Aaron Levie on Why Open AI Wins

0:00Open Weight AI is often for him because of threat to frontier labs. Aaron Levy thinks that gets the economics backwards. The box co-founder and CEO joins Theo Jaffe and Sophia Puccini on MTS to discuss why open models could make the AI ecosystem more competitive, the debate around distillation in China, and why America needs more open weight AI. They also get into the latest frontier models, how AI is expanding rather than shrinking boxes engineering roadmap, and why model routing could become the default for enterprise AI. We are live with Aaron Levy, the co-founder and CEO of Box, which does all kinds of things cloud content management, enterprise documents, permissions, collaboration, a lot of different AI functions. He has an incredible Twitter account at Levy. He's been around on the website forever, and he writes about AI and many other topics. We have two huge stories today, I wonder which should we start with?

1:02Astrology. Astrology. Of course, astrology. Yeah. Huge acquisition in the astrology image, Gen Space, something we want to do. Is it an open source play or not yet? I think not yet. I think astrology might remain closed source for the time being. Okay, that's about it. We have to fight that. We have to fight that. We have to fight that. That's a good story. Today there was an open weights letter that Jensen Huang wrote and Box signed. Can you tell us your interpretation of this letter? What exactly is it aimed at? What are you guys trying to shape? What do you think Jensen was trying to shape? Yeah. Is the letter maybe a Rorschach test of what do you see in this letter? What we saw when we read it was, hopefully, it's the default stance of the folks that signed it. Basically, almost two full main points. The reason why open weights AI is super important is because it actually drives AI progress. We get more innovation. We get more options. You can build on top of these models.

2:03You can train them for your own use cases. Open weights in general, I would argue, is actually a very important part of the AI ecosystem. So much so that I think it's actually kind of mis-framed as zero sum with closed weights. It actually just adds to the number of use cases that people then do with AI. That's kind of part one. Part two, I think embedded in there was a little bit of a call to arms on the US actually needs to be probably even more invested in open weights models. We probably want even more companies showing up to the table with this innovation. It didn't seem like it was directly about, we must support all of the things that are happening in China as much as the US needs to continue to support open weights. We probably need even more innovation here. I know, by the way, distillation is actually not always bad and there's lots of use cases around it. Let's maybe calm down on some of the foot on that. We support every one of those stances. Open models are, I think, very important to the AI ecosystem and actually think it pushes even the closed model providers to innovate faster and be able to deliver more for the market. I think everybody is a winner as a result of open weights.

3:03To be fair to the other side of the argument, I think there's a category, forgetting commercial and economic interest for a second of some of the close players. I think there's a category that obviously has real arguments around the safety elements of open weights. At some point in the future of capability, I don't agree with that point of view, but I think it's a healthy debate to have and to have conversations around it. I'm almost unequivocal to take the stance that more open innovation and open weights innovation is good for AI broadly and the diffusion of AI. Diving into the distillation point, how would you separate distillation that is just normal, basically, normal economic activity from distillation that crosses a line into some sort of civil or criminal violation? I don't know if I would make that argument. I don't know where that line would exist. I'm almost deeply on the side of, I think it's very hard to make the argument that AI models should be trained on broadly the public internet, but another AI model can't be trained on the outputs of an AI model. I think it's a very tenuous argument to make. I don't see how you can pull it off.

4:03At the exact same time, I understand, and if I were running a lab, I would want to like block these things down as much as humanly possible. So I'd be trying to find every defensive mechanism to prevent the distillation of my models, but I don't know that you'd be able to kind of credibly argue that there's some ethical kind of line that has crossed. Simply because then you would be basically arguing against the underlying training runs of these models. Most people didn't get to opt to inter-out of their data being trained on. Maybe it's in some kind of terms of service from the underlying provider somewhere in the fine print, but this is a thing that is sort of like, we are assuming that the general knowledge of the world is going to be trained into these models and that general knowledge, whether it comes from Reddit or it comes from Anthropic, is like, I don't see the distinction between those in any meaningful way. Yeah. I mean, we heard it take yesterday that was kind of exactly that. It was just like, we're focusing on the wrong part of the chain when we blame distillation. It should really be like, how is Anthropic thinking about like, how to restrict their API use or like how to very effectively like prevent the sort of, because we were also talking about how like distillation, they still have to pay for the API usage.

5:08Anthropic is getting paid more for distillation than most people have ever gotten paid for the original training runs. And if you kind of like, you almost kind of do a thought experiment of like, well, what's the difference between distillation in a very direct like I'm using the API across 10,000 accounts versus just quite literally, if we all generate code from Fable or GPT 5.6 and all that code ends up in public repos, what's the kind of compelling difference between those two scenarios that a model will get trained against? And I'm not sure I at least understand why there'd be arguments, why there'd be some kind of ethical line between those two. And then you kind of layer in another element, which is I think there's a lot of people that argue that actually China in this case are doing kind of breakthrough innovation in general independent of distillation that's causing models to progress. But also I'm sure that there's going to be other kind of controversies and conspiracies about what's happening. And I don't know enough to be able to kind of credibly argue that other things are or are not happening in why these models are progressing. I'll let obviously the labs to make those cases and the government to kind of pursue that as they see fit. But all of that to me is independent of do we want open weights models and do we want

6:11this innovation in general from the ecosystem? I think the best case scenario personally would be you get a few people that are at a stage in life where they can just be like, yep, I'm going to drop $10 billion on building an open lab in America. And we have our own version of, you know, moonshot. Like I think that would be a great service to America and to innovation in general. So I've been kind of waiting for the date that, you know, somebody does that. But you know, for now still waiting. Well, there's another foundation under this post. They might do it. They might do it. There's another angle here, which is, you know, most open weights models are coming from China. And some people worry that if we become too dependent, if the American startup ecosystem becomes too dependent on Chinese open weight models, and then they like shot off the open weights, they stop releasing new open weights models that this could be very bad for the American tech industry or that because these are Chinese open weight models, they undercut the margins of American companies. It's a national security concern, say some people.

7:11And I think those are great arguments, but it's like, okay, so what do you do about that? You know, some people say that the iPhone should cost $5,000 and $7,000 because we shouldn't, you know, rely on manufacturing advancements that exist in China. Like, I don't know. If you kind of play out the alternative, then all you would basically be arguing is America should have really expensive AI and the rest of the world should have very cheap AI. And then we'll see how kind of, you know, that works competitively over the next kind of five or 10 years. Yeah. Maybe one of those things, which is, you know, the cats out of the bag on that, we can't unwind that dynamic like open weights models exist. China knows how to produce them. I lean more on the gents inside of the argument, you know, everybody had this moment where they watched the gents and Dorkash podcasts. And that was like the original kind of Rorschach test of like, you know, some people pass Dorkash's like, obviously like that's the exact right position to be taking and, you know, you should be interrogating gents and then other people watched it and was like, why is Dorkash not understanding gents and point and kind of letting him kind of sort of expand on that. I was like watching it and just was like, yeah, gents is obviously right. You can disagree with it.

8:11But he's obviously right that the following will happen. If you block off China, China is like not going to give up on AI. They're not going to just decide that this is not that important of a technology category for them to plan. So if you just assume that China considers AI to be very strategically important and oh, by the way, you know, all these funny meme tweets of like, we have more Chinese AI researchers than the Chinese labs do in the US. Like, obviously they can produce incredible talent working on AI. So this is not something that we own, you know, some impenetrable, you know, talent, you know, based to be able to compete against. So it's incredibly important strategically for China. They actually have a lot of the core raw materials needed between, you know, raw human talent and kind of industrial might for building chips and fabs and what not over time. You know, like the getting the training data, you could just, you know, go higher 100,000 people in China to go generate an insane amount of data. Like there's like ways of getting the data.

9:12So at some point, you can, it's not that complex to think through China becoming a very, very strong, you know, player in AI. And if they're a very strong player in AI and the world adopts their models and adopts their architecture, that obviously is going to be bad for the US economically over the long run. So the point is instead of sort of forcing that and catalyzing that to happen even faster by the constraints that you impose, maybe we should actually be a part of the infrastructure buildout that that's going on. And I, I actually generally kind of land on that because what's going to happen is the alternative is that these models are still going to get trained one way or another. But now they're going to be, you know, trained on a different hardware stack. Ultimately over time, that hardware stack can be the one that gets deployed in sovereign clouds and whatnot. And so I think it's, I just don't think there's a scenario here where you can kind of close off completely to China and somehow you dramatically slow them down. And then kind of hand wavy, we just win. I just don't think that that plays out in this space.

10:15So, you know, all of the, you know, kind of a lot of the sort of theoretical risks of either open weights or, or, or, or, you know, the infrastructure side have to, have to assume that we have such an insurmountable lead against China and that only accelerates past the point where like there's just simply no catching up. And I think as we've seen recently with, you know, K3 and, and other models, you know, that gap is maybe narrowing as opposed to expanding over time. Yeah. That door cash, Jensen interview by the generational, they had such Jensen it like when Jensen was like, I'm not a loser, you know, you're not talking to someone who woke up a loser. Yeah. We're not a car. Yeah. No, that makes no sense. Yeah. That was, that was some of the best content I ever produced. So I, I don't know why they didn't charge for that one. Yeah, it felt like tech reality TV. Like he was actually just getting into it. It was awesome. Absolutely cinema. Yeah. Well, okay. What do you think would have happened if like, let's say the US would have open sourced,

11:17like, be trying to open sourcing like a Kimi K3 level capability model. Like, because I understand the argument of like, okay, we need to deploy like American open source models that are the same if not better than the Chinese open source models. But like, what do you think are the trade offs of like completely open sourcing it to the point where like, obviously Chinese labs would be able to like have access to it? Yeah. I, you know, I first of all, even, even the open verse closed, I actually appreciate all the debate that happens on this. So I'm very passionate about the topic, but I also totally appreciate all of the arguments on the other side, even even though I probably lean the other way in some of them, but some of them do inform, in, in, in forms of my views and I, and I kind of update maybe then to be a little bit more nuanced on my end. But, but, you know, on the open source, US side, I generally think that the money maker in AI is inference. And so ultimately, the dollars are going to flow to the infrastructure stack one way or

12:21another. Yeah. I think in a world where you only had one or two labs and somehow there were these insanely kind of closed secrets about training and, and you had like real intellectual property that was protectable and patentable and nobody ever could know about. And it's like, you know, totally locked down. Then in that world, I think you could probably argue that like, you know, there's another couple layers of the stack that you could kind of close off. But in a world where we're going to have, you know, three to five players in the US that consider it sort of existential to them to have leading models, then, then you have enough of a competitive dynamic in the market where you have to expect that, that the cost of tokens converge closer and closer to the, to the cost of the infrastructure over time. Not to zero. I'm, I'm definitely not a believer that there's no margin there, but, but closer. So, you know, in the 20, 30, 40% range on top of the cost of infrastructure, which is different from 70 or 80 or 90% let's say. So if that's the case, then actually if you had an open model and you powered, let's

13:22say, the preferred infrastructure for that open model or you created the post training environment for that open model or you were the kind of considered the safer brand for deploying that open model by enterprises, yeah, I actually would argue make almost the same amount of revenue just by again, driving the inference of that model. And so if you assume that with things like, you know, I would actually make the case that even for something like an open AI, if you fast followed your frontier models with open source versions of let's say the prior generation at a more consistent, you know, pace, I actually think you would, you would be even more competitive with your frontier models because it would keep more and more of the use cases within your ecosystem and your family. And you probably could just power a lot of the inference of those open models as well. So I would just argue that these are, these are kind of flavors of ways of enabling different kinds of use cases with AI as opposed to entirely different economic structures of AI because at the end of the day, you still are like, like, you're going to be paying for large GPU clusters no matter what open source AI does not mean that somehow I'm going to be

14:25running fable on my laptop. And I'm able to sort of circumvent the need to have cloud infrastructure like the dollars are going to be just not for the flow. Yeah. Yeah. So the dollars are going to flow into something infrastructure provider. And there's no reason that open AI can't be that infrastructure provider and for topic can't be that infrastructure provider, you know, obviously meta and SpaceX now are going to be getting in that space. So I'm not convinced that open weights AI dramatically changes the economic structure of AI other than to just provide even more avenues to innovation and more avenues of use cases that begin to emerge. And that I think would actually be a good thing for the ecosystem broadly. So I'm pro, you know, US open weights. I actually think the lot of the closed labs should should consider having kind of faster follow open weights models. Yeah. You know, to be able to handle the post training use cases and some of the sovereign use cases. And I don't think it would actually be as bad as probably is sort of perceived from a market standpoint. Yeah. This is a really interesting thread. I'm curious, you know, if this would be good for the business model the labs, like

15:28why don't the labs do it? Yeah. And I think that's not because never open source anything. Google open sources, Gemma, which is like something, but it's not like the last series of Gemini. I think there's two categories that people fall in. One is you sort of, you have to have a, you know, your time horizon has to be longer to believe that you can monetize open weights because like tomorrow, if I release a closed model, I just will literally make more money. If I just keep it closed and I force everything to route through my API that on paper, I'm going to make more money. No question. But over the long arc of time, where no matter what, you have to assume that the cost of token per task, you know, gets driven down regardless, then you kind of play things out two or three stages and it's sort of, it's almost, you know, kind of a wash whether you were closed or open because again, it's really just the inference cost that then ends up mattering. That's the first part. And then the second part for Anthropics specifically, let's say, is I actually just believe that they consider this to be a major safety risk. So to them, they're not like thinking about this. I'm guessing they're not thinking about this as like an economic argument that they

16:29should be open source for market share reasons. I actually just think that they fundamentally believe, no, you actually need one entity to control the flow of the tokens to be able to do, you know, kind of prevent, you know, prompt injection and make sure that you can route to different models based on the kind of the queries people are doing. You can't do that in an open weights environment. So I would just almost consider them to likely never to open open source their models for, you know, sort of like, you know, kind of reasons that, you know, are tied to just the creation of Anthropic in the first place. Yeah. I feel like, well, with regards to like frontier labs, this is going to be a segue to Claudopus 4.5 or from Claudopus 5 being out now. But yeah, with regards to like the timing of the release of the models in frontier labs, it just feels like a timing thing now. We know they're cooking good models internally. So with that being said, what were your thoughts on Opus 5, especially from like a, you know, knowledge work perspective, because you have that perspective at box.

17:30Yeah. So we've been running e-vails on Opus 5 for the past, kind of maybe a week or a week and a half or so, a couple of weeks. And it's a fantastic model. It meaning meaningful jumps over Opus 4.8, which already was kind of best in class at the time period that it was launched. And so, you know, the way this shows up in our world, so we deal with, kind of, you know, corporate data, documents, financial, you know, documents, contracts, marketing assets, research materials across every single industry. So this can be life sciences companies, law firms, large banks. And so you can imagine that the things that you're, if you're a large enterprise, you kind of need two major aspects within AI model. You want a deep domain understanding. Like you have to understand life sciences, you have to understand law. Like that has to be packed into the model. And then you have to be extremely good at just being able to kind of work with large amounts of data, process it, deal with tools, analytical capabilities that are kind of, you know, horizontal.

18:31So you have kind of general intelligence, you know, going up the matters and then you have very specific kind of industry intelligence. And Opus, Opus 5, you know, kind of this represents an improvement on both those axes. So we tried it against, you know, a variety of different industry tests that we do. Again, kind of meaningful jumps over over four eight. I'd expect that this, you know, becomes, you know, get a very compelling model across knowledge work as, you know, GPT 5.6, you know, I think equally has. So definitely great kind of work on the, on the cloud front. And I think, you know, again, like what's exciting is, and this is why I don't think it's zero sum. The closed labs continue to stay ahead on the frontier. And I think the vast majority of the dollars in profit will still go to the closed labs in almost all scenarios that this plays out. And so I, you know, I think it's been a great month for both OpenAI and a Thropic on that front. Hmm. Are you finding fable better for any tasks than Opus? You know, like it is, it seems on vibes like very slightly better, but also twice as

19:34expensive. So Opus seems like Pareto superior there. Yeah. So I don't think we've, we've, like, so we have internal benchmarks that deal with kind of general knowledge work. And I would actually agree with you that, that probably due to the cost difference, you, you probably argue that, that, that Opus 5 is now kind of an improvement over fable because of the cost delta. On certain coding tasks that, that we had tested internally, fable completely out, outmatched for eight, like in a, in a way that you couldn't make up for by the, by the, you know, token cost difference, just like, like being able to have superior solutions to hard technical problems. Sorry. For eight. So Opus 5. Prior to having access to five, if we kind of compared, like how big of a leap five was on that, I don't have enough internal coding tests to, to kind of compare against. So, so I had to see if, like, if how much is for eight meant to be, you know, kind of superior coding model versus a lot of the benchmarks that they just came out worth, where

20:36these kind of general knowledge work, kind of a, gentle tool used type benchmarks. So, so I think it, it kind of open question. The one problem that fable had and everybody's kind of tweeted about this already is, is it will often, you know, again, kind of push you down to four eight for certain capabilities. And then that becomes a problem because you're not, you know, it's kind of hard in advance to know, like did the, did the thing kind of, you know, trigger some security warning because it was looking through permissions, you know, access control code. And so it kind of freaked out and, and, you know, isn't then giving you kind of fable level of intelligence. So I do think that in Dropix, going to have to kind of work through that. I've heard of, you know, folks in the bio space that effectively can't use fable because it just again pushes the queries down, you know, too frequently. So you're not getting the, the raw intelligence from the model. And so, you know, I think, I think we have to, like, I, I appreciated in Dropix ultimate proposal on, on how we should kind of work through these types of dynamics with the government

21:36and have a, have a kind of a multi point framework where we kind of agree on what, what are these kind of capabilities that are of different risk levels and test the models for them. But there's, there is kind of an interesting question though, which is like, well, what, who gets to decide that risk framework and, and how do we agree that the risks are actually, you know, real risk that we perceive because I would argue that if, if, if you kind of snapshot it in time fable, right now, the level of things that they, they push down and prevent you from doing, that would be untenable for the future of AI. People will, will simply not use AI if, if this is the kind of ongoing environment. And so, so like, how should we decide now? Like, like, maybe, maybe, they say these risks are very highly nuanced. And so, so that's why I, I generally lean to work like, at all, very, like way too early to be locking these models down, you know, given, given how early we are. So how is it affecting software engineering at box? Like, are you hiring more fewer people? Like how do the skills that you're looking for in software engineers change over time?

22:39How do you expect them to change in their future? Yeah, we're, I'm still very bullish on software engineering. We're, you know, the way that we've used these models is, is simply to just to do way more than we were before. The, the kinds of, and it's just like so qualitative and only, only people that are, you know, probably like in the product roadmap reviews each day can, you know, maybe kind of fully process. But like, there's something totally different when you're, when you're doing a product planning session, when you are thinking in a world of just sheer human, you know, base constraints versus, versus now what AI on locks, like it completely changes your mindset about the things that you'll go and tackle and take on. We have multiple dozens of, of projects right now that we absolutely would not be doing if AI didn't exist. We just, we would not have, we would not have lit up the projects. Like we would have said, no, that's too complex. It's not worth it. Like, you actually have two interesting scenarios. You either say no to the very small things because it's not worth it, or you say no to

23:43the really big things because they're simply too hard. And so most of your software projects kind of stay in this middle band, which is like, okay, it feels like something that maybe you could do in one to six months. Like those are the kind of things that we would use to kind of tackle pre AI. Now what happens is you can tackle the multi-year projects because they're not multi-year anymore. And the things that are kind of like, well, that would take a week, but it's not that big of an issue. So we're just never going to do it. And so it just remains on the backlog, just very far down the backlog. You do those because now that takes two hours. So what happens is you end up basically being able to solve more and more problems that your customers have always had. And that actually just increases your ambition. So wherever possible, I'm generally in the mode of actually trying to add more human talent to this because there's actually just more things we want to go and take on. And now actually the main problem is we have, you know, you just have financial constraints because there's other areas of the business that you want to grow and that you want to hire for. But I generally think if you think that you've kind of eliminated the need for software

24:48engineers, there's just no chance you're being ambitious enough with your product roadmap. And so we're constantly pushing ourselves of like, now what more can we take on as a result of AI? Yeah, totally. Yeah, even like talking to some of these coding agents, they will overestimate the amount of time that it takes to do a project. If you really, they'll be like, this is a weekend long project. Sometimes they say it's a month. Yeah, they're like, this is a month long project. And I'm like, oh my god, like this is going to be so long. And then I knock it out in like three hours. Yeah, that's because it was trained on all of our Slack messages, pre AI. And so we're the, we're the engineer, you know, said that's going to be two years. So I think it's, it's, it's definitely trained to be highly, highly conservative on project timelines. Same with math too. You ask an AI like, what is your timeline to an AI model disproving and open conjecture and mathematics? Yeah. It'll be like, maybe five years. No, it's already happened.

25:48Yeah, totally. I guess like the last question, because we were talking about this earlier is like, since it looks like model releases are just happening way more frequently for enterprises specifically. It seems like, you know, loyalty to one provider is not going to make it basically. So what do you think is like the optimal strategy here? Like, we were talking about how model routing is the future and like all of these services that offer that sort of thing are the future. Yeah. I mean, kind of that's, that's my conclusion, but I'm also extremely biased in that conclusion. Like, like the, the more that you need multiple models to, to do a task or a set of tasks, the more value accrues to the layer that can understand the task and get access to the data and handle the workflow, which obviously is a better outcome for the, let's say, applied AI layer, which is, which is where we tend to sit. And that's a, that's a, that's a future that obviously a lot of companies are aligned

26:49to and, you know, strategically and kind of existentially in some cases. So if you're cognition and factory and cursor, you know, and at Repplet, et cetera, obviously the outcome that you want to have happen is that you actually have many models. They're all good on different axes. Some are like really like the cost-tuned, you know, kind of workhorse models and some are like the super frontier, you know, orchestrators and you want to have an outcome where you actually need multiple of those models to, to, to be able to complete the task effectively or at least cost effectively. That's actually the, that, that appears to be the timeline that we're on. And, and now, again, I think you have basically five credible US players in, in, in model development between SpaceX, Google, and, and, and, and, and, and, who'd ever get meta. And so the, like those five players are all on a, on a warpath for both driving down the cost of intelligence and improving the frontier at the same time.

27:51So having a model router that can kind of be above that and again, pick and choose at different points, you know, which model to use, I think creates a lot of value for enterprises. And actually helps with, interestingly, it actually helps with the diffusion of AI. You know, one of the, the challenges that enterprises have is, is almost kind of like analysis paralysis on, yeah, on how, if you have so many models that are, that are emerging and they're constantly leapfrogging each other, you actually have like a resistance to just landing on one model family or one partner. And so the applied layer, I think that gives you that relief, where it basically says, you know what, you don't have to make that choice. You can, you can start to kind of get your workflows going, get your data in the right setup. And then you want to use Fable one day, go for it, you want to use GPT 5.6, go for it, or you want to route that to, you know, GROC 4.5, also, you know, go for it, we can lower the cost of the overall workflow. And so that's the, that's the layer that I think is, is going to become increasingly valuable over time. And that layer is, is actually kind of well tuned to understand the deep industry and

28:55vertical use cases, yeah, that AI needs to be applied to. Whereas the pure horizontal models are, are, you know, it's much harder for them to go deep in, in legal and finance and healthcare, not because the model can't do it, but because the, the, the serendin kind of, you know, the apparatus that you need around the model is not there. It needs to get access to the right data, it needs to be implemented into the workflow in the right way. And that's why I think you're going to have just a tremendous amount of value creation at that layer of the stack. Yeah, totally. That does seem to be where we're heading. Well, thanks so much, Aaron. We are at time. It's been a crazy week. Crazy. Crazy. I don't think it's over yet. Oh, we, yeah, we're not, we're, we're, we have at least 10 more hours to go of, yeah, of what could happen today. So open, open source, so much to talk about. We're so glad to have you on. Open AI hack. You guys open AI hacked hugging phase just four days ago. Yeah. So that was four days ago. I thought that was a month ago. We have until 1159 to figure out the next big story.

29:55So many, so many things going to happen today. Yeah. So many things. Thank you so much, Aaron. Thanks for listening to this episode of the A16Z podcast. If you liked this episode, be sure to like, comment, subscribe, leave us a rating, we're a view, and share it with your friends and family. For more episodes, go to YouTube, Apple podcasts, and Spotify. Follow us on X, and A16Z, and subscribe to our substack at a16z.substack.com. Thanks again for listening, and I'll see you in the next episode. This information is for educational purposes only, and is not a recommendation to buy, hold, or sell any investment or financial product. This podcast has been produced by a third party and may include pay promotional advertisements, other company references, and individuals unaffiliated with A16Z. A16Z advertisements, companies, and individuals are not endorsed by AH Capital Management LLC, A16Z, or any of its affiliates. Information is from sources deep reliable on the data publication, but A16Z does not guarantee its accuracy.