I think MCP and tools are stupid. If you are looking for complex tasks, increasingly longer horizon, increasingly complex tasks, doesn't matter if it's coding or something else, you are going to be interacting with data sources, right? And you're going to be interacting with things that are installed on some form of a virtual machine. Uh, and we're putting like MCP in between. We're putting tool calls in between it where the model can just write code and interact with the system. And we're starting to see that like Laguna S does this a lot. You'll see this as well in like Frontier models. They're increasingly no longer here. We're going to stuff 50 tools in the like system prompts to no here's a virtual machine with these binaries installed. This code base you can operate in here a folder where you can write you know your memory if you want to. And the model is using code to do complex tasks.

Before we get into today's episode, I just have a small message for listeners. Thank you. We would not be able to bring you the AI engineering, science, and entertainment content that you so clearly want if you didn't choose to also click in and tune into our content. We've been approached by sponsors on an almost daily basis. But fortunately, enough of you actually subscribe to us to keep all this sustainable without ads, and we want to keep it that way. But I just have one favor to ask all of you. The single most powerful, completely free thing you can do is to click that subscribe button. It's the only thing I'll ever ask of you, and it means absolutely everything to me and my team that works so hard to bring the Inspace to you each and every week. If you do it, I promise you, we'll never stop working to make the show even better. Now, let's get into it.

All right, we're here in the studio with Khan from PSA um together with Ibu. Uh, welcome.

Thanks. Thanks for having me, guys. It's good to be >> Yeah. Fresh on the plane. You texted me. You're like, "Hey, I'm on my way to SF." I was like, "You're on a plane right now, right?" Like, "Hey."

Yeah. Fresh on the plane. You texted me. You're like, "Hey, I'm on my way to SF." I was like, "You're on a plane right now, right?" Like, "Hey." after I texted you, I realized that probably coming in with major jet lag. Going to offer some fun experiences today, but let's do it.

Uh, I mean, I think the thing I would tell guests is that they don't actually have to prepare that much because if you're truly working on this every single day, then even like what you hazily remember is going to be new for a lot of the audience that don't live in your world every day, right? Um, so 10 years ago you did a talk at Google Slush, uh, talking about the democratization of AI. Um, and now here you are like open sourcing an incredible new model that we're going to talk about, but I guess what got you into democratization of AI? Like it's not obvious from your LinkedIn or something.

No, it's not at all. Actually, I don't think it's obvious how I got in this space. Um, I owe getting into this space to Andre Kopathy. In 2015, he wrote an article called the unreasonable effectiveness of recurrent.

Yep. >> And that article, I read it and I pivoted my startup at the time overnight to working on RNN's and later LSTMs and transformer models to be able to write code. If you go to this article and you scroll down, you can kind of start seeing like this was the precursor to what ended up becoming language models. So, uh at least what he was character level language models that were starting to actually predict letters. Uh he has an example out here. There's a little Paul Graham generator and you can kind of read it in the text kind of makes sense but it doesn't uh and there's a little there's an example of code a little bit further down. It says Shakespeare

and and for some reason I read this and I went down the rabbit hole of learning everything I could about RNN's and LSTMs. Right? is pre-transformer paper and I had built a completely unreasonable belief uh that neural nets should be able to generalize to anything and everything and that language should be able to generalize you know to a lot of things that are intelligence and the ability to write code and so I started building Source which was a fully open source company trying to build uh what we used to call machine learning on code language models on code and we spent about four or five years on this uh till the end of 2019 And that sounds really cool today, but back then no one cared, right? Like no one cared. We were in the dark. Like we did things along the way. We tried applying convolutional neural nets to like the structure of code. We were you know when attention came out we were applying it to LSTMs and then the transformer paper came out and it was it wasn't obvious and what we missed throughout that entire journey that we were on the right track but we should have just kept scaling up. And today to all of us the scaling laws and scaling up seems like the most obvious thing.

But having spent four or five years of my life on working on language models on code it wasn't obvious. I have a lot of respect to folks at Google and OpenAI and others who kind of took that confidence and and kept going. Uh, we failed ultimately at the time and it kind of was like biggest failure of my career. Right. You blew $12 million of investors money which was a lot back then. Yep. you spent still a lot but uh and you spent years with like a group of 40 people just obsessing over this problem and life took a different turn and and and family kind of became a focus and I kind of kept my heads down and really frankly didn't really look at language models for the following two years kind of big mistake considering following years and be really interesting and then chatbt came out and it was kind of like a vindication it's like people started texting me I found like my old you work decks and this old talks and and throughout that whole journey you know we we kind of really had a strong point of view at the time that like as you're building more capable intelligence it should be open and open open source when we started poolside that actually wasn't the case at all and I want to be very open about when we started poolside we were like there was a premise of two things. one is this technology is not going to stop compounding in capabilities I think to most people obvious today. But 3 plus years ago when we started most people were still arguing if these were stochastic parrots or not. Uh and the second was that reinforcement learning was going to be the biggest driver for LLM capabilities. Today very obvious 3 years ago was not an opinion held or direction held at either OpenAI or or Google or Anthropic or others. And so people kind of looked down on us a little bit or like you know this this really going to work. And so we just started working the problem. uh and we never really thought about open source again. We just kept our heads down and we and we built our like knowledge understanding from scratch, right? We didn't roll out of an existing lab. So, we picked up the papers and started writing code and figuring things out. And it wasn't until the beginning of this year that me and my co-founder Jason picked up the open source conversation again. And if you go back to some of the early things on our website, it was very straightforward. It was we want to get to AGI. We want to support a world of abundance and we want to be the first company that gets there. Uh but we started talking at the beginning of this year because it became kind of obvious that the world was going in a direction that was starting to kind of like kind of like pick at us a little bit. Like it didn't this didn't happen overnight. It was like a little bit we were seeing this and we're like okay the world's going down a path. And throughout this journey, there was something that I kind of used as a as an analogy or things. I said kind of well, if I go back to back in those days, kind of 2015 or 2016, we're working on this. And I picked up a sci-fi book off the shelf. Uh, and I was reading the book about 2035, AGI is achieved, and the story would be over the following, you know, decades, right? And I would have that first chapter where everyone's trying to figure things

We just kept our heads down and we and we built our like knowledge understanding from scratch, right? We didn't roll out of an existing lab. So, we picked up the papers and started writing code and figuring things out. And it wasn't until the beginning of this year that me and my co-founder Jason picked up the open source conversation again. And if you go back to some of the early things on our website, it was very straightforward. It was we want to get to AGI. We want to support a world of abundance and we want to be the first company that gets there. Uh but we started talking at the beginning of this year because it became kind of obvious that the world was going in a direction that was starting to kind of like kind of like pick at us a little bit. Like it didn't this didn't happen overnight. It was like a little bit we were seeing this and we're like okay the world's going down a path. And throughout this journey, there was something that I kind of used as a as an analogy or things. I said kind of well, if I go back to back in those days, kind of 2015 or 2016, we're working on this. And I picked up a sci-fi book off the shelf. Uh, and I was reading the book about 2035, AGI is achieved, and the story would be over the following, you know, decades, right? And I would have that first chapter where everyone's trying to figure things

Speaker A: out. You'd get the chapter of Chhat coming out and then you would get to the chapter where the world was at a fork in the road and one that it picked was one where three or four or handful of companies were going to create all of intelligence moving forward. And when I thought about that story, it felt like a dystopian sci-fi book, not a utopian sci-fi book. And the reality is I'm a utopian sci-fi guy. like uh and so we kind of took a step back and said hey can we play a role here uh now it was easy for us to do so because we were not at the frontier but we were at the frontier I don't think we could have changed our mind uh and I don't mean this like it's when the moment there's too much capital involved too much expectations you've built up things right we're small team you know just improving and improving and so we we knew that we could make that decision now but it would be a lot harder to make as we got closer and closer to the frontier and caught up to others and did a lot of soulsearching and a lot of conversations uh and said, "No, this makes sense." Even if there's big unanswered questions like how the hell do you build a business model with foundation models about open source, big open-ended question that we do not fully have the answer to yet, right? At what point do you no longer want to release open source models because misuse of models has, you know, real potential risks associated with it? Uh how is the government going to respond to open source? Uh but I think it all just came down to one thing and I'll stop the monologue is the fact that I rather live in a world that has a 100 foundation model companies than a world that has five. even if I was one of the five. And the smallest and most meaningful contribution we can make for Hunter to exist is to open up our research and open up like our weights right now and kind of figure out along the way how we can like do more.

Speaker B: Yeah. I think if anything over the past 3 years that has become a bit more true. Um you are one of a cohort of Neolabs that people are now calling that and you know we're we're doing this on the day that Thinky launched their uh new model and you are our performing dev on their on some benchmarks that they released right like they just don't have it yet. Um so it should goes to show that I think like actually this is one of those things where like there actually is room for multiple players and you are seeing a little bit more of the feature maybe more like 20 not 100 but like you are one of the 20. I I really hope so. Right. I think we are I'm I'm excited about their release. I'm excited about everyone kind of releasing because like ultimately like choice competition is is both going to drive progress in the right direction. But the fact that like you know we we create models and while we all you know drink out of the same well of data effectively, we do introduce very different behaviors and biases in our models. Some are intended biases, some are completely biases. And if we shape up in an ecosystem in the world where open models are going to be a part of the token economy like I don't think there's any question about it anymore

Speaker A: uh then we want to be able to live in a world where companies countries people can choose and say hey I am most aligned and I trust most this provider for these kind of things. Yeah, >> I think more than just one of the 20 Neolabs um up until recently, you know, most of open-source innovation was coming from the Chinese labs, right? So there's the Deep Seek of the West is that today, okay, maybe it's thinking machines reflection, but there aren't many, right? So, um, one of the things you guys started sort of in France, Europe, but very much now you're kind of taking that American standpoint and more than just that, the point is the Chinese models that we see, they're not super open research. Uh, the work you put out is, I think, some of the best. So every few months you get not only frontier models but also here's a sort of breakdown blog paper technical report of here's everything for state of whenever to build you know frontier intelligence and you're filling that gap too right so not just only open weight not just western but also pretty open research

Speaker B: no I appreciate it look I think it's I think it's actually the most meaningful contribution rights are a binary let's call them what they are yes we can modify them we can change them but like giving someone the weights does not allow ow them ultimately to recreate what you're doing, right? And and so now there's challenges around releasing data sets, challenges around like release certain things, but being able to share your research like right, how did we do it? What are the lessons we learned that we spent, you know, tens of thousands of experiments of compute on I think very much so. One correction though, Vivu and I I say this because it's kind of been haunting us for quite a few years. We actually from day zero were an American company.

Speaker A: and I do want to

Speaker A: definitely call it lucky, but also a lot of hard work of the team that now like that's starting to kind of show up in results.

Speaker B: Uh just because we probably won't revisit this again, but um and this is a fun recruiting challenge if someone knows the answer. What was the bug? And then we we won't tell the solution, but

Speaker A: so the bug this is you're going to test my memory here. So but I think I think I can recall. So if you so if you look at um so if you take like atom as an optimizer you have epsilon yeah which is right like in the denominator exactly in in the denominator and at the time if I recall you looked at like the early lama papers and things like that

Speaker B: people were juicing epsilon like quite a bit like they were like adding I don't know if it was eus 4 whatever like a high value for epsilon and if you think about this during training it's kind of like bit weird and counter intuitive that we're adding noise to our optimizer by just adding effectively like a random number in the denominator uh right like behind the decimal point and uh

Speaker A: and and it was like one of those fundamental moments where we had trusted this paper that was out there and we're like oh no it has to be this way it has to have this high value of epsilon but it made no sense to us intuitively like why do you have to have this so high? Like if you're just trying to avoid division by zero, why can't the value be extremely small?

Speaker B: and and that was kind of like one of those moments where you realize like okay, finding things out from scratch yourself builds a better intuition because the one thing you learn very quickly with model building is that your intuitions that you start with are going to get beaten up so hard right? like the it's such an experimental science uh that the things that seem obvious you very quickly get to learn like you were wrong and hopefully you figure out why and sometimes you don't even Yeah. Um yeah

Speaker A: so you know, one of the reasons that you uh you know when you released your new models, um Vivoo got really excited. I mean, everyone got really excited, but Vivu led our paper club on it and you guys saw it obviously. Um maybe talk through some lessons learned in that. Uh whatever you can sort of disclose.

Speaker B: and so we kind of said, okay, the state at the moment, like 3 years ago, was bash scripts and slurm and spaghetti code bases for training and like data pipelines that were kind of patched together. And we kind of looked at this and said, well, ultimately model building is a process. You're going from raw data, right? Like pre-training raw material the web, etc. Uh you're doing a whole bunch of filtering, cleaning up, transformations, analyzing. these days that's you know far more complex than it was 3 years ago. uh then you're training a model which is effectively a large distributed systems problem right across hardware that has still it's become a lot more reliable was extremely flaky back then uh and now with every new generation we get our new sets of challenges and and then you go into the next stages right there was no mid-training back then but like you go you know you're post-training and then and then your reinforcement learning and so we kind of looked at this and we says well this looks like an industrialized process this looks like an endto-end process uh that every single part of it kind of has its machinery, right? If it's your big data pipelines, if it's your crawling ingestion of the web, if it's your you know large scale distributed training and then you've got your your reliability. And we said, well, why don't we take some of the world's smartest distributed systems engineers that we knew and make them part of the process of research from day zero, not retrofitting it later on, but like really from the beginning. And that became our model factory. And so our model factory started with a handful of components. Today it's thousands of components and I kind of try to equate it to if you think about like someone who was at the very early days of Foxcon. If they had been there for the following decade, they would be able to rebuild Foxcon because they saw every decision that led to building that system and all the complexity. If you and I walk on a Foxcon today, no chance

Speaker A: right? Because we don't have the lineage and history of decisions that led to that. And so we kind of built early on from the beginning uh with a team that really understood that well the metric that we are optimizing for is the speed of an idea from a researcher to an experimental result that we can trust to then being part of the next model training. and in the and because it's such an experimental science ultimately in the beginning when it wasn't that complex you could kind of patch your way around it right but now at any foundation model company you are running I mean we're small team right we're less than 70 researchers another 35 engineers uh and we're running well I haven't checked the latest count but far more than 10,000 maybe 10 to 20,000 experiments a month and so if you look at that scale of every model run that is like it's ultimately it's it's you need to be able to trust it as an infra problem and so what we have now done over the years has gotten really good at that and just by working it and improving it and obsessing over those kind of end-to-end decisions so now what that means is that you looked at Laguna XS2 that we launched it was 5 weeks from the beginning of pre-training to launch the model that we're going to talk about today was 8 weeks from start to pre-training uh to launch we started the next model literally yesterday because we now finished the post-training required for the model we're launching you know next week or by the time this comes out today uh and we move that compute to the much larger Laguna M model that we're now training and so the model should be an artifact of someone's process it shouldn't be really a thing in itself like and we kind of treat this like the way you would look at like a SpaceX factory where yes the first rocket really hard to build, but the much harder challenge was building the factory and now they're rolling off and no one is really thinking about the next launch anymore. So, it's just another launch, it's another launch, another rocket comes off. And that's what we're trying to do with model building.

Speaker A: look like. Uh and that's frankly so when we talk about like to your question about our models I really talk about the model factory and my coolest example of these things is always that when we kick off a new run doesn't matter if it's a pre-training like big run or if it's now a post like one of 10 post-training versions we do for like pre-release or many experiments is that at any given moment the changes that somebody made that they had experimental results on from the day before make it into that run. So there's not like a cut off 90 days before or like no it's like literally from that moment because we can now trust the machine enough and then you also have to invest in a reliability.

Speaker A: So one of my favorite metrics about like Laguna S is that there was no on call events right like completely zero and actually we haven't had a meaningful on call event like to wake up for as far as I recall this entire year. Uh now there is one asterisk to that in usually the first 6 hours of launching a new model run something breaks because you set a config wrong you made a small mistake etc. So that's usually there's a little bit of intervention but that's always within like in call periods right not not on call and and I think that's starting to now compound. So the model we're releasing now I love it it's amazing but we're already on to the next one. Uh and I think that's the way it should be.

Speaker B: Uh and I think that's the way it should be. I think I also just want to point out, so for context, this was like a month ago. Um, we found it actually in the tech report. So, we just came in with, okay, new models dropped. Haven't heard about it for doing this.

Speaker A: >> We're very much like, ah, okay, look, it's like, you know, on par with Kimmy, Deepseek, whatnot, the small ones, Gemma level. Oh, it's a very cool paper on what goes into building. And then we hit this page, right? Like literally page two of tech report is uh this process allowed us to build the small model from scratch to delivery within 5 weeks applying the lessons and then I'm like oh this paper is not about here's a tech report of benchmarks and here's how many tokens it was trained on like for people that want to dive more from what we're not going to discuss on the podcast. It's all laid out here, right? From custom software that agents can use to interface with training code, pre-training data.

Speaker B: >> Yeah. The paper clock. >> Yeah. Yeah. All that stuff. You know, read the paper here. But um but I I I would like I I love principles. I think that is a good starting off point for maybe telling some stories. Maybe we can go one by one past the principles. I'll just call out the Dexter just called bought by Prefect. Yeah, it's a kind of fun. But yes, I'm actually very familiar with Dexter. uh just anything where like they trigger some kind of story.

Speaker A: >> So well I would say um well experiments code is obvious but I think one of my favorite things is uh I don't know where it is in here but early on uh and I still think this is the case actually a lot of foundation model companies people prepare their training data sets they get packaged up then they get copied over to a training cluster distributed across all of the nodes and then training starts. Uh and we looked at this like 3 years ago and we were like that makes no sense. You lose so much time because the moment you have to rematerialize a data set, you have to make a change, you have to fix something, etc. You've got all this time of like repackaging it, right? Tok tokenizing it, repacking it, moving it over to a cluster, then distributing it across the nodes. The bigger your clusters are, you start using fancy like torrent like algorithms to like distribute your data. So why aren't we streaming data into training, right? something that's very common in like just basic

Speaker B: >> just in time >> just in time like good computer science like principle and that was one of the first things that I think unlocked are the model factory because the moment you start thinking about well a training job doesn't matter if it's a big hero run or small like you know post-training experiment consumes a certain number of tokens per per second right and it's actually not a lot right from a like a data moving data perspective so we said well we have our training cluster And now we've got like our AWS kind of setup where we can build these amazing big data pipelines. We can set things up. We use Spark underneath the hood like all these things.

Speaker A: Uh and so once we made that shift into I can stream data into training all of a sudden you realize a lot of things unlock because now you don't have to wait for the whole data set to materialize. M you now all of a sudden when you're running data experiments about mixing data it's a config because you've got these data sources that are coming in and you just we have the service called blender that's in the report where we then say okay for this run I want 20% of this source 10% of this source I want this much so many epochs of repetition I want this to be you know uh shuffled in a certain way and your training job can start while the rest of the data is even still materializing uh also what it does is because all of this underneath. So for us, we treated the data layer underneath as like an immutable data layer. And that was really important like experiments as code immutable data later means that you can always go back and understand literally down to the single token at which cursor it went in on which version of the code.

Speaker B: >> yeah. and it took us a I have to admit like the first year of poolside we understood that engineering had to get great but we didn't understand yet uh that this is ultimately in support of like a good rigorous scientific progress. We were quite a few we were a very small number of people so a lot of it was yolo ideas and yolo runs. But once we realized that we treated data as immutable and code as always version and you could always track and trace every experiment end to end perfectly, you could repeat everything perfectly, right? You had perfect reproducibility. I can still reproduce runs from 2 years ago if I wanted to, right? It actually enables the scientific progress like the scientific process. And I think that took us probably about a year, year and a half into the company to figure out.

Speaker A: We also had some great hires like our co-head of applied research, Nikolai who joined us from Yandex who've been working on language models since like the early 2020s I think brought that into the company of like hey we want to have even more rigor and then once we kind of had the combination of like increasingly more capable platform that allowed people to do more but had this immutability, we were able to start actually okay every experiment is truly an ablation. we truly need to understand it and and I think we became much more scientifically rigorous in the last couple of years and the infra underneath enabled it.

Speaker B: >> yeah a lot of it's fun like even just the one you share all the ablations two picking the data sets right there's like a random small small paragraph in here where it's just like oh yeah pre-training data we have some um we have an automixer you know it trains eight small models scales them up picks the pre-training data set we don't even need to look at it I'm like wow a lot of engineering vigor there and there's just you know there's just a lot

Speaker A: >> yeah and look and we want to put out more like actually we we treat writing papers as something that we haven't earned the right for yet for a long time. So you earn the right to spend time, you know, publishing research once you're at the frontier because until then you're catching up and every minute and hour in this industry matters. Like I I obsess over not just a wall clock time from idea to result, but just general like time every day that we you know waste is is one that doesn't allow

us to catch up. But in this case, we said, "Okay, we're going to give ourselves I think we gave the team like three or four days while still doing their work. Like give everything in there." And to your point earlier, if you know your stuff, it's easy to like put it out. And so there's so many more things that we want to talk about over time and we will definitely start doing.

And as we earn more of the right, but also now have like added to our mission that we want more foundation model companies to exist, you'll see us like be way more proactive uh and just trying to keep dropping some of those like things that we've learned along the way that can help others like speed up,

which is the other cool side of this, right? It's not like back to your point, it's not just here's the benchmarks of our training. If you want to replicate here's experiments of optimizers, data sets, post training, uh, you lay out a lot of it here alongside here's your system for how to do it, you know. So it's it's really like promoting >> other people can do the same.

And by the way, I also want to make clear right we have been incredible like we've taken a lot of advantage of the fact of all the open research that others have published, right? and and and you mentioned, you know, that the Chinese labs and we I think it's important that there's, you know, from every country and every culture and background, including like Western companies like us, there's different models that come out that people can choose to trust. But I think we do have to give credit where credit's due, right? The incredible Chinese lab has done an amazing job at sharing their research and we have definitely take like been on the receiving end of taking advantage of that. So when you're on the receiving end of something coming to you, I think it's you also have kind of an obligation to give back. Do you have a favorite or underrated Chinese lad that you want to shout out?

Everyone shout out DeepC.

That's a good question. Mulan obviously for

Yeah. >> Right. They started years before Chachi BT. It was just rebranded >> and uh and and so I have like >> I remember how hard it was to work on these things >> before the rest of the world got excited about it. And so I have an immense amount of respect for people uh uh who were working on improving models when it wasn't the sexy thing to do when believing in LLMs, you know, was was going to get you ridiculed.

I remember like back in 2016 when we were doing what we'd call, you know, machine learning on code with with some of these models, you know, we would people would just frankly laugh at us like to be like, you this makes no sense. Like why are you wasting all these like millions of dollars on trying to figure this out? And uh and so I would say they're probably the one that uh I think deserves the shout out, not just because their latest model is very good, uh but because they they fought to get here.

And I think I think every foundation model company you takes time to get here, right? It took us three years to get to the model that we're that we're now going to be releasing. And now the the time in between the models is coming is counted in weeks. It's no longer counted in months or years. But this stuff's hard. Uh, and if we can make it a little bit easier for the next person, like we should all do so because if we don't do so, we're we've got a small window before models are really impacting recursive self-improvement to a level where catching up otherwise might become unfeasible and we should try to in that window encourage as many neolabs or whatever we want to call them like to start. And so one of my kind of current mi uh mission but quality is I want to encourage whoever is a researcher right now who thinks they can actually tackle this to go and leave and become my competitor like start another foundation welcome because I think we need it u I think otherwise we're not going to be in the world where uh I don't want to just be the fifth or the sixth company that wins. I want to look at a world where there's lots of choice.

What else do people not see in starting a foundation model? you know it's there's a lot of comput there's a lot of capital required a lot of compute you lay out moto factory and how to do the training but there's a lot there right that's a

well look it's uh I I in turn this is an oversimplification and I and I I always asterisk it with that because it can land a little bit the wrong way in in people's minds but I actually think you can sum down and I sum up 95% of model building to just doing you're just doing two things you're improving data or you're improving compute efficiency and I know kind of feels like an oversimplification for the incredible like gifted and skilled work people do. But if you really look at it like what are we doing? We are looking at data. We're generating new data. We're improving data. Uh and the only way to do that is to look at the data, right? That's a big part of foundation model building. And on the other hand, we come up with these incredible breakthroughs in inference, in architecture, new attention mechanisms. But what are they really doing? They're bringing compute efficiency. Now we have definitely had some breakthroughs over the years that that allow for more model capabilities but at the limit if you could train a large enough model right like you had infinite compute we probably if you had infinite comput you'd be at AGI probably already tomorrow right like it's not and so uh and let me say that infinite compute with infinite ability of much faster networking because networking and so being more of the bottleneck than than compute but uh so I do think that that's those are the main things and to just realize that this is engineering. I think it's become more obvious. But I think for quite a few years, people have held foundation model companies and researchers and others on this pedestal of like you're doing credible magic or rocket science or only like you know Nobel laureate physicists can do this. And don't get me wrong, there are some really hard problems that need to be solved. But a lot of the work that all of us are doing on a day-to-day is not sitting down trying to solve a math theorem. A lot of the work that we're doing is just really doing the basics, right? Writing good code, looking at data, improving it, running experiments, looking at plots, trying to see like, hey, trying to shape our intuitions. And a lot more people could be highly capable researchers. Uh, and I think that's it. It it feels it feels far for people to do so. I've seen in our own company, we've seen engineers become researchers because the model factory allowed them to be have a much lower hurdle of running experiments and trying things. And one of the guys on our team who started as an engineer building our our agents is a legit reinforcement learning researcher now making real progress. Uh and that happened in the span of like 6 months. Um that would have not been what I think most people assumed was possible, you know, a couple of years ago.

Um that would have not been what I think most people assumed was possible, you know, a couple of years ago. Yeah. Um I think one of the interesting moments is when you can sort of um self-host like you know for a programming language like if you can compile the language in the in the language the equivalent is can you use your own tools right you have the pool CLI you have your own models um presumably you're not only using your own models there's no way but like um you know what's that percentage over time

this is the first model that we're releasing that is starting to meaningfully contribute to our own work. It's not a it's not state-of-the-art model yet. Fable and there are very capable models, but Laguna S is really interesting. I'm going to actually pull up the quote. Penging, one of our co-heads of applied research uh said something uh last week as the as the model came out about 10 days ago, much better than frankly we had hoped for expected. and he said, uh, I have the feeling that a lot of the

Speaker A: gains in intelligence, but more from different behavior, more verification, less taking things for granted, not declaring victory early, and being way more persistent. And to be honest, those are more predictive than raw intelligence for success and human also to some degree. And this was he wrote me this on 5th of July on a Sunday and it's kind of been burned in my brain ever since because the Luna S model as you'll see it it and why it does so well in benchmarks and why it does so well in using it on a day-to-day basis is that it's just incredibly persistent. It reasons a lot. I do call that out. We have work to do on making it more efficient. We have to do work to do on offering different reasoning modes. But this is the model that has been able to do things that I never thought it could do. 118 billion 8B active model, which is not that large. It fits on a DGX Spark and still runs at, you know, 30 40 tokens a second on a Spark. Is able to solve Erdos 397 independently. It's able to do complex programming tasks. It's able to I asked it this morning to make me a Wi-Fi scanner without using any external libraries on my Mac. And it's like figuring out like the core WLAN API by really persistently trying to understand it without access to the internet. And more and more I love checking. I've probably spent 8 to 10 hours a day with this model for the last 10 days. I'm not exaggerating. I was on my 11-hour flight yesterday. I spent 10 hours reading trajectories and traces and like of the model.

Speaker A: And what I take away from it is exactly what Ping said. we are going to be able to squeeze so much more out of smaller models than I think we had imagined in the industry because yes there's intelligence and larger models are more intelligent like no doubt about it we should continue to scale up uh but the behaviors of being really persistent of being able to backtrack when you're wrong of like understanding how to interact with your environment show us that we can get a lot more out of it and this for me has created a bit of uh question in my mind the last couple of days. If you think about where we're using models today, right? We are using models say for knowledge work represents 25% of the global economy, you know, $25 trillion of work. As we scale up models and they become more intelligent, we are excited about using them more and more for pushing the frontier of science. And if you look at the frontier of science like true breakthroughs in science, they have been linked, they are linked to more intelligence in many places. Einstein figuring out general relativity is able to bring ideas together that other people would have not brought together. And I think one of the many dimensions of intelligence is the ability to do that. And it's something we clearly see that as models get larger and more capable, they're able to pull more ideas and threads together that a smaller model wouldn't be able to.

Speaker A: And we're starting to see examples of that in medicine and like in bio and other things. But if you think about the majority of knowledge work that we do and it includes building software, I'm a software developer at heart first and foremost probably. Uh although I probably can't say it that much anymore. I don't write production coding years. Uh is that what makes us good is actually our persistence. It's our ability to encounter a problem and backtrack and say I need to go figure out this bug. I need to go research this. I need to go look at the documentation. I need to like try different five different ways to see like if I can solve it. But it is not necessarily bringing three ideas together from radically different fields. And so if we are now seeing and I think Laguna S is an example that we are able to make a relatively small model much more capable than I had definitely predicted or any previous like benchmarks had shown for any model remotely this size or even larger uh at least on coding tasks that it's because of the behaviors and so now the question I have and I don't have it answered is I know at the limit so infinite model size right extremely large model and the cost of that model is going to be very expensive to run. We know this, right? So larger model ROI. Uh so I know that at the very limit I'm not going to use the world's largest model one day quadrillion parameter whatever crazy like skill we scale up to do a basic coding task. Already today I'm starting to size down for certain tasks. So it means that there is an optimal. It means there's some curve that goes as we go up the model size for knowledge work at some point we're at the peak and after that the return on investment of using a bigger model uh just doesn't make sense.

Speaker A: Now I think the question is before I would have thought that peak was really very far away. This model for me is the first sign that maybe that peak is at trillion, five trillion, 10 trillion. Maybe we can squeeze way more out of these models. I'm no longer thinking that we need two or three orders of magnitude on the largest models to be able to, you know, solve knowledge work uh the accounting, the legal, the code that we write. And so if if that holds true, it is an argument for the commoditization of models. It's an argument that open-source can win and like succeed in this world. And now it's of course a self-serving argument and it's a hopeful argument. But theoretically at the limit it works. We just have to go discover in the next couple of years of how much more we can squeeze out. Now I do want to put a big asterisk. This does not mean I'm against scaling models. I think we ultimately only succeed if we scale our models as large as our competition. I do not like I think we should not put our head in the sand and say we're going to be king of open-source small models. I think that's frankly it's a copout. It's trying to be king of your own kingdom but not realizing what the rest of the world's doing. All of us rather use a smarter, faster, you know, more model. But it's kind of a sign of hope. And so I don't want to overly state this is a good model. We have a long way to go to get to the state-of-the-art. But what hopefully people take away when they use this model is that the behaviors inside of it are what push it to be far more capable less than necessarily the number of parameters.

Speaker A: Is that mostly post-training? Like right. So it's entirely post-training. Are we done improving anything on pre-training? Is like pre-training done? No. Okay. Uh so I just wanted to cover pre-training and then we go post training. Pre-training is not done. Uh I mean look there's a part of pre-training of just dealing with skill right every new order of magnitude of model skill you are going to get new things you got to solve for that's those are ultimately you know engineering challenges uh I have a I would say a not commonly held opinion that reinforcement learning will move earlier and earlier into pre-training yeah is mid-training not even mid-training like like mid training today right is uh like if you look at um so we've been working on this for years already uh and I think the best I think the first time we saw it out in public was the deepseek zero paper uh this is a year and a half ago I think if I recall correctly um where you know you can very early on in a model as it starts capable of being able to use language etc induce reasoning uh and so the question that I kind of have is like we have this we have the data set that's the web uh and the web I think we can arguably say probably has the totality of humanity's knowledge somewhere encoded in different places. It's a huge variance degree of quality from garbage data and like once you look at pre-training data you really get humbled of like what the web is to like you know the most greatest scientific papers and best blog posts and like you know best transcripts and whatnot and so now what we are trying to figure out and have been doing a lot of work on and it's a place where maybe not as open as we're on other things but we will become more over time. uh we've been spending a couple of years really doing research

Speaker A: how can we turn the web into not just next token prediction but into a way to teach the model to think earlier in its training. uh and and i think there's a huge amount of gold to be found there. uh i think we are right now in you know we've got some drugs in the industry. one of the drugs is distillation. another drug is, you know, more environments like and they're great and they make us feel good and they make the models better and like we're all addicted to them and we'll use them, right? uh in in various different ways. uh and but ultimately i think we are still barely squeezing out of the web what we should be getting out of the web.

Speaker B: i think just next token prediction pre-training is not enough. uh and so i think we'll see some very interesting things still happen. uh and that rl in post training to induce behaviors to improve things like i think we the whole world knows how to do this now. i think we're we're scaling it up. everyone is but i wonder if we need to go as far as we're going today with environments. i'm not sure yet if you mean you're going too far.

Speaker A: i think there is a ability to generalize more from the web. uh but i also i'm very encouraged like when i look at laguna s and just post training is is what is the big impact there. uh and i see like oh wait a second just by making some of these behaviors much better we're able to get so much more out of it. it just changes a little bit the way you think about intelligence. the analogy people draw often is the rl phase is where you don't learn as much new knowledge. you reshift. yeah. yeah. so, you know, you reshift distribution and you can have it reason towards what you want. um, on your point about mid-training, a lot of mid-training is still just continue pre-training in a domain, say medicine, then you do rl. so, still just pre...

Speaker B: it's just better data, right? like i mean mid-training, i like how we invented this word like it's effectively just like you know second phase. it's a second phase of pre-training with like a really dumb way to do a curriculum, right? like ultimately what you'd want is a curriculum from token zero to token 30 whatever or 40 train tokens that really truly is the optimal curriculum for the model to learn. but mid training is essentially a two-stage curriculum on the web because we do not have the compute uh and uh effectively to to try to ablate the perfect curriculum, right? and so i'm pretty sure that you'll start see people talking soon about some other terminal two or because now now we do this right we talk stage two and stage three and stage four mid training and like but ultimately all we're doing is we're trying to assign a curriculum to the web data that we have to uh to allow the model to learn better. i think at some point as things get compute as as models get cheaper to run as next generations of compute this will become more of a continuous spectrum. i also think the reason by the way you have mid-training and like stage two and stage three is organizational right it's this is i think a thing where that we really try to avoid with the model factory is like mid-training exists because there's a mid-training team now right there's people or like people decide to focus on like a mid-training effort uh but what you really want is engineering and and skill of experiments that allows for a much more continuous spectrum that you don't you have infinite stages now we're not there. compute's not there. organization design is not there for it yet. uh but i think we'll get there. uh we'll look back on a couple of years and be think, "oh my god, it was so cute that we did our pre-training data like this in such a naive way." like we barely ordered it. we didn't really do a good job at like the kind of building that curriculum.

Speaker A: and i'll confirm that you know when i talk to some researchers that this is a lot of the focus now is like how does pre-training change and what is the next objective other than uh next token prediction. i assume you don't have the answers but you have some ideas. we have some ideas. we're not ready to talk about it yet. uh we've been working on them for years. and i think that's the one thing that's also like you asked earlier about like what's not obvious about building a foundation model company is that you are constantly balancing

Speaker B: versus like your my breakthrough pure research and and finding that balance and and adjusting the percentage to it based on where you are in the race is really important. i mean so like this is a nice way i was going to bring up auto research at some point. that's another andre invention. uh, which is like i honestly like how many objective functions can there be, right? like just try a thousand of them, set it running, whatever. it's also like you know what you're looking for. you're looking for losers like that. like it's also a thing people take bets on, right? when you say more neolabs, you're doing a version of we'll do foundation models, scale them up, next token predictors. a lot of other neolabs that we see want to take a completely different approach, right? at some level, you're right, it's all uh compute efficiency and that's the net objective, but you know, some are okay, different architecture, like vastly different amounts of compute spend. so, some are different. they're not just they're like, you know, 99% not balancing here's the vanilla and scale up. they're 99% on here's novel research that'll change everything. and i think look i think it depends when you started as well, right? yeah. when we started like the novel thing we did was reinforcement learning on code. no long that's no longer novel by far but we were like we that's where we obsessed over when no one believed in rl. so you have to kind of when you start the company you have to have to have your own idea. you have to have something that's different that allows you to speed up. right. for us it was rl to llm that later became common like you know knowledge but in the beginning it wasn't.

Speaker A: it's cool. you know, this was like your original 2023 blog of purpose and like you do lay it all out here. the blog is pretty underrated, right? the whole rl on code was very very early on. we even had to argue with people like we we say here things like to push beyond current capability to train your own foundation model. we had to argue with people that it mattered that you had your own like you know base model. uh you can't fine-tune your way to success, right? major capabilities emerged from training a base model made accurate and useful during fine tuning

Speaker B: which like for perspective at the time you know we knew closed models open anthropic were huge the open models we had were like mistral 7B a 30B a 70b no but when we actually this the date on this thing is wrong when we published this it was april uh

Speaker A: april 2023 i think this is just happened on a migration file on ar on archive mistral had started we started on the same month right so this wasn't there was only i think llama out at the time and that's it right and so uh but i I agree i think we want we want as many diversity of ideas and i do think if you're starting today you you want something that gives you an edge right uh and what i do think we sometimes over i think every arch like at the limit every architecture works and rnn works it's just not compute efficient right like if you had infinite compute, you could probably just, you know, take a basic rnn from back in the day and you could get pretty far. uh now there have been, you know, meaningful breakthroughs, attention, other things that are there. uh but you i think we are we're still we're still very early in figuring these out. the things i'm most excited about i'm

Speaker A: most excited about people doing extremely low precision training, right? So like the ternary stuff that we're seeing and it was very cool. The bonsai stuff yesterday was super cool to see. Uh, I think there's you can find tweets from me going back to 2023 which is like kind of the notion of like well it's an obvious trade-off. Bigger model, lower precision equals, you know, smaller model with higher precision by definition, right? It's just what is like how does that actually play out, right? What's the actual size limit? So you now have companies that are trying to figure that out. Like those are the kind of things that can change our industry if they're done right because ultimately like our bottleneck on compute is is is a maple bottleneck uh and and a networking bottleneck and the moment you start doing those things. So I'm excited about that. We're not doing any I mean we're doing the usual like uh Laguna S was trained in FP8. Uh only thing that in this run I have to admit that wasn't FP8 was the all to all in the new run we just started yesterday. 80 fpa it was all to all that was just like cut off day like we're not perfectly comfortable wanting to do it.

Speaker A: Uh you've got amazing work by Neimatron in NVFP4 training like I think it's underrated what they've done there. Uh I'm excited to get to NVF4 training. Uh doesn't make sense yet cuz we're still training on hoppers, right? We're like relatively small. We're 10k H200 cluster company right now. will be scaling to a lot more soon, but uh and really a lot more if someone is thinking about applying for a job. Uh but like the yes, I think it's there's so much more juice to squeeze out of this and hopefully Laguna S shows people that a model at this size can get a lot more and and we did this thing in 8 weeks. We think there's a lot more juice to squeeze out at any model size. uh we're now scaling up because it's the most optimal thing to do for us as a company, but if I had infinite time, I would love to push more the capabilities at other model sizes.

Speaker A: I don't think we've properly announced what your new sizes. So, we have XS which was 30Bish. Uh old medium was 200B which is going to be deprecated it seems. So, new Laguna S Laguna small 118118 billion total parameters 8B active. So very sparse. It's a scale up of the excess architecture. Uh it's the kind of classic or call it classic these days like 3:1 ratio of sliding window attention to global attention. It's just it's a nice size uh for a couple of reasons. One is just very costefficient for us. It was a good way to do we wanted to get our progress out quickly. One of the things that we've seen is that it's a balance inside a foundation model company between focus on releasing and shipping and like your new novel research. But with the model factory, we are able to like treat the release of a model as less of a time investment from the team because it's just oh at this moment in time you know do the pre-training run done apply the latest post-training and uh and so this is I think a nice weight class. It's one that also will fit on the DJX Spark, which uh I have a small like soft spot for. I love having that little thing like, you know, run a good model.

Speaker A: Yeah, we covered it on this pod GTC last year. Nice. I think a GPT OSS 12B was the first because it's a large single GPU, which was the H100, right? Rent one H100. Now you've got 128 gig max, Mac minis, Sparks. It's it's the home sweet spot. But I think what I'm most excited about is that this model hopefully shows people what is possible in this size because uh when you'll look at the benchmarks and and start using it, you'll realize that we are outperforming models two or three times their size. yeah, and they think so for example today's Thinky model is like a trillion pars. so yeah, exactly. And look, and by the way, I'm excited about I have it just came out. So for those of you listening to this it like I saw it on my phone coming in two seconds, so I haven't even had a chance to read the post. but somehow you are not only you're you're better than thinky which is like one of those benchmarks but also like on certain benchmarks like the tow bench one like you're actually state of the art. We're look we're doing uh I'm not sure if we're state-of-the-art on I mean 3 banking I haven't checked where we sit on the leaderboard uh but I think we are within our weight class I feel very comfortable to say and even in some weight classes twice larger that we are probably state-of-the-art. I also want to caveat this like you know best model still in the world right now is definitely you know give me a fable give me a 5.6 to your point earlier we also use other models.

Speaker A: I think that so the interesting thing you mentioned earlier is you're starting to shift a lot of your actual usage to it, right? Benchmarks are like they're good to compare, but they're not super realistic. They they have to, right? This is how they're going to dog food benchmark. No, you have to like you have to use your own models and and you have to have your own internal evals and benchmarks and and what the funny thing is is like within first 30 minutes of a new checkpoint coming out that's you know the kind of first post train after a pre-train you yourself can feel in the first 30 minutes of where this model is going to be like you don't know exactly but like when this one came out we were like oh like this is different like uh and and I think that's I think that's the example uh but it's a little bit like you know like your kids I don't know if kids by you know parents like they see their kid and it's perfect and they love it and then like you know they don't see all the rough edges you always get that when you build your own it's the most fun part is that you like you love a little bit every model that you do we try to kind of say this thing constantly it's like it's the worst model we'll ever train and so I know the team now is like already on to the next one uh as it should be because this is a race uh and this model is a moment in time that hopefully shows people that we are serious about this race, that we want to work really hard at it, that we want feedback, right? Where is it good? Where is it not? Like one of the nice things about having your models out in open way out in the world is that you get a lot of feedback. How do you think about building it with like um working with a harness, right? So open code, codeex, you have your own pool CLI tool. Um basically getting people to use it, the code design of model harness, training it in. So you need to do some multi-harness training like if you uh especially at these smaller sizes like you want to do a little bit of multi-harness training for these models to just get the right and it's very little like you don't need a lot but it's just like to get the right behaviors that you see in your harness transferring to the harness that like you other people might use it in. Uh we internally have been kind of this calling this polishing which is like you've got your model you do a little bit of polishing so that like it's able to work well on other harnesses as it is in your own. No doubt it's going to be better in your own harness. And and it's just because of like where are you putting your reinforcement learning compute, right? You're putting your RL and and your synthetic data, you're putting it to your own harness because it's the one that you understand the best and you're able to kind of push the most. Uh because that end-to-end control is what allows you make it better. Uh then transferring those capabilities is more about just making sure the model, you know, induces the right amount of reasoning and like, you know, uh understands some of the maybe more complex weird tool call formats that might exist somewhere else. Uh, and so we do do some multi-harness polishing as we call it. Uh, it's not really what drives capabilities, but it does create a better experience. I think frankly I think everyone probably does these days, but it is totally fair to see why your own harness is going to still

Speaker A: be better than others. And I think we see this with all the foundation model companies. Uh, and it's just that when you are pushing capabilities, you don't really want to trade it off by putting 10 harnesses in your RL runs because it's just complexity. It's complexity of engineering because these when you're trying to do good science, when you're trying to really understand what make my model improve, you want to make one variable change to something you understand and a harness from someone else you don't know or understand in the same way as you understand your own. Right? They might have different subaents or different prompts. Well, if it's open source, you can look at the source.

Speaker B: Yeah, but it's time, right? Like like I really cannot stress like I know I'm like a weird person on this because like I have friends like can we meet up or can we do this or can we go? I'm like no because ultimately this is a race and time is the only thing that matters. And if I look at our team and say okay what is complexity worth introducing on our general trajectory to building more capable models which generalize to other harnesses quickly and by the way our model works well in auto harnesses. I really encourage people to do it. It works well. Like I've we've been testing in open code and kilo code and others and like and in clock

Speaker A: I know Honda.

Speaker B: Exactly.

Speaker C: Everything's getting bought.

Speaker A: Uh and I and I think part of that is like you know and there's some amazing I'm excited like I think her mess is actually ridiculously cool harness like uh and so

Speaker A: But our model is still better than some other harnesses who do that in coding like tasks because it was RLED with it. Now I do encourage people I think our model by the way is perfectly fine and good that the differences are probably maybe too small for anyone to notice but we see it ultimately still on benchmarks uh by a little bit. So I think it's both are true. Foundation model companies with their harnesses will really push them because it's just operationally frankly the best way to have scientific rigor and improving your models but also someone who takes our model and really does a lot of work on improving a harness is going to out compete us as they should. Uh, and that's just because the harness is the stop gap between what the model is capable of and what it needs as additional instructions and what it needs as access to data and tools, right? And that's ultimately I think what a harness is. It's like is it able as you build more capable models, you know, you're improving the instruction following the models. And so additional harness is just saying hey if you encounter X Y or Z behave this way. And so even if you would say that two models with two different harnesses can equally reach the same capability that you care about a harness that is really tailored towards a capability will do it more efficiently. It's kind of like a person who's getting a manual of how to do the task in the most efficient way with the right tools and the right data sources versus a really smart person like go figure it out. They'll both solve the task, but one will do it a lot more efficient. So, I'm a big fan of all the harness development that's happening in the world. And we want to work with more harness like creators to also make sure that like if it needs some additional training like polishing that we will do it.

Speaker B: I mean, um I think when you say it's a race, there's a question of what are you racing to? Um are you racing to be the best coding model company or the best coding model plus harness company? I think that's a do those are different things. Or fill in the blank or neither. Uh so we I I raised AGI coding for us since day zero of our website has been and we've said this over and over again. We think focusing on coding and long horizon like software tasks is a path towards AGI because it forces us to solve the hard problems. It's it forces us to solve the ability to do extremely long horizon complex work that requires lots of reasoning, external tools, data, etc. And one of the things I can show you, so we'll we'll have a web chat on with this model and I've loved this model for deep research, just using it in my coding arts. It was never trained for it. It was never like looked at it, but it's great at it in my opinion. Uh because ultimately the skills transfer, they generalize. Now where we are not focused on today is to make sure that the world's greatest medical knowledge is encoded in this model or the world's greatest legal knowledge but it actually did we won't be publishing this benchmark because we didn't have time to really do proper but it did really well on legal bench uh and at least on our first runs uh and we are very rigorous when we publish eval we have checked them for every little thing we have run them many times we've passed like we've gone and we like we try to be extremely honest with it so if we haven't spent enough time on benchmark that we use internally that is public. We just said we won't publish it. Um and so I mean the other way is just to give it to artificial analysis and let them run it like third party.

Speaker C: Oh, 100%. And we are going to be doing this as well and and still it takes time and effort, right? Because you're working with people to understand like you know the infraures and and like the tools they're using and like are they set up well. But I agree you absolutely want to. I'm a big fan of companies like Vals and AA and like others that are doing this stuff.

Speaker A: You're the first to bring it up.

Speaker B: Yeah, I think they're great. they've got like I I love like a lot of the work they've done and put out. Uh and so and there's I think many more and please create more eval companies like create more evals. I think it's so valuable for the industry.

Speaker C: Actual monopoly kind of I feel like oh duopoly maybe you know

Speaker B: I think it can be broken.

Speaker A: Yeah because I think it can actually be broken really easily because creating an eval isn't sexy work but whoever does it everyone is happy to get a good eval. You've never like if an EVA is well constructed, everyone's celebrating it and everyone's willing to pay for it and everyone's willing like the foundation market.

Speaker B: Oh yeah. Yeah. I think creating eval Yes. But like in terms of being like we are the industry standard ones that will run TB and make sure that you didn't you didn't cheat and I'll run it the same way that you run it versus your competitor run it.

Speaker A: Yeah, that is very true. And we need that and and it's actually nice that that's like a few standard places that we all have to like, you know, adhere to. It keeps us all honest. I think that's super important to do so. uh and uh but yeah no I think our goal is to build the world's most capable models uh and right now we are focused on the coding agent capabilities uh long arising work but what you see with that is that you get a lot for free I've always said it's a lot easier for us as we get to you know sot and frontier on coding to then say okay

Speaker A: now we're going to obsess in using the model factory to add more data for places that you know we're not as strong on like could be medical or legal or any other areas um and similarly I think what we see and we see this with reasoning models a lot if you give models access to the right knowledge sources and they have capable ways of reasoning they're able to go very well into domains that are less known to them or even seamless in their training data so uh but yeah are we agent like model plus harness no we're model company uh but i think models today cannot be trained without harnesses it's not possible so it kind of is just like where before it was just the weights in the container well now there's an agent harness that's attached to it uh and but i think there's a big difference in being an agent harness as a model company than someone who's truly building an agent company i think they can do far more than we can

Speaker B: yeah um understood yeah i think that that is my minor push back if you are truly identify as a model company then make the best model for open code right instead of for pool or whatever um i think that that's not as you know that's that's minor compared to actually if the goal is AGI actually make the best model for hermes

Speaker A: right like because that is the next stage after coding i'm look and we are uh we're working actually like very closely with them because i do think like it's and uh you have to care you have to invest in it's why we do the polishing and we spend time on it uh and i think over time yeah you're you're right that you want to kind of balance that out but ultimately you just want general capabilities that everything works equally in every in every harness

Speaker B: uh just on the topic do do you guys do much with like hermes open claw nano claw whatever

Speaker A: pi pi no pi is a different pie

Speaker B: it's more coding

Speaker A: i'm a big fan of pi though i have to say i think it's a really pi pi you sound closest to pi in terms of pool and pi in terms of like the minimal surface

Speaker B: the minimal it's because i don't i have a allow me for one more strong opinion

Speaker A: i've been saying this now for two years i think mcp and tools are stupid

Speaker B: oo that's cool now you support mcp

Speaker A: i support mcp and we support tools and everything they make absolutely no sense to me uh and like and then i'll explain a little bit why and then i think i can probably get people to come along on this one if you are looking for complex tasks increasingly longer horizon increasingly complex task doesn't matter if it's coding or something else uh you are going to be interacting with data sources right and you're going to be interacting with things that are installed on some form of a virtual machine uh and what we are doing is that we're putting a layer in between those things we're putting like mcp in between we're putting tool calls in between and this is even more about tool calls than mcp where the model can just write the code and interact with the system and we're starting to see that like laguna s does this a lot you'll see this as well in like frontier models they're increasingly no longer here we're going to stuff 50 tools in the like system prompt to no here's a virtual machine with these binaries installed this code base you can operate in here a folder where you can write you know your memory if you want and the model is using code to do complex tasks and when it uses code it is not one or two tool calls or three things that are chained together it actually starts you know using if statements and for loops and and making things conditional and so i actually think we're moving from we already are moving from tool calls you know to effectively models writing code little scripts and you see this a lot when you get the python you know

Speaker B: exactly like in just the arrow arrow you know written code and the file i don't know what you call the eof yeah exactly uh you already see this happening more in models because when you start training them in rl the models want to be free they want to be able to do the thing they want to do in the most efficient possible way and it is not calling one of the 50 tools in there like system prompt and so i'm a very big fan of give the model a minimal harness as minimal as possible give it a container in which it has its own codebase right they got a model's codebase that has access to the api keys and data sources and and little libraries and and documentation that it needs and just let it run free at the task um and i think that is the way we're going i think we will in 12 months not see a single system prompt that is stuffed with 20 or 30 or 40 tools anymore

Speaker A: no comment no no push back there i think uh i think there will be it'll be supported for a long time just because that's a lot of people are trained on that now but maybe you guys don't have to support it in your models going forward so uh but yeah i mean if you can i do think that's writing code is more generalist and it's a it's a means to an end

Speaker B: and we do support tools and we support and this is actually the first model we're doing parallel tool calling in which we needed to catch up on so like that's there and like so it's it's there but i it's a personal nitpick like i want the models to have as many degrees of freedom and just like you know be free and and do capable things

Speaker A: yeah so and then so that was on the path towards like okay how do you use poolides models and laguna models for my hermes or my open cloud all those things and so typically what i look for is um computer use or vision that's a that's a very big one you guys have a blog post on that but then also the persistence i think is very strong value as well as long context which you guys have a million token context anything else

Speaker B: so for us look so for us vision understanding is is the next thing right we don't have vision understanding in these models yet

Speaker A: i was going to is the we don't have vision understanding in these models yet uh and so this is something that we've we've started efforts on like we think it's it's super important to have visual understanding that's company vision uh and so no we've got work to do there and and this actually one of the things i loved about the thinking model like from the two minutes i scrolled the blog post very committed to multimodal including audio

Speaker B: yeah they're state-of-the-art audio as much as it's a trillion parameter state-of-the-art audio but also all trained from scratch right no encoder in the sense to me that's that's that's one of the strongest reasons why you need to trade from scratch is you would just have a different tokenizer you'd have different i'm fully aligned like zero dis zero disagreement from me here like uh just add the modality and and and don't put you know keep it keep it simple uh we're i don't think we'll touch audio for a very long time it's kind of in the name too you know ink link ink

Speaker A: it's kind of in the name too you know ink link ink true yeah yeah why so what's so hard about audio it's not about what's again it all comes down to focus i see right like saying no to things means that there's a researcher and compute that can go to making general progress and our view is like general progress is going to come from the ability to push these models to far more capable reasoning far more longer horizon tasks uh i don't think audio adds to that i don't think it it pushes us close to AGI i think it is a necessary modality as you get close to AGI i think visual understanding sits in the middle of those things i think visual understanding can absolutely do so but it also unlocks capabilities that are just valuable today so but this is the point right you want more diversity you want more different foundation model companies to focus on different things i think we are just kind of like a horse with blinders on just like yeah you have your path we have our path we want to catch up to the frontier and uh we don't want to distract ourselves with anything else

Speaker A: Yeah. Uh I will call out that one of the branches of research is deeply OCR which is can you just throw away the text tokenizer and just only vision.

Speaker B: I find this I look geek the geek in me is like looks at this stuff and it's like okay look like look at the number of bits in code.

Speaker B: which I'll say, you know, you can see somewhat works looking at anthropic. OpenAI has a lot of vision, multimodality. Um, Anthropic kind of just didn't, right? Fable's a big step up in image processing, but like they're not known as the multimodal company, right? the language model coding company that has multimodal capabilities that's never super flex and you know goes pretty far.

Speaker A: I look I in this I think I I think anthropic I mean they've done many things right but I think this maniacal focus on just pushing capabilities scaling up models is I couldn't agree more. Uh I think it's it's that's the first hurdle and once we get that then we can improve a whole bunch of other things. uh and but at the same time on the other end of the spectrum it's really exciting to see people you know building these spatial models right that are uh in the world models that are being built like for very different you know use cases uh but I think ultimately it all comes together at some point.

Speaker B: okay so scaling models this is Laguna S for small you have good naming extra small small medium large

Speaker A: um still scaling so the new medium started training and it's much bigger than the last medium M started training yesterday. Uh so it's a uh 39 day pre-training run. Uh and uh how do you know the days in advance? Just the computer >> model factory but >> right and and like at this point like with the model factory like it's >> I thought it was interesting. So in in the Laguna medium and extra small you even quoted number of GPU hours for how many days and whatever for different size and I'm like oh you can also work backwards to how much that costs right? What GPUs how many hours

Speaker B: >> and you realize it's not a lot. No, it's not.

Speaker A: Uh, and you know, you started with Deepseek of the West and and uh I think that's uh the Deepseek moment, right, was a moment when people realize you can train incredibly capable models for not a lot of money on the training run. But I think that's the falsehood, right? Like the training run is not the expensive part. The training run is a very anticlimactic event, right? Like we just had a Slack message come up yesterday to say the new model is training and here are the links you know so you can can follow along the evals and like that's it uh all the work that goes into that moment it's kind of like how people talk I know nothing about sports but how like athletes talk about like you know it's all the preparation it's all the going to the gym and then the game is just a game I think that's a little bit like with model

Speaker B: yeah people had overindexed on deepsek was trained for $5 million or whatever it was right it's like there's the amount of R&D before that the infrastructure is built >> up older things the data but no so Laguna M is training and yes there will be an L and there will be Excel and and what you'll see that what you'll see with M right M is much larger than the last M right so these monikers are a little bit our version of the

Speaker A: >> he was making fun of people for saying small is 24B or something >> no small small for Mistral now is over 100B

Speaker B: >> what >> yeah I can pull it off >> I mean our small right is 118 so I don't want to say anything else like

Speaker A: >> I mean I think it's also okay you're small as >> we all know that the single hardest thing for any foundation model company is naming I don't want to say that we're good at it either. I mean, it's this is Laguna, you know, Laguna S 2.1. It's it's >> But at least people understand, you know, medium is bigger than small. Until you mess that up, like you have a password.

Speaker B: we try we try hard. Uh while we're on the topic of naming, this is going to be at the end, but might as well. Uh why poolside? Why Auna? So when we started the company, it was going to be called Snowball Apps. Uh it was after the snowball effect because we expected this company to become a snowball effect and it definitely has been a snowball effect for us.

Speaker A: >> turns out it's an Amazon trademark. I kid you not that my co-founder's next suggestion of a name was let's call it Bedrock.

Speaker B: and so at this point I was like okay no you are amazing at naming things if you were at Amazon. Uh and so uh early on in the company uh before we were incorporated uh we were at an annual conference of a very big major tech company and we had been discussing with them and you have to realize the company at this point is me my co-founder CEO Margarita we know the first person is going to join us we haven't like incorporated yet uh and we were discussing an open AI Microsoft style deal with this big tech company like they were going to provide us with a lot of compute. We would give him, you know, perpetual access, a whole bunch of kind of things. And uh we found out the name was trademarked Snow Labs while we were at that conference and having this discussion that we had no right to have, right? We were a couple of guys who had nothing yet, but this big company was willing to entertain the fact that we might partner with them. And uh we were discussing this and it was in their annual conference in a public setting and the chief scientist of that company said people can hear us here like we should move somewhere else. Let's go to the restaurant poolside. And for some reason me and Jason looked at each other in that moment and said oh uh and then later that night we the name stuck with us and we said let's call the company Poolside. And ever since we never ended up doing that deal and we used it as a reminder to never turn down our round down our ambitions because that would have been the easy path. Uh and the hard path was what we did which is start and try to raise exorbitant amounts of money when you're just a couple of guys who are

not even building it in Silicon Valley who don't come from any you know the known labs and things like this. And so everyone assumes poolside because AGI everyone sits poolside and it was a playful name and we liked it and it was a little bit different but actually the name is like a reminder for us to never round down our ambitions. Uh and whenever you're faced with those decisions to just pick the harder path.

Yeah. Uh I mean that's a great story. I I know you told it before, but I just wanted to on the record. Uh but that's that's what I did the first time I met you. You told me you sat me down. You were we were in the hotel somewhere and you were like we're raising 500 million. I'm like and then you gave me the whole vision and then you actually did it and I was like uh well you know I don't have that much opportunities to to to ask like just how do you do that kind of raise to that kind of to those kinds of VCs what are they looking for you know like yes vaguely AGI but like what do they want when these look it's the world's definitely changed right when we were raising that $500 million round the majority of investor conversations were still trying to explain that these models were not just stoastic parrots and that they were going to keep going. Uh I've seen the world go from OpenAI is going to win it all and there's no one else who can build a company, right? I mean, Anthropic struggled, you know, to raise their $500 million round. It's kind of like well reported. They pulled it off gladly. Uh and so I think when we raised that, it was about a year and a half ago at this point. The world was very different than it is today. I think the world today there's been there's been this function where the number of people who believe AGI is real uh is probably a super linear or definitely some form of an exponential function itself. M and I think this is important because if you hold the belief that we had three years ago and a year and a half ago and we looked for people who shared that belief which is like this technology is going to fundamentally underpin everything that's economically interest or economically valuable and scientifically interesting for like the future. Then the value function afterwards is easy to understand which is okay if you get there you are one of the commodity one of the players who can build this commodity. uh and over the years building that commodity has become not just about building models but also about building infrastructure and other things and so I think today because the number of people is bigger and the outcomes have been proven right I think the incredible like financial success that Anthropic is having right now and like the growth that OpenAI's had and others and Google no longer make this a question of is their product market fit uh which really a couple of years ago was like part of the question like how big can these things tell people that like you know you be up these amount of revenue numbers in our industry right now people were still like would laugh you out the room.

uh now I think it's a function of who in the world believes that it's going to be an igopoly of intelligence and who believes that igopoly can be broken by other companies and I think that's what divides investors more than anything else for the ones who believe in AI uh and then you've got a whole layer that you know is kind of self-selecting out foundation model companies because they're like look I I can't make a the money I put there you know compared to what I can put in application companies very different I think there's incredible application companies and there should be many should be built but I do think we are still in a world right now where this is the early innings of this can still be the early innings of who is going to you know be part of the set of people who win this is intelligence is the most in my view going to be the world's most demanded commodity. It will more commoditize in margin and price. Uh and the world wants choice and wants options. And so I think treating the world as like oh there's only going to be two players I think is very shortsighted from investors. I think that group who thought that was a lot bigger at the beginning of the year than now.

Mhm. I think the last couple of months have woken up a lot of people and going holy like the the world both can use a lot more intelligence but also like the world is far more complex. we should have multiple choices, more options, things that can be turned off that can't be that the restrictions that people put on models now I think is is another area of this, right? Like the fact that we are entering into a world where model companies are saying you're not allowed to use me for foundation model company development. They should be allowed to do this. It's capitalism. It's their business. It's their work product. Uh but it is insane. It is it is wild that we are like okay with that.

Do do you have more problem with Anthropic saying it or the White House saying it? You know that you're picking two two different uh you know limitations and restrictions there.

Look I think I I I'll put it this way. I think we we want to as this technology gets more capable for the better and worse we do want to yield to democracy to figure this out more and more. I think any single company making unilateral decisions uh is is dangerous. It's a concentration of power in a small number of people with very limited checks and balances. Uh and that has never worked out well in history uh in any way, shape or form. Uh and this is not a criticism on the existing foundation model companies. This is just more commentary on like how I'd like the world to be. Uh I think in a world where the technology gets more capable, government needs to play an active role in determining you know where where is there real risks of misuse right and I do think we need to separate safety between misuse uh and you know doomsday scenarios that you know I think are no one knows if are going to happen or not. And I think just like very practically I think uh I'm glad to see there's a lot of conversation now starting to happen again at the government level of trying to figure this out. Uh and and now what the final decisions are maybe I'm happy about them, maybe I don't. Maybe I agree, maybe not. But ultimately like that's kind of democracy always, right? Like at any given moment I might not be perfectly happy with one or the other, but people chose to vote in someone to make those decisions. And so I think over the long run over you know over a 20 year time span the world kind of directionally goes correct and and democracy does work at least what's the famous quote of like it's the you know the worst of all it's the best of all the worst systems or or something like

it's the worst form of uh organization except for all the others that we've tried.

I love that. Uh and so that's what I hope for now. I do think we are in a critical moment of time and so speaking up for anyone is important. I think you know researchers who are thinking about starting their own foundation model companies start you know people who want to share their opinion and be vocal if that's with their representatives or just out on X like do so. Uh and but concretely to your point uh I think we are not at a level of capability right now that we should start restricting you know open models in any way shape or form. I think it will hurt innovation if we do so.

Is there a point at which you will change your opinion there?

Yes. I mean look and there has to be. Yeah. Right. Like you cannot if you sit with a straight face and say this can be open forever in every way, shape or form. Uh it is just as I think agreeous as saying you know the opposite of it all needs to be closed down right now. Like I think at any ends of extremes of spectrums is where we go wrong.

right and in in in society in any way, shape or form. And so the answer is always more nuanced and the answer is never black and white.

I think as we encounter like real world scenarios where we have to say hey we have to be more careful we need to reevaluate if that means training a model differently and opening it up having different versions some things that you know that are restrict totally okay because I don't think anyone should be irresponsible. What I do want to call out is that people have been calling for the fear of misuse of these models since GPT2, right? And I still remember like we cannot release GBT2 because the whole world will get misario.

And so like this is not a commentary on Dario, it's a commentary just in general in the space. And so we have we have not been very good at this so far and we need to get better at it. And I do think that the work that's happening with like safety institutes and better eval is probably the right direction.

Yeah. I mean I I want to say something in defense of this. It's better to air on the side of safety and then roll it back rather than the other way because the other way it's a one one way decision.

I think I think that's I think that's true.

the caveat there is also the competition, right? You you don't have global error on the side of safety, right? You're talking

Yeah, exactly. So, you don't get to do uni unilateral safety because someone else will just be more unsafe than you.

Yeah, exactly. You can pause innovation here. It doesn't mean it's it's pausing.

Take over. So, they're complex, right?

And and I think we are much better off talking about certain capabilities that we can you know commonly agree on and internationally agree on that we want to you know limit or not have available then we should talk about it in black and white of models available yes or no like the moment you start getting these big blanket statements it's that's when you start getting at the risk of like I always think back about when we banned advertising on cigarettes. good thing. I'm not saying I'm against back, but it effectively established an igopoly of cigarette companies because no one else could ever compete. Uh, and it was the probably the best moment to to the tobacco industry that that ever happened. And we don't want to do that right now.

If we if we pull up, you know, walls behind innovation, and this is a self-serving comment because I'm not at the frontier yet, but it's not just related to me. I think it's related to everyone in the space. uh you are deciding right now in 2026 based on the current capabilities of models that this is something that only two or three companies can build and that to me reads like chapter 14 of the most dystopian sci-fi novel that I could read because from there I think you can play out all the scenarios that happen in the world and none of those are the ones that make me you know excited about the future uh and I think that's the thing we should all think about like what's the future we want to be excited about what do we want to have and I think that's a future where intelligence is a commodity, everyone can access it. It becomes cheaper and cheaper, right? And I think that's important. It can like impact more of the world and it's not one where, you know, a single company puts their thumb on their scale of what it outputs uh to or turns it on or off. I think the one entity that has more power than the US government here is Nvidia because like basically whoever gets the allocations gets the compute. You can take it down to TSMC or

TSMC below that. But I I just want to test provocative statements to see if you you have any response.

I need to think on that one >> which actually I think they are regulated, right? Like you can see the government

Can they ship to China?

Okay. But you know they're not China.

Look, I think this industry has existed because of what Nvidia has done.

Yeah. >> Right. I know that we people like it's easy to give him flack, but I also want to say like I remember when we started sourced right in in and in in 2015 post that Kapathy article it was able for this progress to happen because we're able to put consumer GPUs in servers and they allowed us to do so and then like and you kept going further and so this is something like foundation models are so closely linked to their hardware and their systems. Why do we see these step-wise progress happening? And we see them happening because of the next generation of networking and and systems that come out, right? The difference of a model you could train on hoppers versus GB300's is the difference between a trillion parameter model and a five or six trillion parameter model. And so these things really coexist I think very closely to each other. And I think the more interesting question I think for for the future is going to become of like how do we what can we unlock in terms of model capabilities like as we start co-designing these things even more and we're seeing that with like the next generation of systems and I think the world the world you know abhores like capitalism does a really good job at trying to like you know push towards things that are that allow for more competition, right? And and Nvidia allows for competition. It's not, you know, but if a government says no one else can build foundation models effectively through the regulation, that is very different. Is it hard to go build an Nvidia? Absolutely. Is it hard to build a foundation model? I think it's very hard to build a foundation model. But we should like make the playing field one that where you know if someone wakes up tomorrow and wants to do so they're like allowed to do so they're allowed to use the tools to do so. And I think there's still a big difference between what we're seeing in the discussions about model companies versus what we're seeing with chip companies.

The gap also seems to be the expertise in who regulates it, right? Who at the government decides what's too safe, too smart, too dangerous. Um but while we're throwing spicy questions out there um do you have anything that comes to top of mind that could be changed? So you know should open AI anthropic open source models is it open weights? Is it what we do in RL that determines you know your safety barriers? Is there anything that should be done there?

um yes. Um one of the things that I'm excited about that I think we're more and more talking about I don't think anyone is doing yet is uh mix and match of hardware during RL training, right? Like the you think about like the notion and we're seeing this in inference, right? The prefill and decode. Yeah. uh just work better with, you know, a general purpose uh a GPU and and a more specialized like chip, right? Like the Grock chip at Nvidia, the LPU and the GPU combined and there's different versions of that in the industry and and RL is batch size constraint, right? So like you are ultimately and you're batch size constraint because you don't have infinite tasks, right? When you've got the entire web, you can be much more flexible in scaling up your batch size because you've got the entire web. But for RL, you have, you know, x millions of tasks that you are going to be training on. And so you cannot blow up your batch size massively, which means that you actually can't scale compute to a certain extent with RL the same way you could scale compute with like pre-training. Uh, and so I'm very excited about anything that improves that. And I think one of the best ways to start improving that is the things that we're already starting to see in inference, which is the separation of the the prefill and decode to different chips to come to reinforcement learning, right? Uh and I think we'll be there soon. Um and I think more people should be working on this. Uh because then all of a sudden we're able to just be way more efficient in how we train our LLM from a wall clock time. Again, coming back down to the fact that it's a race, right? the race is measured not in how many GPUs, but the race is measured on on calendar time and that's probably one of the biggest impacts we can have right now to speed up our industry. Uh and so that's one like technically I love geeking out about and talking to people.

Speaker A: I had a tour of their data center and um physically you can see how PD disaggregation is mapped out in the in the data center and you have to own your own hardware to do that.

Speaker B: Yeah. No, look, I think it's I think more innovation in space is just like is the coolest thing. Uh and uh and so I'm I'm excited because that's frankly like all of us are like why why did we finish you know post-training this model whatever two weeks before release or no sorry between release between pre-training then you know mid-training SFT and then the time it takes for release. My biggest wall clock bottleneck right now is RL time, right?

Speaker A: And it's just because I can't scale it up further because I can't add more GPUs to it because of that bad size constraint. There's a really cool uh blog post that just came out that was showing uh RL done in even lower precision than any of us are doing. I thought this was really cool. So this what date is it today? We're on July 15. So this came out 5 days ago. Uh and I thought this was very cool. uh I think you know lower precision RL while keeping it stable we're we're still doing this in FP8 and so uh I was excited to see them sharing this work and and bringing it out uh it's definitely something that I'm excited to be doing once we move to uh to Blackwell GPUs

Speaker B: but yeah cool part of open research you know you you take and you give

Speaker A: exactly yeah I'll just quickly mention there was a paper that did a basion on uh levels of quantization and they roughly concluded that 4bit was the sweet spot.

Speaker B: But I don't remember this was a couple of years ago, right? I think I remember this year

Speaker A: one year came out. But like I'm like, okay, maybe NV FP4 is it. You can't really like the the lowest you can go is turnary. That's it. Like there's not that many.

Speaker B: Well, I mean there's there's still quite a difference between NVFP4 and 4bit, right? In terms of what's what's possible, but I think NVFP4 is, you know, underrated in terms of of what it is. I'm I'm quite excited that when it came out uh it's uh you know just getting that extra like that trade-off between uh between range u yeah

Speaker A: is very cool.

Speaker B: Uh couple closing questions. I have a quick question.

Speaker A: Okay quick question back to technical side. So any big takeaways from XS 2.1 medium to training the new small just general in terms of training models. You mentioned a lot in the earlier discussion about okay in pre-training there's a lot you can squeeze out right you can learn a lot more from the web um at the same time you took 30b and scaled it up to 120B right um is there any gating on how small is too small so I'm I'm just going to ramble for a bit I'll come to a question at the end but you know part of Karpathy's thesis was cognitive core right we've seen vibe thinker nan beads 3b 4bs that reason a lot and then you know the idea is you offload to a different model for the work this these are small reasoning models so have you found anything interesting in model sizes like 20 30bs on device 100bs on single GPU um can you squeeze out more there's a lot more to squeeze out uh like I think not to make too many forward promises but I think we can squeeze a lot more out of the excess size as well uh and I think we learned learned a lot during s training that will allow us to improve excess like size even further and I think already since then we have learned things that could have made s even better uh I think there is a lot more still for like our our space to squeeze out of models much smaller u I don't think that's an argument against scaling it's just an uh and one by the way I think this is a nice thing that you know it's really it's not very helpful to have a post-training recipe for a smaller model and try to apply it to a bigger model

Speaker B: but um recipe for post training for a bigger model applied to a smaller model is almost always just a really good like improvement and and baseline. You can still tweak it more. Uh but I don't think that's necessarily like obvious. Uh and uh and so you once you make your bigger models better, you often have a a quick lever to quickly improve your smaller models again. Uh but will we be able to squeeze a lot more out of smaller models? Laguna S gave me a lot of confidence that I think we can. Uh and I think it's around that discussion we had earlier about that it's about the behaviors, not necessarily the raw intelligence uh that you're trying to improve the models for. And that's on all axes of uh there's like an axis of how long a model will reason. So how long can it stayic and there's also efficiency, right? You want to ideally push on both. And the the thing to clarify you guys aren't doing right now which we do see at Frontier Labs is the distillation, right? You have a big big model that you don't really ship to users and what you put out for inference is typically distilled from that which gets you quite a bit of gains, right? Look, I think it's it's something we don't do right now because of like why we're also like building these models, right? These models are for us part of our research path. So, we've you know, Laguna Medium was much larger than the last two models that this one and last one that we've released. Uh and we've trained even bigger models in the past. So, there is the engineering component of like a bigger model and at every kind of order of magnitude size, you'll learn new things in pre-training about stability. But at smaller model sizes you are able to just iterate a lot quicker like internally right on your research. And so uh for us distilling down to a smaller model doesn't actually serve the purpose. These models are kind of it's not the right term but to us they're dualpurpose models. They are progress for us to wait to see did we improve in the model factory and something to put out into the world. Uh and so that's why we don't do it. We've done distillation experiments and there's like really cool things you can do and I think if you have lots of user data uh then uh you can go even further right uh in that but I think there's something to be said in having a a quick cadence of models trained end to end from scratch so that you as a research organization can learn the lessons and not wait. That was actually one of the big lessons we learned over the years when we used to have a much longer cadence between model trainings like six months and we would train just like a big big model, wait six months train another bigger model uh you would be compounding so many changes of improvements that at the by the time you're training your next model it's a bit of a soup and you don't really know what ingredients led to the outcomes. So when you are training far more frequently models uh and this holds true for both post-training uh and from pre-training from scratch you are much more able to get an understanding of what led to the improvements uh and I think that's important like ultimately we are all still there is no true science yet of you know deep learning for large language models uh but we are all I think you know trying to gain insights from our experiments because it's those insights that lead to scaling laws that lead to the kind of improvements that allow us to be more compute efficient and get more capabilities.

Speaker A: Yeah. Um, amazing. I was going to end off with a little bit more history. Uh, you spent some time looking at uh metrics for engineering team productivity. How do you think about engineering team productivity today?

Speaker B: I mean, it's wild, right? I mean, it's the it's like the golden age. Like, it's the fact that you can just take an idea and build something by waiting overnight for an agent to do the work. I don't know. to me

Speaker A: like how do you measure on you know because literally in a you look I think it's a good question it's one I haven't thought about in a long time uh but you know you're you're pretty qualified to do it I'm going to no it's a fair point let me take a second to think about it look ultimately what is code what is

software engineering is to go from something that is valuable for an end user or sets of end users uh like an idea an extra bug fix a feature to like delivering that value. And I think what we're doing with these models becoming more capable is that we are massively like both cutting out middlemen and compressing the time that it takes to deliver that value. And ultimately that iteration cycle for any startup or any company is what allows you to win, right? if you're able to solve a bug in two hours versus in staying in the backwalk for three weeks. If you're able to like be on a customer call and learn, hey, if this feature existed, it would like, you know, they'll be willing to pay more and it's more valuable to them and you ship it in a week instead of in a month. And so I think ultimately maybe the same things that we looked at years ago preLM still apply. And it's just the notion of cycle time, but in this case, it's the lead time from the moment you have a valuable thing that you're looking to do for someone to the moment that it's actually shipped to them. Every other metric is ultimately a leading indicator for that lagging indicator, right? It doesn't matter if you're looking at amounts of code, PR, reviews, all of these kind of things. And so I think in this case we are starting to move so quickly in some of these things that we can just sit back and look at what was traditionally the lagging indicator which is the lead time from contrition ticket to like you know like an end result. Um what I would look at in this new world uh that maybe we didn't think about before is how much can a single person do with that. Right? One of the most like if you look at AI native companies they're not designed like the engineering orgs of you know pedalm age. They're actually designed with often just the builder, right? Uh and as close to the kind of customer to the ability to ship. Uh there isn't necessarily a huge team in between that sits there. And I think that is I think is exciting like organizations where a single IC can just you know get much closer to that. So I would look at from where the value sits that's identified to the moment it's shipped and how many people are involved in that and you want the amount of people involved in that to be less and you want the time end to end to be shorter.

Okay. Um, is there a way to eval when you're interviewing somebody?

O uh because that is you know I look the most compressed version I think the the common answer to this is agency.

Yeah. How much agency does a person have? Uh, I think in the age of AI getting more capable, agency becomes probably one of the most important qualities for anyone. And I think agency is something you can look for in you know what people have done in the past because agency is something that if you have it, you are demonstrating it right. No one has just agency and is sitting back and not like exercising it. The whole definition of it is that it's exercised. And so understanding like what were things that people did in their lives in their professional and their personal project that showed agency and your personal backstory shows a ridiculous amount of agency right like I think that is ultimately it it's the you know the Silicon Valley you know quote of the last you know year and a half or so it's like you can just do things right that's I think what you're looking for.

I think then aligning high agency people is very hard because they all want to go their own way that's the whole point right? Yeah, but I think the notion like I think the notion of a good leader, right, in an organization is to be able to bring people together around like a common outcome. And I think what you want to do with anyone who's high agency, I feel very lucky I've got an organization with incredibly high agency people. Like I mean I'm not the one who built the model, right? I cannot stress this enough. Like it's the team that like achieved this and it's a team that is incredibly high agency. And so if you look at like what does it take to bring that together it's it's ultimately a common goal and a common set of boundaries because if you allow to just go you can do everything you become an exploration algorithm and this is what we see in big tech right in research and big tech everything is an exploration algorithm everyone can do anything as long as and then becomes political about gathering the resources so when you say this is our common goal and these are the boundaries that we've set right we're not multimmodal we focus on RL like we do these things and you're upfront with people before they join the company, you know, like you get a lot of agency. You can run where you want, but these are these are the places where we say this doesn't this are the lanes that make sense.

I think it actually gets the best out of people because like innovation comes from constraints. We did this with relatively little computer and relatively little money compared to some of like you know the others that are out there. uh and I've thought back on that quite a bit recently and thought actually it was a good thing because those constraints kind of forced us to become much better in certain other axis that might others might have not right we uh we purchased relatively little external data uh I was going to ask about that

exactly right that was a constraint but it's a constraint that pushed us to move on other areas to improve and like and there's lots of versions of that so I think high agency people you want to empower you want to get them really excited what they're doing but you also want to say hey if you join this mission this is the outcome I need you to achieve but these are the places that we don't go and maybe if you care about those places go somewhere else.

yeah, great um call to action who are you hiring?

we are hiring on every possible role in applied research and engineering in the company uh so from pre-training all the way to eval to post-training architecture like we are still in a world where you know individuals can have massive impact and I think our pitch to join us it's we spoke a lot about the mission how we think about things but I think we are one of the places where it's the highest ratio to individual to impact right less than 70 people built this model less than 150 to engineering and researchers like together did this effort and that's a very broad definition because I put myself in the 115 list and so being able to do this kind of work on a mission that you're aligned with uh and you can have that in every individual still has a huge impact.

and being able to publish being able to open source the model. Yeah, look at all of those things are are part of that, but I think ultimately uh when you you can today pick between joining a very large foundation model company and but you are one of many uh and not by any fault of them but just by definition the denominator has become really big and our denominator is quite small and so the level of impact you get to have is really high and I think ultimately all of us frankly the most incredible high agency people I know what are they optimizing for they're optimizing for impact uh they're optimizing for impact and am I aligned with the mission and and if today you heard about the mission and aligned and you're optimizing for impact I think we're a really good place to try.

okay, I think we ended there was fantastic uh statement you did amazing on 4 hours of sleep uh you guys so you know podcast eval definitely appreciated I literally like my eyes are like starting to go like this I'm like let you go let you go back to thank you for setting this Uh we wanted to get this in because we think it's a great model and I think it's a great story to tell.

Thank you.