's like.

You get all these documents and like, oh, my God, we got to sign last Thursday. And you're like, okay. I put all the documents, David, into Notebook L.M, which is a Google product. That's right. We put the old documents in the old cap table and the correspondence with the founder. And we said, tell us about this deal. What are the details? And now I've got an analyst who works for me, two years out of school, who when they go talk to the council on the other side, they were able to ask the right questions.

Hey, what about the share price? What's this? What's this earn out? What's the difference between an earnout and a bonus program to, you know, to do retention? All these conversations would have been a $1,500 an hour lawyer, et cetera. And the velocity is incredible. So are you seeing the team size be radically different in 2025 for a hollow than say team size you had at climate.com, which I think was, what, 12 years ago? Yeah. Yeah. I mean, when we sold climate, we were 160 employees. And then we grew up to, I had hundreds of people. I think I ended up with, I don't remember, over a thousand people in my org after we were acquired. But it's very different. So a lot of people assume you just have small team sizes.

And I think that that's under the assumption that the companies are doing the same things and look the same as they did a decade or two ago, which is like making an app, a Web 2.0 app, was a company back then. And you needed engineers and front end and back end and DevOps and QA and all sorts of stuff. But today, there's the opportunity to build deep tech companies, to build technically difficult

companies, to build businesses that historically may have been the realm of thousand plus person companies with 100 to 200 people. And I think that that's really the key thing that I focus on. Everyone always talks about, oh, you can have a whole startup for two people and raise $2 million and get to profitability. It's so different. VCs are dead. I don't think that's true.

I actually think the addressable economic opportunity for startups has actually grown. So startups that otherwise might not have been able with 50 or 100 people to go after a multi-billion dollar seed market, which is kind of the market that we're in, can now do so because of the leverage they get with AI. And that includes software and automation, automation in the lab for us. We're doing more automation in the greenhouse, better decision-making because of the AI. So there's a tremendous amount of leverage and efficiency that allows smaller, nimbler teams relative to big projects to address them. Venture capital money is not going to suddenly dry up and shrink down because everyone's thinking, oh, everything's a web app. It's not true.

I actually think we can do a lot more with startups out of Silicon Valley or out of any small company that knows how to use AI and that knows how to leverage these tools. Yeah, I'm in total agreement. if we were sitting in a world where there weren't problems and we solved every problem set, we could be having a different discussion, but there are so many problems we've yet to solve and so many discoveries we've yet to even conceive of that it feels to me like it's going to accelerate because of these.

You can get started cheaper and you can go faster and you can solve problems that maybe a VC wouldn't have funded because, hey, $100 million to go after this is a different number than $5 million or $10 million. And so I think you're correct. I'll give you another set of example. It's like, if I were to say, design an automated drilling machine that can get me into the earth to go find rare earth minerals.

And, you know, the fact is we have a need for rare earth minerals, right? So that's a problem that society faces, that civilization faces. We ought to get them out of the ground. So we know of a few mines, but it turns out we know one millionth of one percent of what's in the earth's crust where it is. We got to go get it. Right. And by the way, everything exists everywhere, meaning I can go get all these rare earth minerals.

They're not just in like Inner Mongolia. The reason we mine them in Inner Mongolia today is because they're at the surface level at some particular location. We only look at the surface in mining. If we went into the crust, all of the stuff we need, all of these rare earth minerals exist in North America. We just got to go deep enough. And we don't have the technology to do that.

But the energy cost is a square of the depth you go when you do mining. So the deeper you go, the cost is. squared. Incredible, yeah. 200 feet is the square of the cost of going 100 feet. And it's very hot, then you need new material science to do it. And then how are you going to get all that dirt? Where's that dirt going to go? And then how are you going to refine the ore? Every one of those steps of that deeply complicated, highly technical project can now be run in silico. And I can do generative design on a machine. I can do generative design on the project plan. I can basically take what would have what would have otherwise been a thousand plus person analyst, engineer, research program type

effort, and reduce it down to software written by software. And that's what I mean when I say like we can address tougher problems than we historically think about addressing. And I think we're kind of at the surface level today of thinking about the problems that we can address. Silicon Valley is still adjusting to this. We're still thinking like, oh, let's make a web app in a minute. Okay, cool, good for you. Everyone's doing that. We're going to get sick and tired and bored of that. What's going to happen is everyone's going to wake up pretty soon, and they're going to be like, holy shit, we can do so much more.

I can design a project plan to get me to the moon in five years for less than a billion dollars. I can design a project plan to get to Mars and ship an entire city to Mars in less than 15 years for less than $50 billion. So that's the sort of like visioneering that I think we're just beginning to do as we're realizing the potential of these tools. And that's a limitation of entrepreneurship and, creativity and just, like you said, the problem set, maybe we, imagination. It literally is. And in some ways, you look at the AI and you think, oh, it's going to do tasks for me. But it might actually open up your mind to what's possible and other vectors.

So it's almost like this coach sitting next to you saying, hey, maybe you should consider this. Maybe you should consider that. Here's some outcomes you didn't consider. And I have this happen all the time now when I'm making investment decisions. You know, you don't know a specific category. I didn't know about diamonds. We have an investment in a diamond company on marketplace. place. When I asked it to explain the diamond supply chain, I had no idea that diamonds get

processed in India and they get polished and they go from Africa to India and then to Europe and Middle East and China, America, et cetera. And I was able to catch up so quickly. So the discussion I had with the founder was like I had been in the diamond industry for a year. Right, right. It's really weird to catch up that quick. How do you make employees understand this opportunity because I'm seeing a bifurcation in my company. There's people who are AI first, and then there are people who are not. How are you as the manager, the CEO, explaining this to people? How do you do the professional development to get everybody into the mindset that you and I and,

you know, Laura and other people are doing, which is AI first? Eric is my, I don't know, I think you've maybe met Eric Andrako. He's my VP of engineering, I guess we call him, or AI. He's been with me forever, worked at Climate Corp, ran data science, at Climate Corps, built all our models, led the teams that did that. So he's working with me at Ohalo, and he's done a couple of these kind of lunchtime seminars where we actually did one, where we then followed it with a hackathon. So every employee at the company came and we did a sit down, taught him how to use cursor, and then said, pick a business problem that you're dealing with

at work, whether you work in the lab or whether you work in the greenhouse or whether you work at a desk and come up with an app, a web app, and then you're going to create and deploy that web app, by tonight. And then everyone kind of went through that exercise. So that was a good kind of set of experiences for people. The other one is, I really like this idea of challenging people when they ask for headcount about why they can't use AI to do the things that they are asking for the individual to do. That, I think, needs to be built in culturally into organizations to help managers, help leaders, and even help individuals figure out, hey, wait a second, I never thought about using AI to solve this problem, like designing a greenhouse or, you know, designing a new lab layout or things like that.

And then I would say the, you know, the other thing we're kind of doing now is these AI first offsite. So I'm doing one with my reports, my leadership team the week after next. So I'm doing kind of a full day away and we're going to go through actually systematizing how we are going to work with individuals in the organization to try and make the organization more AI first. What are we going to ask of people? How are we going to demand outputs from them, you know, what are the things that we're expecting of folks. And so we're actually trying to build from the top down. I need to make sure everyone in leadership is using these tools and then figuring out what are the methods for making sure that they're then passing that through

the organization. So it's like dissemination of any management process. I would say, you know, you've got to start at the top. First of all, you've got to make sure that you as a leader are the right person to lead in this effort. And then you can kind of get your team, basically take those processes and distill them throughout the organization, make sure you have the right levels of checks and accountability and performance standards and so on. So I think that's pretty critical. I really like what you're doing in terms of starting at the top, leading by example. You know, the leaders are in the foxhole doing this work on the front lines. And then it's really smart, I think, to take people to an offsite to do it so they can pause, you know,

the punch list that they need to do in the office. You get them into a physically different setting. It kind of opens their mind, gives them permission to try something new. This is, I think, think very smart. And define the processes. So rather than just to, hey, we're AI first. And then it's like, okay, no one knows what that means. So go to an offsite with your management team and say, what does it mean to be AI first? Okay.

So for example, when anyone asks for headcount, ask them, why can't AI do the job? Right. How have you tried to use AI to do the job or to get the things done you're trying to get done that you need to hire someone to do? Ask those questions. And so maybe every job description that needs to get approved by leadership, and the first paragraph is why aren't you using AI? And they got to fill that out.

And then they can make the request for the job. Those are the sorts of business processes that you need to build that really institutionalize this notion of AI first culturally. Yeah, and you just think a company doing what you're doing, what I'm doing, and any AI first company is doing, you compare that to a company that's not doing it. And you start to think about the gains and the velocity and the shots on goal you're going to get as an entrepreneur,

or it's just going to be three, four, five, six, seven times more efficient, more shots on goal. And if you're not doing it, you're just going to fall behind. I have a similar thing where I tell everybody ADD. That's the framework we use internally. Have you tried to automate this AI? Have you deprecated? Is there a reason why we're doing this? We're doing this for a year.

Is this actually necessary or not? And then delegate it is, you know, is there an outsource company? We use Athena, for example, for a ton of stuff now. And it's really incredible. The number of applications we get for funding, $20,000 a year, it was taking us 20, 30 minutes to kind of review every application if we want to do a good job. Now when the applications come in, or we get updates from our portfolio companies, the first thing that happens is AI scans it for these fields, for these trends,

and then puts it into the database. So you're starting with that first 15 minutes of data normalization and summary. done. And man, we went now, I think it's we're at three minutes to process each application. And that means we can process more applications. And you know the outcome of that. If you increase the denominator, man, your investment quality is going to go up. Okay, since we have the Sultan of Science here, we should talk about two things that you're an expert on. One is biology and like the AI impacting the physical world. And then we've got a couple of friends of ours, some besties making robotics. And so I'm wondering in laboratories and when you're doing literal science,

how is AI impacting that? The reason AI is fairly transformative in biology is because biology's been digitized and is being digitized. So it's being turned into kind of digital form represented through data. So going back to pre-DNA sequencing days, right? Generating DNA sequencing data was super, super expensive. That cost has come down by hundreds of thousands fold, right? So today we can sequence the entire human genome for probably under a hundred bucks. And remember, that was a hundred million dollars. Was that Craig Venter did it? Like 25 years ago out of Harvard, yeah. I think Bill Clinton was president when he went on stage and introduced it. And so, you know, the way DNA sequencing works is you take DNA, you multiply it many, many, many, many, many times over. That's called amplification.

and then you chop up the DNA into little short sequences. And those little sequences get picked up on a chip, the DNA sequencing chip, and they removed one read at a time. And ultimately, you get these little short reads. And so you've got all these little sequences and the 140 base pair long sequences that comes out. That's the DNA read. And what you then can do is statistically reconstruct that strand of DNA.

And there's multiple strands because there's multiple chromosomes. So you're basically able to put together the complete genome ultimately of the human or whatever it is, whatever the sample is that you're reading. And so all of the little pieces of that, and that's just optical scanning, Jason. So I don't know if you realize,

but it's like there's a chemical that creates a little bit of a different fluorescence for A, C, or G. And so then you're picking this up on this little optical scanner and you're running millions of these a second. So you're just taking images, images, images, converting those pixels into an ACT or G,

and that'll give you a little readout on the DNA. So now, fast forward 25 years. And DNA sequencers are super cheap, super fast, super efficient. And now with large scale compute, we can very quickly assemble genomes based on these little short reads, these shotgun sequencing reads. And that gives us insights into the sample that I'm picking up. Okay, why is that interesting? Well, that's interesting because then I can compare the DNA of an organism to the physical characteristics of the organism. So in the case of a plant, you know, what genes and what variations of genes are causing the plant to be tall or short,

have big roots, little roots. But as I gather more and more data, I can actually start to predict complicated phenotypes or physical characteristics of plants by having AI figure out combinations of genes and create a model that predicts a phenotype as a function of different genes. So sometimes it's not just one gene being on or off or one gene being one variation or another that makes a plant tall, but it's maybe 50 or 60 different genes

and there's different combinations or variations of those genes that can predict the height of the plant or predict the depth of the root or predict how much yield you'll get out of that plant and all these sorts of things. So what's happened in plant breeding, which is an area we spent a lot of time in, is you can now say,

hey, I want to get plants that are tall, that have deep roots so they can survive drought, that can make a lot of food, that can grow really fast, that are really energy efficient in terms of using sunlight, disease resistant, all the things you want. And then I can collect all this data we have of how different plants performed in different environments.

And I can look at the DNA data of those different plants. And the model starts to predict what combination of genes are going to make the plant you want. So it's almost like AI is giving you the DNA sequence of the organism and saying, here's the perfect organism, here's the perfect DNA sequence. And then I can go look at all the plants

I can use in my breeding system and then there's other technology to breed plants to get them together to make that thing. This is happening similarly in microbiology, so where they're making little E. coli cells or yeast cells that were then programming to make a molecule. So I can program E. coli, for example, to make a biologic drug. And that biologic drug, I wanted to make the perfect protein that's going to bind to a cancer cell and then trigger my immune system to kill that cancer cell. And so there's this whole world of predictive protein structure that's being used to create new biologic drugs.

And so we can look at the performance of different proteins historically, which ones bind. I can make a prediction that this protein would bind to this specific cancer cell in this way. Based on that prediction, here's the DNA sequence I would likely need in the E. coli cell to get it to make that protein. I extract that protein and I turn that into a drug. And so there are so many applications of how we are now able to treat DNA or genomes almost like a language model. And so people are calling this concept the genome language model or the GLM. GLM instead of LLM. I like that. Yeah. The GLM is effectively predicting or scoring a DNA sequence based on some objective. And that is being utilized in drug discovery, in plant breeding,

in human health. We had the Collison brothers on. You may remember this. They're the founders of Stripe. And Patrick was talking about the ARC Institute. Yes. And they funded the ARC Institute, which has put out an open source model called Evo 2. And Evo 2, what you can think about it as, is a grammar checker for DNA. So you can put any DNA sequence into it, and it will tell you whether or not that gene will be dysfunctional or functional, or if there's an issue with that gene.

And the way they built it is they put all the DNA data that they could gather from all over the world, from all these different organisms, and they basically trained this model to say, that DNA sequence looks like it makes sense or doesn't make sense. no idea why, no idea how that model works, but it's actually able to predict the functionality of a gene. And we use it in production at O'Hollah.

Wow. So we actually use it to look at the plants in our breeding system. We feed the whole DNA sequence of every plant in, and that model will tell us, hey, these genes are dysfunctional or may have issues, and that gives us a score for that plant before we've even phenotype that plant. And so there's a lot of ways that we can kind of start

to utilize these models in different production environments to accelerate biological work like plant breeding. You know, one of the things I wanted to ask you about was I've read this book back of the day, and I'm sure you're well aware of it, the man who fed the world about Norman Borlaug. Yeah. He won the Nobel Prize, yeah. Right.

He did the dwarf wheat project over 20 years to make, you know, really less husky and more calorie-dense wheat had some, you know, second or third-door effects that maybe weren't as intended. but it's very analogous, I think, to what we're doing now. That took 20 years. Maybe you could describe a little bit about the impact of that project on humanity, and then how long it might take if you were to speculate if you were doing that today. In the 1950s and 60s, the population of Earth was outstripping the supply of food and calories,

particularly in South Asia and particularly in what we today would call emerging or developing markets, third world, places like India, China, Mexico, parts of Latin America. So the population was growing faster than those areas were able to produce food to feed their population. And malnourishment was a major problem. So we needed to increase calorie production. And that calories can come from rice or wheat or potatoes, whatever I can farm to produce calories to feed people and help the population grow and survive. So plant breeding has been around for a long time. In fact, the reason we named our company O'Hollo is because of an archaeological dig on the Sea of Galilee, where they discovered, they call it the O'Hollo Two sites, these little pots with seed in it from

23 or 26,000 years ago, which demonstrates that there was actually farming happening 26,000 years ago and potentially storing of seed and plant breeding, selection of seed, which we didn't think happened until the last couple thousand years, and we actually didn't think farming started until, you know, 9,000 years ago. And agriculture was this kind of thing that allowed civilization to explode all over the world. Because we were now, instead of just finding food on the ground,

we were able to cultivate the land and make food. So that's why we named the company, Ohalo. So plant breeding's been around for a long time. And the whole objective with plant breeding, historically, is you put some plants in the ground, some seed in the ground. The plants that are bigger, you select those, so selective breeding. And then you plant the seed from those plants next year.

And every year, if you're picking the bigger plants, the plants are going to get bigger and bigger and bigger. and the yield goes up. In fact, corn, I have a little thing in my office, but I don't know where it is, started out as a little grass, this pig, and modern corn is like this, because we selected into this giant ear of corn, and that creates a tremendous amount of calories. And so the whole point of plant breeding is, you know, getting more resources produced with less resources going in. That's the magic of plants. They can take carbon out of the atmosphere. They can take sunlight and rain from the sky and they can make food, they can make molecules that humans can consume. It's a magical,

beautiful, brilliant part of nature. And so at the end of the day, what we've done historically with plant breeding is we've just allowed a random evolution of software, the DNA, to drive the yield improvement in that plant. And then when we got DNA sequencing, suddenly we got better at this and we could select the plants using genetic markers. So we would look for specific genes that we knew would have disease resistance or drought resistance or whatever. And so that data gave us this ability to be more predictive. So data made a prediction and I selected better plants. And it turns out my breeding got better and faster and I got more yield more quickly. And now we're using tools like whole genome sequencing and these AI models to make more complicated predictions, not just about the plants,

but also the crosses I want to make between two plants and the combination of genes that the AI is saying will unlock this particular feature and doing it more quickly. And then we can use AI to control the environment we're growing those plants in. We can use AI to control the conditions to accelerate the breeding outcomes. And then we can actually use a lot of lab-based tools and automation to accelerate a lot of the work and do it in the lab instead of in the field. So we move to greenhouses. Now we're moving to lab what we're calling in vitro breeding, where we actually accelerate

breeding even more by doing the whole cycle in a petri dish, for example. It's really incredible. So now the predictability. and the speed has gone like, you know, significantly higher. And that's a lot of the technology that we've developed and work on at O'Hallo is trying to accelerate both the predictive power and the turnaround cycles in plant breeding, which unlocks like a much more tremendous improvement in yield across many, many different crops. And so, yeah, that's core to agriculture.

So that dwarf wheat project that took 20 years and, you know, like probably hundreds of thousands of human hours of effort, it probably done in a year or two? Yeah, a very short period of time. And I would say, what's even more interesting is that you could have taken all of the genome of all the plants at the time. So go to the beginning of that project. Here's the way to think about it today. You take the whole genome of all those plants.

And based on data, you know, about how those plants look and perform, the software would predict the series of crosses to make the specific plant that they ended up with. And you would do it all right away. That's extraordinary. Because you're basically able to read all the series of. that software and understand what combination of genes, or one way to think about is what sub-programs are you putting together to make the plant that's going to have this kind of magical high-yield capability. And I think that's really what's so profound is we've digitized biology in a way that

allows us to not just unlock plant breeding and improved yield, but also unlock benefits in human health. You know, we've talked a lot on our show on All In about Yamanaka factors and some of these other incredible discoveries that are possible because of digital biology that we're now using to discover new molecules that can unlock, for example, the Fountain of Youth, reverse aging, reset cells to a youthful state with a specific protein where that protein does some magical thing to the epigenome in ourselves, resets the epigenome magically to being young again, and all of a sudden you look like 25-year-old Jason Calacanus again.

When you look like Leonardo DiCaprio. Exactly. I go back to my Leo era, which was a great era. And so you could conceivably also, if you knew my blood work, you could then say, oh, your vitamin D is low. We're going to give you the strawberries with more vitamin D in it. We're going to make a custom strawberry for you. Like, the mind starts to wonder, like, wow, we could, maybe we don't take vitamins in the future.

You just make better strawberries with a little more density in the things. And when we go to buy stuff at the supermarket, are we going to be buying a custom? Well, there was this guy who was the CEO of Monsanto many years ago. Shapiro was his name. he kind of gave this whole pitch when he became the CEO of Monsanto that GMO technology, they were going to take genes from another organism and stick it in plants. And it was going to allow us in agriculture to get those plants to make vitamins that aren't natively available in that plant.

And ultimately, plants were going to be this kind of agent of pharmacy, of drugs of health. So for example, one of the first products was what was called golden rice. They got vitamin A to be expressed in rice because river blindness is a major problem in South Asia. There's not enough vitamin A in the food supply. So throughout South Asia, there's this river blindness condition. And the idea was by making vitamin A in the rice, which makes it have a little bit of a golden color. It's why it's called golden rice. That population would be spared river blindness.

And you'd be able to deliver vitamin A in the food supply there. So, yeah, that was an old school idea with GMO technology. I wouldn't say that we're like trying to make strawberries that have, That's not really the goal. At the end of the day, the work we're doing at O'Halo is just trying to improve the yield and the disease resistance and the drought resistance, climate adaptation of the plants, make them more sustainable. And then in many crops that we work in, we're making seed for the first time. Many crops, people don't realize this, are not farmed by planting seed in the ground. They're farmed vegetatively. So you actually chop up leftover potatoes and put them back in the

ground in order to grow potatoes like Matt Damon does in the Martian. And so our system actually allows us to make seed and potato so farmers can replace today they're putting about two and a half tons per acre of leftover potatoes on the ground in order to grow 25 tons of potato. And with our system, you can put less than 10 grams of little tiny seed to replace that two and a half tons of potato. And it cuts out all the chemistry that's used, all the fumigants, all of the fungicides, all of the insecticides that are sprayed that are toxic on the environment, on the potatoes that end up in our food supply deals. Those all would go away by just using seeds. giving the farmers the ability to plant seed.

Amazing. So the cost and efficiency of creating these crops is going to go down tenfold or something if you succeed. Yeah. Yeah. I mean, which would be unbelievable living in a world where calories are irrelevant. That's a good point. I mean, this is a big part of like my big belief system, as you know, is I do think we're entering an era of abundance, abundant energy, abundant food, abundant resources, abundant, everything, water. Because at the end of the day, if energy, cost drops to zero or drops close to zero, you can turn salt water into regular water. DeSalle is an energy problem, yeah. That's right. And what a lot of people don't realize is actually you can create heavy ion beams

where you accelerate atoms at a very fast speed and you can bombard them with other atoms and create heavier atoms. So theoretically, we could make any element we want if energy costs were low enough. The limiting factor there is the cost of energy to accelerate all those atoms. So you're talking about the replicator in Star Trek. Could be the replicate in the future. I actually think we're probably 100 years away from that. Right. But if we get fusion energy working where we can convert hydrogen or water into electricity,

where 10 meter by 10 meter by 10 meter of water gives you enough electricity for the whole Earth for a year, theoretically, we could make elements. We could also have this ability to make all of these, you know, mining devices that go on the earth and get all the elements we want. We could take salt water and turn into regular water. And so we're very much on the brink of this era of abundance. Right now, we produce 3,500 calories per person per day on Earth. So, you know, we have, if robots are making all of our stuff, we could get robots to make

housing for us, make roads for us. The cost is effectively a function of electricity at that point. What's the cost to power the robots? So again, if the cost of electricity comes to zero and all labor gets automated, then we enter this era where it's like kind of unfathomable today of what's possible. Yeah. It's kind of mind-blowing to think of, as we reference, up on the hallucination and like being able to trust the output, that was something that was pretty

acute at the start just two or three years ago. Seems to be getting better because AI seems to be checking its work and showing its work. I wonder how does that relate to what you're doing? And are you, have you come up with techniques, tactics, best practices around this issue, which, you know, for people who are in, let's say, finance or the law, I hear them saying over and over again, hey, we got to get this right. It's not like writing a press release or a job description. You know, you can make a mistake in a job description. It's not the end of the world. But I suppose in the work you're doing at a hollow, you know, mistakes are costly and you can't make them. So how do you think about that? I think in the first run out the gate, it's like let's use these systems

and they make stuff and it's not always right. But as you know, we now have models of models. So models of models can have multiple models. You can run output in parallel. You can have chain of thought with QAQC built in. For example, you could also have an agent that just does QAQC. So the agent would read stuff, the agent would confirm stuff, would verify stuff, or you would have multi-model output. And then if all the models agree, that would be the output you would go with. If there's a disagreement, then you have a system for...

And by the way, a lot of these concepts, these are architectural, software architectural concepts. They don't need to be built by the end user, meaning they're getting built by the application layer and by the foundation layer. So companies like Google, companies that are making application tools like cursor, they are building in a lot of these checks and a lot of these multi-model layers that resolve to a better output than you would get if you had just a single-shot, single model, do a run for you.

And so I think that's, I think I'm not worried about hallucinations and quality over time. I do think that's a core part of the service. that should be provided by the application layer, or if the foundation model or the model itself is a model of models, it'll be resolved within the architecture. Yeah, it's so incredible to think when we started our careers,

you know, the issues that we had to deal with as technologists, storage, bandwidth, you know, the limited number of developers in the world, the limited number of servers we had to rack and stack. And now it's just all of these things have been abstracted away to where it is the imagination. And you and I as movie buffs,

you have this great background here, the Computer War Tennis shoes, the Disney film. You and I were talking, maybe it was two or three years ago, about the creativity of movies. We share a passion for cinema. And we were just talking about

how people would be able to create their own versions of Star Wars, Star Trek, whatever. And I just thought of you the other day. I was scrolling on TikTok, and somebody made, and I thought it was real. There was a fan film. And then I realized,

oh my God, this is all being done by, AI. They made little vignettes of the moments in Star Wars we haven't seen, Darth Plagueis, you know, interacting with and training Palpatine. And I'm looking at it, I'm going, wow, that doesn't look perfect, but it will be next year. And they're going to make their own version of that story that, you know, or fork it and say, here's what would happen in aliens and Prometheus. Totally. You know, and I'll make my own. What was their life like? Yeah. What was their life like? What was that the backstory? Yeah. Let me,

Let me paint a picture for this that I've been thinking a lot about. Because I'm thinking about what is the artist and what is culture in this context. So I'll start by restating what you said, which is I do think that there's a future where AI can dynamically on the fly create media for us. So I can dynamically create a movie. And I can tell the movie creation system, maybe it's my Apple TV, hey, I want to see the story of Star Wars. I want to see it all in 30 minutes.

I want to see it from the emperor's point of view and what he's watching the whole time and how he's remotely monitoring. And I want it to be a musical. Okay? And so like, or whatever it is. Yeah. And so I can,

and then it just renders it and I watch it. Right. So what is underlying that? So, and then people are like, oh, that's so depressing because there's no culture, there's no shared experience.

I'm like, that's not actually true. Because what you could do is the artist that creates the fundamental core of that cultural moment can define, define what is variable and what is not. And the artist will then get a whole set of tools that they don't have available to them today. So the artist could say, don't allow musical rendering,

don't allow black and white film, don't allow 30 minutes, it has to be at least 45, never show the emperor's point of view, you know, allow creation of new characters, don't allow creation of racist characters. You know, so the artist suddenly has this- The world building. The world building. They're basically coders. They're developers. And they're creating constraints and boundaries and a primary storyline from which other

storylines can shoot off of or other perspectives can be realized. But the cultural story still exists. The story of Star Wars. They could say the story of Star Wars is always that Luke is saved by Darth Vader and he kills the emperor. And that always is core to the end of the story. You always have to start with Luke. And you could be really clear about what has to be and not.

media. And then whatever the artist says is core, that becomes the shared experience that we all have culturally. And then how you watch it, how you read it, how you render it, how you experience and enjoy it is different than me. And so I actually think it doesn't diminish the idea of culture and shared media. It does in fact give the artist more tools, more creativity and just creates this like incredibly beautiful way to bring people together with shared notions. By the way, when people go to church and read the Bible. They're not all hearing the same sermon. They're like, when you go to a gospel church in the South, it's not the same experience as going to a Catholic church in San Francisco. Right. And even though they might all follow the same story that's described in the Bible,

their experience, they're rendering their point of view on it is quite different. Interpretation. But they all share that Christianity as a common kind of ethos. And I think that that's what this sort of AI-driven media landscape allows us to do is it gives artists broad creative potential and then it gives consumers broad creative rendering experiences, which is going to be really beautiful. Awesome. It could make more, it might not, you know, people who did not see themselves as creative individuals might get to experience creativity,

which I've always felt everybody should be able to be an artist. Everybody should be able to make a movie. And maybe you just are too intimidated by the tools. Maybe you don't know where to start, but like I don't play guitar, but I've always aspired to. It would be so wonderful to be able to take part in making a couple of dire straight songs. that I could shepherd around

because that's my favorite artist and Dire Straits has been defunct for a long time. I could, for me to get one more Dire Straits album, to get one more Brothers in Arms, to get one more Telegraph Road, that would be worth

$10,000 to me. Like, I would literally pay $10,000 for one more song like Telegraph Brothers and Arms that hits me on that level. And maybe the eye could give that to us. An incredible discussion.

Thanks, Bestie. I really appreciate you doing this. Thanks to our friends. at Google for making these incredible tools that we know and love. And we will see you next time on startup basics. Thanks, Jacob.