I would actually go as far as saying that intelligence is going to become a commodity. If you're a foundation model company and you're not building physical infrastructure, you're cosplaying your business. Project Horizon is us building out one of the largest data center complexes in the United States. We're talking about a scale of compute. It gets counted in the hundreds of megawatts and soon gigawatts. So we call it reinforcement learning to learn RL to L. This is the first time I'm probably publicly saying this.

Hi, I'm Matt Turk from First Mark. Welcome to the Mad Podcast. Today my guest is Iso Clant, a co-founder of Pullside. Poolside is a foundation model lab company focused on software engineering that's currently reported to be raising a $2 billion round at a $14 billion valuation, including a reported $1 billion check from Nvidia. We dig into Poolside's ambitious project horizon, an AI factory data center at multi- gigawatt scale, and why AI labs must own energy, compute, and intelligence. ISO also unveils reinforcement learning to learn, a new path that goes beyond pre-training and classic RL. Please enjoy this deeply insightful conversation with ISO.

ISO, welcome.

Good to be here, Matt.

I should say welcome back. You and I chatted almost 2 years ago now when you guys were just getting started.

I was actually listening back to our podcast and I hadn't realized that it was the very first podcast we had done as poolside and yeah, two years have flown by. Fast forward to today at the end of 2025 the race between Frontier AGI Labs has only gotten crazier frothier. So where do you guys stand? What is the reason for Poolside to exist in a world of hyper competition with other labs?

So it comes back to why we started the company. Right? So if if you go back to early 2023 when we started Poolside, we started it because we had our own point of view on the research. So if you remember early in earlier in that year, you know, GPT4 had just come out and the narrative in the world was all we have to do to reach AGI is scale up language models, do more next token prediction, a larger number of parameters and more data. And we agree with the importance of scale and till the date still do. But our point of view was that reinforcement learning was going to become the most important scaling access for model capabilities.

Extremely contrarian opinion at that point still when we were doing our podcast as well and probably for the following, you know, 12 months as well. Today that's of course very different. uh today I think that's become consensus and I think the world has has understood that we now have an ability to continue to scale models towards more and more capable like direction and really close the gap between human intelligence and AI and and that mission was the original mission of poolside is still very much the mission of poolside.

but along the way I think what we had predicted would happen played out we were going to continue to scale up more compute was going to be required but at the same time the first paradigm of kind of pre-training of predicting the next token on the web was becoming sigmoidal and was slowing down in terms of the gains that it had. So over the last 2 and 1/2 years we caught up on the pre-training side and really on like the what we refer to as the table stakes of foundation model building and language modeling while building some pretty serious advantages on the reinforcement learning side and and now we're at a moment where those two things have come together.

All of us like as Frontier companies are you know developing our big model runs are still at the same scale. So there's series of advantages that we've built in our research. There's advantages that we've built on our engineering and those are probably some of the most important sustained ones. Uh but now we're getting to a point where scaling up our models to frontier sizes uh is also necessary for us. And hence we've recently had some announcements about the sheer scale of compute that's coming online.

And so we're competing like we're we're in this race. Uh I think I I listened back to to our podcast and I said, you know, we're going after open eye and entropic. That's still very much true. But our starting point was a little bit different, right? We said, you know, to get to highly capable intelligence that can reason, that can do planning, that can understand the world, you're almost better off putting a set of blinders and focusing on on one proxy for that. And for us, that proxy was software development. And as we've gotten more and more capable at doing longer horizon software development tasks, we've also gotten more and more capable at frankly all type of knowledge work tasks.

the world from a model perspective. I think we are all over time converging to a similar point. I would actually go as far as saying that intelligence is going to become a commodity. In that world, who we want to be is we want to be trusted by by enterprises. We want to be trusted by businesses to to power the knowledge workforce that started with coding agents and is now going beyond that as well. Uh and I think in the space, you know, so much is going to transition, right? We're like a pre and post electricity moment. So I don't think it's a winner takes all market but it does seem to be that it's a small number of companies who are who are able to get there.

As part of that race towards AGI you've had uh some very big news in the last couple of weeks. First of all there is a rumored large fund raise up to $2 billion where Nvidia reportedly would be investing up to a billion dollars at 12 billion pre 14 billion post. It's a very large round that seems to be very tied to another big news that you guys did formally announce which is Project Horizon. What is Project Horizon?

Yeah. So, so Project Horizon is us building out one of the largest data center complexes in the United States. And this comes back to something I said earlier. We we talked about intelligence becoming a commodity, right? Our our view for pretty much since starting the company over 2 and a half years has been that there's three layers of the stack that fundamentally are going to matter. It's energy, it's comput, and it's the intelligence built on top.

And within this world, if you think that intelligence is going to become less distinguishable between the companies building it and becomes a commodity, a commodity probably more like oil or cloud comput like bread at the bakery, is there's two things that matter. Your ability to scale it and the cost at which you deliver it to your end user. Now the ability to scale it was frankly the primary motivating driver for project horizon but cost as well.

and it's important to think about what you could do 12 months ago versus what you could do today at the scale of compute that we were talking about you know 2 years ago in our industry you could call up a data center colo and say hey I want this much compute in 6 months and there would be space like physical space where you could deploy it or there'd be someone who has the capacity available to you today we're talking about a scale of compute that gets counted in the hundreds of megawatts and soon gigawatt but there's no one you can call

right these are built to suit data centers that are so large that no one is building them before having a tenant

and so as you're approaching the frontier and your models are getting more capable and you want to serve that intelligence to the world and you want to scale up your training as well over time now your lead time from deciding to do that to being able to do that is no longer calling up a hyperscaler or calling up another partner and having it, you know, in months. Now, we're talking 12 months, 14 months, 18 months with huge capital numbers attached to it.

And so, my co-founders say this thing to each other and probably not really for for podcast material, but it's it's a if if you're a foundation model company and you're not building physical infrastructure, you're cosplaying your business.

Mhm.

And and this is kind of the the really fundamental nature. It's not that we went out and said it'll be cool to build infrastructure. It was a necessity for the mission to be able to achieve it.

So you think you just get boxed out, right? So obviously OpenAI is a whole target project and they just announced like I think a fourth data center that they're about to build. Enthropic has a special relationship with AWS. I think AWS is building one for them. That's I guess that's what you were saying. You need a tenant in that case. The scal is willing to do it because they have anthropic as a tenant. So as an indie lab for lack of a better term like you have to own your own destiny and that it involves building your own data center right just to play it back.

Yeah. I think as a foundation model company you can go two paths right you can choose to to deeply partner with a hyperscaler and kind of you know become uh you know have them become an owner in you and and really go all the way uh and I think that's one direction but at the same time the world is getting to a point where I don't think anyone has any doubts anymore that we're now on track to to reach human level capabilities and and intelligence and in that world $29 trillion of knowledge work rewrites itself right scientific progress starts pushing beyond levels that we've ever seen

And all of it is intelligence on compute. And so we are far more in a in a foundation model company is far more in a physical infrastructure business than most people realize because the ability to scale that compute, right, and and bring it to end users and frankly do so cost effectively, right? The cost of your tokens are going to matter more and more. as our as our intelligences all become closer to each other, the ability to scale this up and and do so to end users where you know the cents per token are kind of a determinant if someone's going to buy it is is critical.

So we already started several years ago uh asking ourselves the question like what would it take to move towards this and then over time you know we learned more we observed more and and then we started acting towards it and and the acting towards it is project horizon. So project horizon in a nutshell is today announced a 2 gawatt campus we can actually go far beyond 2 gawatts on the site that we're in. It's a partnership with an incredible family out of Texas called the Mitchells. The Mitchells have one of the single largest parcels of land in the United States. is half a million acres.

And it always blows my mind because for context, LA is 300,000 acres. And on that land is renewables being built. There's grid connection. There's water, but there's also a 20-in main gas line. And that offers us a possibility to start scaling up energy that powers data centers incredibly fast and incredibly large scale.

Mhm. And so what you're seeing from us is that before the end of this year, we're starting construction on a 250 megawatt data center. It's natural gas powered uh with grid connection as a backup and it will house an incredible amount of compute, but it will be the first phase of many that we're building out.

Great. So just to unpack some of the things you said, you mentioned the impact of owning your physical infrastructure on the economics of Frontier AI labs. to help us understand understand that obviously one key question in the world of AI has been gross margin where do you think owning your own uh physical infrastructure lead you towards in terms of like overall structure

I'm not yet convinced that gross margins in our industry will look like a SAS company like 80% plus I think when we're talking about um a commodity as intelligence that we build value added services on top right as a foundation model company on one hand you you sell your tokens right your barrels of oil, your raw material. And on the other hand, you're building up value added services, your products that you bring to end users and customers that you know unlock value for their businesses or for consumers in the case of others.

And in that world, I think it looks a lot more like a cloud company. So I think when we're going to look at the scale of this, uh, it's going to look like cloud companies with a zero behind that. from a margin perspective probably pretty similar because if you think about what is an AWS or GCP or an Azure it's effectively virtualization of hardware with services on top. I don't think foundation models are are that different. So I think from a margin profile will probably sit far closer towards that like 40% mark that you see in in cloud companies than you do like the 80% on the software side.

And now when you start thinking about what sits in the stack of cost of intelligence, you effectively you've got the land, um you've got the energy, you have the data center, you have the chips inside the data center and and just for context, if we for instance break this down with illustrative numbers that are roughly in the right ballpark, uh to energize 250 megawatt is about half a billion dollars worth of physical infrastructure. Mhm.

So in the case of gas turbines, it's gas turbines, but there's different paths to it. Building a data center like the power shell that brings together all of the equipment where the compute can run inside, you're talking kind of north of $2 billion, $2 to $2.5 billion. Now the compute that goes inside today would be about $5.5 billion.

being the chips,

being the chips and the networking and and and everything that kind of is what we'd refer to as the fit out of of a data center. And so you got about an $8 billion like cost of 250 megawatt. So when you hear in the in the industry you might say a gigawatt is 40 or $50 billion. That's what people are referring to, right? It's it's the breakdown of those costs.

Now the life of a chip has a certain amount of years to it. Uh and in our industry there's there's lots of you know argumentation about how long they last and it's a demand and supply question. A data center has kind of traditionally, you know, a 15-year life, but it needs retrofitting as as new generations of chips come online. And your power infrastructure can last a very long time, but requires maintenance.

Now, in that $8 billion project of 250 megawatt, you're annually spending somewhere north of 300 to $350 million of opex. And so, you can split that about 50/50 between the cost of financing and the cost of energy and operations. energy like being the most.

And this is where there's already an interesting insight. All of a sudden, you realize that the energy part today is not the biggest part of the stack, right? You're talking $160 million a year in energy. And of course, we're I'm taking numbers that are West Texas numbers. So energy can be, you know, double that in other places in the country, but still compared to the total, you know, cost TCO of of of compute, uh, it's relatively, you know, minor. Uh, numbers are still very big.

Now the reason I I mentioned is that what is going to happen over time right like what's always happened in capitalism is that margins compress everywhere now margins have already compressed for energy we already you know try to make energy as cheaply as possible in the world data centers effectively have already been increasingly more commoditized over the last 20 30 years GPU compute today is still very high margin business and and over time we'll find a similar level of compression

and as the chips become more compressed the other parts of the stack are more important as well. And when you end to end own all of this and build all of this, everywhere where previously margin sat in between, you know, falls away. If that's the person who you were paying for, you know, the cost of land, if you're talking about the cost of like the energy, the cost of the building, the data center, kind of all the layers in the stack. This adds up to a very large amount.

because the foundation model itself is, you know, it has capex for building it. that's your R&D cost and you you know you advertise it over time. Uh but the physical infrastructure is really frankly where the majority of your cost sits and so I know I went super detailed straight away but I think it's useful for people to kind of get this sense of where it is. So when you take all those margins out all of a sudden you can start seeing that you can serve your tokens 20 30 40% cheaper than someone else.

And this is going to matter because it's we're shifting from, you know, human labor that leads to economic output to intelligence on compute combined with human labor. And we've barely started this in the world. And so we're going to be in a world where this is a commodity that will look far larger than we've seen cloud compute be. Frankly, is often the largest bills of companies, right? What they're what they're paying in that world. Yeah. Controlling that cost is a big impact. But cost is really only part of the answer here. You have to be able to scale it. You have to be able to say if I next quarter need more compute, can I bring it online?

Because there's no shortage of chips. There's just a clip yesterday from Satcha uh on a podcast that I saw where he said, "Look, it's it's energy and and data centers that are my bottleneck. It's not chips." And it's true, right? Nvidia does an incredible job at supplying the market with chips. and and TSMC and everything in the stack and scale up to to huge levels but physical space bringing power online and building data centers well that's when we kind of in tech all of us start realizing the real world hits us

That sounds great. Unpack that. So I think the industry has seen the XAI data center being built in record speed which I think has reset the bar for for everyone. How do you guys go about building that data center?

So if you think about data centers, there's kind of been a traditional way of building them and I wouldn't say new way because it's been around for for over a decade. But on one end you have kind of stickuilt buildings. Everything is done on site, right? So all the materials are brought there. Thousands of tradesmen who who are assembling the data center and on the other hand of kind of like Verdive here in uh in the west or players like Day One and their parent company out of China who've gone full modular. to think of like your your data center units are rolling out of a factory.

And what was interesting in in that both of those approaches uh have their limitations, right? The moment you're bringing thousands of people to site uh and you're assembling everything and and especially in what is effectively the Peran basin, the desert, right, in in West Texas comes with its challenges both on mobilizing workforces, but also the weather, the dust, uh the logistics, the supply chain. But then the factory side has not yet scaled up at a level in the west. I would say this is different in China uh where you can actually you know bring out gigawatts like over the course of you know like couple of years

uh and so we we kind of decided to meet in the middle uh and took an approach where we said we're building the actual building stick built so that's being built on site uh think of it as an extremely long corridor uh with leaves attached to it but let's take a modular approach but one that fits on the back of a flatbed truck. So how do we kind of do 2 and a half megawatt computeworthy skids for think of this for electrical for mechanical and cooling and for compute. Those are kind of the three elements that sit inside a a data center

and do it with manufacturing partners here in the United States. And what you start finding here is and once you start breaking the problem down you realize that the electrical modular components can be done by lots of great manufacturers. This is not unique. This is so you actually have great players including very large ones uh that can help you there. when you can kind of get to the cooling and mechanical part. uh it's a smaller number of companies but again have been doing this for a long time highly reputable uh and you can work with

and then when it gets to compute side there's recently been an unlock uh because on the compute mixed with cooling side what you're finding is that now we have you know direct liquid to chip cooling the designs are changing there's a lot more that you can do in a far more contained manner and so we decided we said we'd rather have a little bit more redundancy on every individual kind of room that we build so that we can bring them incrementally completely online.

So where most of the data center is being built for AI factories today, you'll see them bringing them online at kind of 40 megawatt chunks. We said what if you designed it in a way where you could bring 2.5 megawatts of compute online, you know, as you're doing construction. Think of it again on this hallway as you're bringing on like units of kind of a thousand GPUs or depending on the GPU type can be less in the future.

uh you're bringing this online and the combination of those things that we refer to as as incrementally delivered hybrid modular data centers is kind of bringing the best of both worlds. It usually derisks your ability to deliver. Uh it drisks your ability to bring compute online. It also means that if all of a sudden next month you need more compute, you can make decisions on 2 and a half megawatts like I just want a little bit more. On the other hand, you're taking advantage of the fact that for the kind of bigger construction, the building and stuff, you take the traditional approach and on the manufacturing side, you can partner with multiple manufacturers and you can do so actually very geographically advantageous as well. Most of the things that we are bringing to our site are actually manufactured in Texas or very nearby.

And so now you're actually reducing as well like your supply chain risk and your time like to sight risk. The biggest advantage though of all of this is not just speed, but it's also your workforce that you can mobilize because we're in a quite remote area and that remote area offers all of these advantages, right? Like pretty much infinite land so we can, you know, build a single level horizontally uh an incredible amount of source of energy, but also being remote meaning that mobilizing workforces there is more challenging.

So all of a sudden, you know, when we're building 250 megawws, we're doing it with a less than 450 person on-site workforce with very skilled tradesmen and and and and construction workers that were bringing over there and we're housing over there versus a world where you might need traditionally 2,000 plus people to build the exact same thing. So it's all about de-risking and speed.

Fascinating and very different from the world of RL. But let's precisely um switch to RL now since we chatted a couple of years ago. You mentioned that the world has changed and um everybody was very pre-training focused uh for the last few years and sort of feels like in 2025 the combination of pre-training and RL now has become sort of like the the both the state-of-the-art in what what a lot of people do. Walk us through your approach.

So when we did this podcast two years ago, you know, we spoke about our approach of of reinforcement learning from code execution feedback. Yes. I think at the time I called out we had tens of thousands of these environments where the models would go in they would be given synthetic tasks explore solutions and then we rewarded on the solutions that

uh over time that has grown a lot across two axis. One is right now we have over a million of those environments uh and tens of millions of revisions putting us in like the 30 million range. So just yeah two orders of magnitude larger over the last 2 years and that's because as you increase the diversity of the type of problems that the models need to learn in you can increase the capabilities of the models. The second axis is that the models got so capable in their reasoning and longer time horizon capabilities multi-step like complex planning and execution that RL moved from kind of singleshot reinforcement learning to agentic RL.

So meaning that now we give a much higher level task. We send in an agent that agent goes and you know tries to solve for the problem by using its tools. So editing code, executing commands, updating packages, like doing everything that a software developer would do. And when you see it, you know, you see it today in coding agents, it looks even quite natural to how we do things, right? Increasingly more like how we do it. And then it gets rewarded along the way as well. So what we're seeing is that we're entering a world where the reasoning capabilities of models and their longerterm horizon planning and execution is now getting so capable that the tasks we provide them are no longer simple tasks like the questions at the end of a a textbook but now it's like the projects that you you know you're made to do in school uh where you have to do a lot more steps

where we have again found ourselves in a point where we now have again a contrarian opinion on the future. Uh and it's really not by by trying to be contrarian. It's just that we I think have had a pretty had a couple of you know a long time thinking about this. Right. I started thinking about this first in 2016 when I was building source and did the first RL with with LSTMs with language models and now the last two and a half years with with these larger LLMs and our view now has become that the world is going in a direction again that we actually might not fully agree with on so what you're hearing a lot about at the moment is the scaling up of environments with human expert rubrics.

So it's not just going into coding where we've got verified rewards. Now we're entering into areas if that's you know how to operate on a spreadsheet or if that's how to do a complicated task as a as a chemist or you know marketing uh professional where you use kind of a rubric created by an expert uh to make it part of what a model can use to judge the answers and it's highly effective uh and by the way we do it as well. It is a highly effective way to to get models singular capabilities on singular tasks uh to push it up.

But the narrative that that's going to scale us to AGI, yeah, we don't agree with. We think this is the right thing to do in the awkward teenage years ahead of AGI. It's because it allows you to create really economically valuable outcomes for your customers. But the singularly collecting of skills this way by creating environments and experts, we don't think will scale to artificial general intelligence.

Our view is that reinforcement learning will have a moment similar like we've seen in language modeling. So, what makes language modeling so beautiful? It's the fact that by predicting the next token on the web, you have this general self-supervised objective, right? You just can throw the entire internet at it and of course continue to filter it into more quality and improve it with synthetic, all the things we spoke about last time, but it forces the model to learn. But the reason that never got us to AGI is because well, the internet never actually included the data set of the thoughts and actions that created it. It's the final piece of code. It's the final article you write uh but not the thoughts that you had and actions that led to it.

Not the trial and error.

Exactly. Uh and so our view is that there is an generalized objective for reinforcement learning that doesn't require you know human judges or environments or reward models or anything external but can generalize over language. Uh the same way that we have seen you know next token prediction generalize. A way of thinking about it is if you have traditionally we've taken the web and we've used synthetic techniques to reformulate it to improve it. We've moved forward like we taken the web and generated forward. What if you could generate backwards?

So what I mean by that is what if you could reverse engineer the web or decompress the web into the thoughts and actions that created it. Is there such a general objective for RL that can be found? This is where we've done a lot of our research in the last year and a half and we have now started seeing a lot of promise towards that being possible.

Does it have a name as an approach or

Yeah. So we call it reinforcement learning to learn RL. This is the first time I'm probably publicly saying this.

we don't think they're mutually exclusive right we think that you want to have a model learn how to think and reason as early as possible in its training and then you want to actually have it learn in the environments to sharpen its skill sets, right? And it's not unlike us, right? Like it's we'll have learned a lot coming out of university, but then we're put in the job and we're actually doing it and we're learning from experiences.

So our view is is that you know we pre-training and predicting the next token on the web is an incredible bootstrap of of understanding language and and helping us get you know to a level of intelligence. Reinforcement learning to learn RL internally also refer to as the Bondi techniques. It's kind of our our our code name for it. We think will push models to a level of reasoning and thought that will happen far earlier in their training than it does today. And then you have reinforcement learning from code execution feedback and and from other verified environments that help learn really what is you know to sharpen those skill sets in simulated environments. And then the fourth stage of training over time increasingly will become learning continuous learning from real world experiences of these agents. And so those are kind of the four stages that we think training will go to.

You just mentioned something interesting about the model and the agent being more or less the same thing. So to to weave that question into the you know current sort of debate topics a few days ago said that agents were 10 years away. So what is your take on where we are with agents as it relates to AGI and um is there an opportunity to build agents without being a foundation model company?

So I haven't yet listened to the kopathy interview and I really want to because I don't think you noticed but I got into the space because of karapathy. So in 2015, Andreopathy wrote an article called the unreasonable effectiveness of recurrent neural networks, a blog post about language models. And that singular blog post let me start my company sourced and got me down this obsession for now a decade of like what you know language models and and can do. And so I have an immense amount of respect for for his opinion.

And so while I haven't listened yet to the interview, I can't comment specifically on on what he said. But what I can say is this, right? What is today the definition that we use of an agent? It's a model running in a loop inside an environment with access to a set of tools, right? And it's and it's doing longer trajectory work. And how are we training agents, right? Agent capabilities as foundation model companies is that we take that agent, right, the binary, the piece of code that that runs this in a loop and has access to could be a container or could be access to a bunch of tools and we train it together with the model.

And so this is why you see the most capable coding agents, right, really coming out of people who are training with the models because and this by the way will likely hold true for lots of domains. But when you already have a model that has all of the a intelligence and all of the capabilities to do an agentic task, right? So it doesn't require more intelligence. It doesn't need to be pushed further in like reinforcement learning. Well, then the question, what's the differentiation in the agent? Well, differentiation in the agent is whoever is building it needs to have access to some form of proprietary data, some form of proprietary environment, something that allows that, you know, that loop that the model is running in to to be, you know, more advantageous than another competitor.

But what I am careful about and we see this in coding, right? Maybe we see this ourselves is if you decide to start a company to build a coding agent where you are not able to improve the model, right? The agent is not able to train together with the model and foundation model companies like like ourselves are deeply focused on doing so. You you don't have that edge and so the cat-and- mouse game that you're playing becomes a lot more difficult.

But we have a phrase at poolside which is over time everything collapses into the models. And I think increasingly that's what we're seeing. Frameworks or agents or products that were 2 years ago you know around today are have you know either gone or they've sometimes their UIs have collapsed right? And so I think you have to ask yourself the question like where do you sit like what's your unfair advantage that you have and if a foundation model company decides to focus on it will that advantage you know still sustain or or will it fall away because they're able to train the models to be further and and more capable in combination with their agent and yours.

So what do you tell people that argue that AI progress is actually plateauing?

The frankly the same thing. I've I've had this question for two and a half years now since starting poolside. This is not the first podcast where you know there was a are we hitting a wall? We are I think at a point where we are continuing to see with every new generation of chips an ability to make the model larger.

Uh and it's important that this links to every new generation of chips because it is not a world where you can throw infinite dollars at scaling. It's that's a false narrative and you can think of it very easily yourself because if I take the size of a a model uh which is still effectively the biggest determinant of how much compute it takes to train it because of the limitations on the networking and because of the limitations on the flops on those chips. It is not that I can linearly add more GPUs and train increasingly a more larger model. If I do so the time it takes becomes exponentially longer.

But every 2 years and now the chip cycle is even getting less. We have an incredible new chip and an incredible new networking stack that is improving that all of a sudden makes the next generation of size model possible. Now I don't think this will scale infinitely. So on one hand we have a a free lunch like as as new generations of chips come out we can build and train larger models uh and it as continues to show that it improves the model capabilities

on the other hand and I think far more interesting as we're now getting into the world of scaling reinforcement learning we're able to train models for longer duration with more data right because I want to have this the size axis and the other thing we have to duration axis how much data we're showing the model and every single year. We've also become multiples more efficient with the data we had to make models become more capable in a lesser compute budget.

I think it's important to understand that it is not an infinite dollar you can just keep throwing intelligence. It's kind of bounded by physics of every chip generation. As reinforcement learning is becoming increasingly more scaling axis, we are able to improve models within those generations far more. But we have not yet found a generalized infinitely scalable, you know, dollars we can throw at reinforcement learning either. And so both on increasing model size and on increasing data from RL, there's still limitations. And if those limitations didn't exist, Pulside wouldn't have been able to catch up, right?

What's the current status of those products and what do what do they do product models?

So our view has been quite simple actually is that we will make models publicly available when we are clearly best on on a very valuable axis right and I think that's that's important it's not just about having a hype on Twitter for a couple of weeks and and having something out you need to bring something in that can scale and that is valuable to the world

so before and this gets to your commercial part of the question we weren't at the frontier and now we've are getting increasingly closer and closer and uh and that's al also opened up our market. So before we were at the frontier, we picked a market where no one else could go, which was the defense industrial base and government because one of the things that we're willing to do uh kind of from the enterprise DNA we have is take our entire model weights and the full stack all the way with the agents anywhere the customer needs it to be. So we today operate on workstations that go into skiffs. We operate on servers that are you know onrem or even as far as airgapped. We operate in commercial cloud like AWS in private VPC environments. Uh we also oper in more commercial just cloud regular environments and then on GVC cloud top secret cloud kind of the places where the the weights have to travel to the customer.

And we did this because it wasn't just about the model is because we knew that as the world was moving towards agents doing increasingly more work in the enterprise it was going to become incredibly important to build out everything around the model. So today we can rattle off you know a very long list of enterprise features that we've built over time so that these agents and models can actually operate in complex regulated environments. And we battle tested that in the defense industry because they're extremely large organizations they're highly complex they're highly regulated and they often need segregated deployments for different missions that that they're on

and so we've been scaling that up in that industry. Now, as the models have gotten to a point where we say, "Okay, now we feel that we can compete, we're going to wider enterprise." And so, you'll see us increasingly more showing up in kind of the large enterprise names amongst financial services, industrials, like technology companies.

But we found something over and over again, which was that a lot of these enterprises and organizations, they have incredibly high valuable problems that can be solved with AI today. Hell, in many cases with AI capabilities of 2 years ago or a year ago, but they're not successfully doing so. Uh and a lot of that has to do with the gap between what the intent of the project is versus being able to bring it all together, the right data sources, the right context engineering, often additional reinforcement learning that's needed. And here we started building up a very strong forward deployed motion. At Poolside, we have former Palunteerians and who came over and and really kind of instilled some of that DNA, the DNA of what it means to find a high value problem and come in with high agency and help a customer solve that.

You even have like FD R you call them FDR. FD is for chumps like you forward deployed research engineers.

Yeah. And and look what we found is that there's a there is a gap between the skill set of doing traditional forward deployed engineering right where you're focused on uh a high value problem and using kind of piping data sources together and and and building you know interfaces on top which is by the way incredible thing. I have a lot of respect for everyone who has done this. The research engineering side is interesting because what you're looking for is the people in that who also have the experience and the natural tendency to work very well with models and with agents and can who work on additional reinforcement learning if it's necessary. And so we're taking these highly capable research engineers and we're putting them inside our customers. And that's for us something you're going to see us scale up increasingly more. It's why we've uh going to be opening an office in the UK actually next week. Uh we're going to be opening an office here in New York and and really kind of scaling up that motion of forward deployed research engineering.

So, zooming out the next 12 to 18 months, what happens at Poolside? What can we expect?

Models reach the frontier that that's uh that's right now I think we see a straight line towards this. Uh you'll see the the scaling up of physical infrastructure to both empower training even more larger and capable models but also serving them to the world. uh you'll see us uh have forward deployed research engineers and increasingly larger number of enterprises uh globally you'll see us just continue to work towards the mission right

I mean for for us we never want to lose sight of the fact that we are building this company and and and the economic engine associated with it because we think that the world that lives after this is one where a lot of abundance can be created abundance through scientific progress that's going to happen but also abundance through the fact that costing of goods and services ultimately are dependent on the cost of what it takes to create them. And as we shift intelligence to compute, we can find ourselves in a a world where we can drive that cost down.

And if you look back at the last 100 years, there's no moment that I think any of us rather live than today, right? And so that's it core mission. It always has been. And you'll you'll see more of that. Uh and uh yeah, looking forward to probably being here in 12 months to to tell you about what happened.

Looking forward to it. Uh ISO great to catch up. Thank you so much. Congrats on everything uh that you guys have accomplished in the last couple of years. Excited to see the data center and like the new models. And thank you for spending time with us.

Appreciate it. Thank you, Matt.