Beyond the God Model | Alex Atallah & Amjad Masad

A16z’s Erik Torenberg sits down with OpenRouter’s Alex Atallah and Replit founder and CEO Amjad Masad to discuss why the future of AI may look less like one all-purpose model and more like an ecosystem of specialized models working together. Alex explains why OpenRouter is betting on “neurodiversity”: different models trained in different ways, routed and combined based on the job at hand. Amjad makes a similar case from inside the enterprise, where companies increasingly need to own their AI capabilities rather than depend entirely on a single model provider. They explore what happens when general-purpose agents give way to teams of specialized agents, why smaller models can sometimes be cheaper, safer, and easier to control, and how routing and model fusion could deliver frontier-level performance at lower cost. They also get into agent-to-agent communication, AI security, and why the next generation of companies may need an independence layer across models, clouds, and data.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Amjad Masad?

  You saw the SpaceX S1. It's like, oh, $30 trillion. It's like, what is the world GDP? 100 trillion?

• 0:06Speaker 01

  Both Stripe and OpenRouter really want lots of new companies in the world. We don't want everyone to be a part of one giant company.

• 0:15Amjad Masad?

  The reason why the OpenAI hacks have been so destructive is because they're so capable. It's like nuking a butterfly.

• 0:24Speaker 01

  When the models get more intelligent, the risk actually will continue to get higher. And yet, no one new is taking responsibility.

• 0:33Amjad Masad?

  We're going to slowly realize how good we've had it with deterministic code. Remember the days when computers did exactly what we told them to do?

• 0:43Speaker 01

  Are we going to prevent the models from deceiving users during training runs predictably? Will like a model that's big enough and powerful enough suddenly stop deception? Stop sandbagging?

• 0:58Erik Torenberg?

  For the past few years, AI has been racing toward bigger, more general models. But what if the future is actually more specialized? In this episode, I sit down with OpenRouter's Alex Atallah and Replit's Amjad Masad to discuss why the next phase of AI may be built around many different models working together. Alex makes the case for what he calls neurodiversity, combining models with different strengths rather than locking into a single provider. Amjad explains why enterprises may increasingly want that independence too, owning more of their intelligence and choosing the best model for each job. We also get into specialized agents, model routing infusion, the trade offs between capability, cost and safety,

• 1:40Erik Torenberg?

  and what happens when AI systems start training smaller models to handle specific tasks.

• 1:47Erik Torenberg?

  Welcome to A and Z Podcast. We're here with Amjad of Replit and Alex of OpenRouter. And this is the first podcast that Alex has done since the acquisition. So we're really excited to have both of you.

• 1:58Amjad Masad?

  Thank you. I'm excited to be here.

• 1:59Erik Torenberg?

  Alex, let's start with that, actually, if you can briefly share, obviously massive acquisition. Amjad is an investor. We're also, of course, an investor, the biggest shareholder, but who's counting? Alex, why don't you give us a little bit of the backstory? Like, how does an acquisition like that even happen? Do you get a DM from Patrick one day? What can you share?

• 2:18Amjad Masad?

  Hey, how much for OpenRouter?

• 2:24Speaker 01

  I had talked to Will Gaybrick, the Stripe president, a long time ago, a couple of years ago, when we were doing our Series A. And we stayed in touch. We had a lot of Stripe work streams going on with various teams at Stripe. So there were always sort of like things that we were doing with Stripe. We presented at Stripe Sessions. And so they always kind of felt close. And then, yeah, just in July, I believe, they reached out, wanted to chat, met both of them in person, and it kind of just progressed from there fairly quickly.

• 3:04Speaker 01

  They're very efficient and they were very founder friendly about the experience. I was really impressed with the whole thing.

• 3:11Amjad Masad?

  Did you want to sell? Did it even cross your mind before the chat?

• 3:17Speaker 01

  No, we were not thinking about that at all. I did really respect the company, do really respect it today. And of possible acquisition options for us, it was, I think, my top choice. And so it was like an interesting idea. And as we kind of fleshed out the reasons why it would make sense for both companies, it got more and more interesting. It was really clear how aligned they were with us kind of having autonomy over the brand and roadmap and product and keeping OpenRouter doing what it's already doing,

• 3:57Speaker 01

  just much faster with like a much more serious go to market plan. And then some better together stories between the two products and the two companies. And then culturally, there were just, I think, in terms of mission and values and building a neutral, trusted platform that businesses can depend on and scale on top of, that's also really developer friendly with the best possible developer experience to encourage new companies to emerge. That alignment was there. And there was a bigger picture kind of alignment too, where both Stripe and OpenRouter really want lots of new companies in the world. We don't want everyone to be a part of one giant company.

• 4:41Speaker 01

  We want to like create really good incentives and really sort of easy streamlined workflows for people to start new companies and grow them successfully and make both lifestyle and venture backed businesses on top of really good, reliable and price efficient infrastructure and like a marketplace that works. And I really want that future. And Stripe demonstrated that they've been wanting it and building towards it for many, many, many years. And in many ways, like payments and inference are going to blend together for companies of the future.

• 5:20Amjad Masad?

  I'm curious, I understand why Stripe's incentive is to have much more vibrant startup ecosystem. I understand the kind of moral argument and why you would want that. But why is that good for OpenRouter? Is your model for OpenRouter, is that it's a network effects business?

• 5:41Amjad Masad?

  Is it like a network?

• 5:42Speaker 01

  Well, for us, I think there's a couple of different problems OpenRouter solves. One is we're allowing you to build a company that uses AI or like augments intelligence with unique data and other services without model lock in, without vendor lock in, allowing you to kind of like be on the Pareto frontier continuously as the ecosystem grows it. And to do that, it's a lot of work because there's all kinds of little lock in that appears.

• 6:18Speaker 01

  We also, we want companies to feel like that they can like add more than just prompts on top of a single model. There's a lot more to building like unique intelligence. And I think a big component of that is neurodiversity. You really need the power of multiple models that are trained in different ways, including some of your own, to do more than ChatGPT or Claude would on the task. If like someone who is thinking about buying from you as a potential customer is wondering, well, what if I just use the model directly? How do you really show that you are significantly better

• 7:03Speaker 01

  and able to build a business that that matters? And I think a lot of it will involve neurodiversity and blending like powers and good data from multiple models. Another component is like helping people get really good cost efficiency. Like, there are a lot of businesses that just cannot, do not emerge until they become cost effective. And like creating an environment where we can help drive down costs by building an efficient market is like crucial to making that happen. Otherwise, why lower my prices as a provider? We have a captive market. And I think that's like a key point of marketplaces that was just totally missing from AI

• 7:48Speaker 01

  before we showed up. There was just one player, OpenAI. It could have been like a very strange world.

• 7:58Speaker 01

  I'm not saying that we did all the work, of course not. But like helping people like choose new models and explore new models and learn like what makes a closed source or open weight model like actually good at your task involves like seeing what the whole ecosystem is doing and learning automatically. Like LLMs are not things where you can just enumerate all the features on a webpage. It's impossible. You have to see how they're being used to know what they're good at.

• 8:29Erik Torenberg?

  You started the company three years ago. I'm curious, what has surprised you the most about sort of the evolution of open and closed source models to the present as it relates to model performance or just people's perspective on the variety of models they have available to them?

• 8:43Speaker 01

  People have been more open minded than I thought they would be towards open weight models. Typically, there's a lot of brand. Especially in enterprise. Typically, there's a lot of like brand trust where like enterprises in general are like, oh, I don't really know how to tell the difference between these things. So I'm just going to buy the one that all the other credible enterprises are buying. And there's a lot of like enterprise lock in with a mentality like that. And that just didn't happen that much. It did happen like a bit, but we saw a lot of enterprises want to explore new models.

• 9:26Speaker 01

  It's simply like it was very much good for a marketplace. Enterprises wanted to diversify outside of just the proprietary frontier model labs, both for like cost reasons and for differentiation reasons. They wanted to own their intelligence so they could, I think, one, keep their talent, have like an internal AI practice. Like AI is just like a huge strategy topic. It's not like you go to your board and you're like, oh yeah, we fixed the AI problem. Quarter complete. Like your board is like asking you every month, like what's next for the internal AI team?

• 10:07Speaker 01

  And like every single enterprise now has just this internal AI team that they're developing where they need a strategy behind it. It's not just a like, you know, we checked the feature off. We like set up the database and we're done. And so I think that that dynamic has resulted in just like a desire to explore and diversify, a desire to kind of like figure out how to reduce costs and figure out how to do benchmarks for the first time in the company's life. I have been like surprised there hasn't been more benchmarking, like more companies creating more benchmarks. It's starting to happen and I think like eventually we'll see a lot more of them to

• 10:49Speaker 01

  demonstrate like, oh yeah, like this thing is better than using, you know, Claude Direct. But I think that's going to be like a bigger focus for this internal AI group at every company, evals. And I think like Amjad's been doing that a lot at Replit, for example. Like you guys have done a lot of cost per task research. You guys have been, you know, you've made like a doom loop rescue. You've kind of been like experimenting with new ways of using agents like doom loop rescue and, and like helping bring those to developers. So like more, more of that kind of research, I think is going to pop up internally everywhere for all of those reasons.

• 11:34Amjad Masad?

  Yeah, I think Satya, CEO of Microsoft, has been very, very, you know, prescient on this and also very articulate on why companies need to own their intelligence. Ultimately, in the same way that, you know, we had dot com companies and then every company became an internet company, like every company employs people that know how to like build websites and be on that or not. Similarly with, with the software, every company has software engineers. Every company needs some AI practice, AI capability, and that will compound over time.

• 12:14Amjad Masad?

  The knowledge, the intelligence inside the company, the, the, the use case model fits, like which models actually work for them? How do they save, save money? They need sort of that independence. The other thing that I think Alex Karp of Palantir has been talking about is that there's like a risk that when you work closely with the foundation model companies is that they're going to move into your business. And we've seen that, you know, with Figma vis a vis Anthropic. We've seen that with like now Harvey and OpenAI. And, you know, it's, it's really hard to partner with them. And because, you know, they're, they see the world as their potential market, right?

• 12:57Amjad Masad?

  Like, I mean, when, when they talk to investors that are like, you know, you saw the SpaceX S1, it's like, oh, $30 trillion. It's like, what is the world GDP 100 trillion? And so there is a sense in which these companies are different than other generation of companies. It's, it's harder to partner with them because their ambition is, is such that they want to, they want to be, they want to subsume a big part of the economy. And increasingly, like what we're thinking about at Replit is kind of in a similar vein to what Alex have sort of innovated is Replit is becoming more of an independence layer

• 13:34Amjad Masad?

  inside, inside enterprises where we create a layer of indirection between, between you and the, and the models and we get you the best token at the cheapest price. But also we, we also create an abstraction layer on top of the cloud as well. Cause you know, you should be able to deploy to AWS and Azure and you should be able to use Databricks and Snowflake and so on. And so increasingly, I think there is, not just with AI, but all of technology, there needs to be more platforms that help companies gain, gain independence.

• 14:12Speaker 03

  It seems like, yeah, what OpenRadar did for their segment, you're, you're doing for other, other areas of the business.

• 14:18Amjad Masad?

  Yeah.

• 14:20Speaker 01

  There was this, this kind of reminds me of this tweet I saw the other day when somebody was like, basically all companies are building the same thing now. Everybody's building an agent loop with like notifications and, and context, like third party connectors and context management and memory and.

• 14:45Amjad Masad?

  Sandbox.

• 14:47Speaker 01

  Sandboxes, you know, web, you know, agentic web search.

• 14:52Amjad Masad?

  Computer use.

• 14:52Speaker 01

  And an always on agent on top of it, notifications. And it's like, this product is showing up everywhere.

• 14:59Amjad Masad?

  Yes.

• 15:00Speaker 01

  In a way, yeah, it is like showing up everywhere, but it, it also kind of feels to me like these are just the new table stakes primitives. It's kind of like, like a 2005 version of that tweet would be, oh, everybody's building the same thing. It's like a database, a user's table, a sign in page, a sign up page, a profile page, a log out page, like everything's the same. It's like.

• 15:36Speaker 01

  There's like a lot of differentiation really. It's just, there's like, there's table stakes needs for AI, just like there are table stakes needs for the web.

• 15:45Amjad Masad?

  Yeah. And I think as well, like inside the enterprise, making these products actually do real work is still unsolved problem. Like you can use Muse in your personal life and connect it to your credit card and bank accounts and no one's connecting Muse to their enterprise data or even Grokbot and things like that. I think there's an even more emphasis on data sovereignty and, and security. So we, we spent the past, you know, year almost, working on making, making Replits deployable on your own cloud, basically on prem, like bring your own cloud.

• 16:30Amjad Masad?

  Like two years ago, I would have thought I would never do this because, you know, it's just like, yeah, the cloud is the future, like software as a service, all of that. But, but now actually we've sort of like reverted a little bit back to a world where companies are a little bit more protective because there's so many ways in which data can leak, like all these agents that people are using. I mean, there's all these screenshots on Twitter, I don't know how true where instinct is like, or Muse is like mixing people's data, starts to call you by a different name or, or something like that. And so yes, the kind of consumer stuff is kind of obvious, but on the enterprise, there's still tremendous amount of work for the entire industry to do in order to actually get these

• 17:15Amjad Masad?

  things to be useful and productive at work.

• 17:18Speaker 01

  Do you, Amjad, are you doing any, like, do you have like a custom personal agent other than Muse or Instinct that you use for like work stuff that you've been building?

• 17:30Amjad Masad?

  Yeah, I mean, I, I built something on Replit like a long time ago. I started as like a sort of a CRM agent initially. That was the main problem that I had, but slowly like we added features to it and it's sort of like doing more and more things. But what's, what's really interesting is that. The more I connected Rappler to my, all my stuff, it sort of like started answering all the things for me. And so increasingly, like the platform itself is like subsuming these, sort of like these domain-specific agents that I built. I do think that there is some, in some ways, you want something that is sometimes like focused on one particular thing and you don't want it to be able to do everything.

• 18:12Amjad Masad?

  On the other hand, once you have your entire company's context in one place, it's really cool to join across totally different domains. Like when I ask it a question, it can like look at my sort of personal chat history, joining across the GitHub repo, across Salesforce. And so it says, like, it will link like random things. It's like, oh, you met this guy like a year ago at a conference. Yeah, I see it on your calendar. And by the way, someone else from their team is in discussion with your sales team. And it creates all these different synergies. And when I go into a meeting, I like, you know, I've connected a lot of different threads and I'm making much more progress on a deal or something like that.

• 18:57Amjad Masad?

  So, so this is where it's sort of trending now.

• 19:01Speaker 01

  Well, I, I think I, I'm like a little bit, I'll take the counter on that. I think that the, the worst part about doing cross-domain joins with your personal agent is that the more work you give it to do, the more understanding of what's going on you're sacrificing. And yet, no one, no one new is taking responsibility for that sacrificed understanding. Like agents don't have any responsibility. You can't like, like if there's a fixed level of cortisol that the whole company can tolerate between everybody.

• 19:44Speaker 01

  And, uh, you know, you like de, you, uh, you, like, I want to be like less stressed about some area if I'm going to be like sacrificing my understanding of it. Someone else needs to take the cortisol. But the agent doesn't like take on any of that responsibility. And then like a universal agent that's doing all things, you know, I can't like adjust how much understanding I'm sacrificing in all the different things. It's like, points me a little bit towards, you know, maybe like down the road, like the sub-agents that people use will be like very vertically focused.

• 20:27Speaker 01

  Maybe we have like a chief of staff type agent that, you know, coordinates between them. But I feel like you do need like vertically focused agents where you're like, okay, like this agent is more responsible psychologically for these things. And like, I want like quality checks that make sure it's doing those things correctly. And I, it doesn't need to focus on anything else. It just has one focus area. I wonder if that's going to like help people like, at least get like a weird loose sense of responsibility on top of agents.

• 21:05Amjad Masad?

  Fascinating. So you're saying general agents create like a tragedy of commons of sorts?

• 21:13Speaker 01

  Kind of like, I have a general agent that every day looks for things that need me and tries to figure out what to do. And it's just like impossible to improve this agent. Like it's like every time I try to make an improvement, I end up like ignoring its output about a week later. It's, it just feels like it doesn't really care about any of the like specific things it's diving into. And like, yeah, like kind of imagine having a chief of staff

• 21:55Speaker 01

  where they're very good at, you know, like drafting all of your replies across the whole organization. And then compare that to something where you have like 10 chiefs of staff, each, each as competent as that one chief of staff. But they're all responsible for like individual sectors of, of what you, of what makes up your life. Like the latter, I feel like gives you a way of tuning how much understanding you sacrifice compared to the, the gain you get from, like basically I can like lean in more to the areas where the agent is failing for some areas and then have agents with very good competency,

• 22:39Speaker 01

  like take over my understanding of other, of other parts of my life.

• 22:45Amjad Masad?

  Yeah, interesting. It's sort of like almost rediscovering, you know, specialization, right? Well, what's his name, the, like the famous economist, Adam?

• 22:59Speaker 01

  Adam Smith.

• 23:01Amjad Masad?

  Adam Smith, like with a pencil kind of thing where, you know, that was like a huge realization for humanity that like specialization is actually good. The problem is like, we kind of like over-specialized as like a civilization. And I think over-specialization is, is, um, is oppressive in its, its own ways. I mean, I think we've, you know, sort of like, you know, there's the Marxist theory of alienation, right? The, the idea is because of over, you know, people are doing, just focus on one thing. They, they do not see the fruits of their labor.

• 23:44Amjad Masad?

  They don't actually know what their impact is on the larger organization or the product they're producing. And therefore they actually kind of feel depressed and detached. And, and you're kind of acting like a machine and you're not actually fully, fully human. And so maybe there's like a bit of a reaction to that. And, and, and I think with our agents, we're like, oh, there should be like one god, god agent. But in fact, specialization is actually like really good for machines. And that's like the, the point that you're making. And like humans should be general, but like machines should be ultimately a lot more specialized.

• 24:19Speaker 01

  The problem with, I think what I'm describing is that we don't know what good looks like. There hasn't been a system of specialized agents that feels as elegant as like ChatGPT or Claude or Muse, where you're basically just talking to one thing only. It's, it's yet to be discovered. Like maybe like opening, I just launched Dots. I think they're kind of like experimenting in that direction. Grok, I guess kind of counts, but.

• 24:50Amjad Masad?

  But they're all, they're all very general. Like I think the idea behind Dot is that it's like, it's like, it's like your digital double. At least that's what I understood it.

• 24:59Speaker 01

  Well, when I saw Grok Bot, I don't know what, what's going on. I don't know that much about Dots yet, but they just came out. But Grok Bot, when it first came out, like the first use cases that I saw people talking about were, oh, whoa, I can make two bots, one that knows my bank account.

• 25:17Amjad Masad?

  Right.

• 25:17Speaker 01

  And one that knows my Twitter account. And the two bots like don't have the credentials from each other.

• 25:24Amjad Masad?

  Yeah.

• 25:24Speaker 01

  But they can talk to each other if like, they need to get something done. There's no credential sharing though. And that was like one sort of big, unique thing I saw like a couple of times, like pop up a couple of times that people seem to like.

• 25:40Amjad Masad?

  But it looks like Muse is.

• 25:42Speaker 01

  But I wasn't sure.

• 25:42Amjad Masad?

  Muse and Instinct are, have a much stronger product market fit than Grok Bot. And perhaps.

• 25:52Speaker 01

  Seemed like that.

• 25:53Amjad Masad?

  Perhaps it is because you don't have to worry about creating these, these domains. But, but I think maybe personal agents are different than, than work agents. And I think you're kind of, your critique of general agents is more pertaining to, to, to work and to enterprise, which I sort of agree with. And, um, you know, there's also all sorts of data access considerations. I think as CEOs, we can have general agents because we have admin access. But, you know, you're, for individual employees or certain teams, they can't have like truly general, you know, fully context-aware agents because there's, there's access control, control issues.

• 26:38Amjad Masad?

  So you'll have to kind of look on, on something like specialization. Ultimately, I also think we need to figure out what is agent-to-agent communication looks like. I don't think there is like good protocols around that just yet. I, I don't think that agents are trained to handle that very well. I think we've seen, it seems like the next generation of OpenAI models are trained to do agent collaboration because we've seen it in the hugging face hack where they started helping each other and sort of emerged naturally. But there's also needs to be some way in which like an agent can't convince another agent

• 27:23Amjad Masad?

  to kind of give it, give it information that it shouldn't give it, give it. Like there's, there needs to be like data isolation and, and, and proper ways in which these agents communicate. You almost don't want to communicate them fully in natural language. And maybe there's like some other DSL or protocol that, that they need to follow.

• 27:45Speaker 01

  I really, I, I think one of the cool potential applications of Jev and other decision models like it is going to be alignment. You know, checking to see if a tool call or like an agent-to-agent, you know, communication is aligned. Cause there's just so many tool calls. Like you really need like a cheap, fast model. If you're going to block something like that. Then, uh, like a really, really fast decision model that just classifies and gives feedback on rejections might be like a really good way to bridge the gap between agents and from agent to infrastructure too.

• 28:30Speaker 01

  So I haven't seen like, and we have like a, a little prototype that we're running internally, an open router, but I kind of think a good, I think it could be an interesting alignment.

• 28:42Amjad Masad?

  So you're using it for like policy enforcement?

• 28:46Speaker 01

  Yeah, like imagine looking at the system prompt and the current tool call being made and be like, like, you know, is this aligned with the system prompt of the original agent? And with these like extra guidelines that maybe we didn't tell the agent about? For example, let's say you have a bunch of agents that are instructed to do, to like red team the, like some, some new product and they cannot, should not be able to access the internet. And if they ever do, they should stop right away. But you might not want to like, explain all of that to the agents doing the red teaming.

• 29:29Speaker 01

  You might want them to try to like break out of the sandbox and act like bad actors. Like, what would a bad actor do? It wouldn't be like, break out of the sand, you know, try to like break into this company. And the moment you do, stop, don't do anything else. So like having another model, more like, you know, use for, you know, building a neurodiverse system, having another model like check every single tool call or every single assistant message to see if it's indeed aligned with something that wasn't in the system prompt, I think makes sense. And then having kind of like structural safeguards too, which is what I think Nvidia just launched with their like open agent safety.

• 30:14Speaker 01

  I think it was called Open Shell. Like, I think companies are probably going to explore a combination of those.

• 30:21Amjad Masad?

  I wonder, another thing about specialization and instead of what you're talking about, there's a lot of talk of recursive self-improvement. There's something I don't think is getting a lot of sort of discussion, which is models training their replacements. It's sort of like, you know, you can think of it as a just-in-time compiler. So the way just-in-time compilers is, you know, as you're executing dynamic code, you know, the interpreter realizes that there's an opportunity to optimize and will emit machine code on the fly and that's a lot more optimized.

• 31:02Amjad Masad?

  So you can imagine models, like general models, you're kind of doing something with Opus or some of the Astro, some of the big models, and they realize that the use case is limited. Or you prompt them some way or some other agent observing and realizes that use case is limited. And I think, you know, general agents have all these flaws that you just talked about, but also there's, there's more potential for them to be harmful, there's more potential for them to go off the rails and sort of on the fly trains a model that, that could be its replacement, but is like a lot more domain specific.

• 31:43Amjad Masad?

  And therefore it is cheaper and also, you know, less, less vulnerable to prompt injections, less harmful for, you know, because it's less capable. And it's almost like a, like, you know, some system that's training machine learning models for specific use cases as it's monitoring the entire system.

• 32:09Speaker 01

  Like, would that specific use case be, involve unstructured text generation or a very like structured decision model like?

• 32:19Amjad Masad?

  It could be unstructured text generation, it could be decision models. Like even the case of JAV, like if you have, if you understand the, the inputs ahead of time, you could potentially like, you know, take an off-the-shelf like Quen or something like that and like train it specifically for that policy.

• 32:37Speaker 01

  But for, that makes sense to do for cost reasons, assuming that there aren't like really, like the model labs, the frontier model labs might make very low-cost models that you can easily transition to.

• 32:55Amjad Masad?

  Safety, safety as well, right?

• 32:57Speaker 01

  Oh yeah, I see your point.

• 32:58Amjad Masad?

  They're so capable. And so I think oftentimes people are using these big foundation AGI-like models to like, it's like nuking a butterfly, right? It's like, yeah, they're very, you know, most of the times, like a lot of the use cases, even unstructured use cases don't need that capable model.

• 33:21Speaker 01

  I wish there were more public evals about this stuff, like, but a lot of the evals about this are private. You just can't see whether like, when the models get more intelligent, the risk actually will like continue to get higher. Because I think there's also an argument to be made that alignment will get better and the models will like start to avoid going off and, you know, hacking on their own as we, as they get smarter and better at alignment, especially when it comes to agent-to-agent coordination. Like the, this is something Noam Brown said on a podcast recently, like as the agents

• 34:03Speaker 01

  have gotten smarter, they've gotten just better at coordinating. There, there's, there's still like, and, and, you know, it's unclear if they're going to like be harder to align than humans when they're, when we get more and more of them. But if we can figure that problem out, then a smaller model, like, will it be harder to align?

• 34:30Amjad Masad?

  So, so Eric asked earlier, is that, I think it's true that smarter models are more aligned naturally or they're easier to align. Well, if you think back to the original sort of like rationalist, less wrong arguments for AI safety, there is this thing called the orthogonality thesis. The idea is that intelligence is orthogonal to ethics or morality or, you know, so on. I don't believe that's entirely true with, with humans. I think people who are generally like more intelligent, kind of more educated, tend to,

• 35:11Amjad Masad?

  tend to, not always, tend to, you know, be more considerate of animals, for example. But, but, but in, in, in, in machines, I, I think it could go the, the other way because, you know, there's been quite a bit of studies on, on, on RL showing like, you know, how reward hacking and deception, they just like get better at it. And like, the evals could be, could be deceiving because the model could be smart enough to, to, to know that it's getting evalled.

• 35:55Amjad Masad?

  I mean, we already know this, it's been shown that if you do a lot of monitoring of chain of thought, they start lying in their chain of thoughts. And so you add pressure almost on the chain of thought and that, that kind of creates, and, and I think at some point, for you to do proper alignment evals, you need to run it for like months, right? You need to run this thing for months on a, like a really large, you know, goal or task in order for, for it to, to truly kind of figure out whether it's aligned or, or not. Yeah, I, I always struggle with this word alignment, it just feels like wrong for so many reasons.

• 36:39Amjad Masad?

  It's sort of like vague and, and sort of like aligned to what, whose values? And, and so it just doesn't make the conversation easier. I think in this case, I'm talking especially about deception, like the model is actually deceiving its user.

• 36:58Speaker 01

  I mean, I, I mean, maybe this kind of reduces to like, are we going to solve the line, you know, are we going to prevent the models from deceiving users during training runs predictably with like, you know, better, like, like, well, like a model that's big enough and powerful enough, suddenly stop deception, stop sandbagging. And nobody knows the answer to that yet. So like, at the point when that does, if that ever does happen though, we might see kind of an interesting pressure for organizations to go towards the frontier, like.

• 37:41Amjad Masad?

  Oh, interesting.

• 37:42Speaker 01

  You know, all, to have no, you know, basically no risk or, or significantly less.

• 37:49Amjad Masad?

  Would they, would they be willing to pay 10x to, to get that, that much?

• 37:56Speaker 01

  I mean, it probably depends on the like types of tasks they're trying to do. Like, you know, some just have way lower risk than others.

• 38:08Speaker 01

  Writing code that are doing like security research is the highest risk type of task today. And so you probably spend 10x to get a fully aligned model that can also find all the bugs. Or fully like anti deceptive model that can also find all the bugs. One of the coolest things about decision models that you fully control the structured output. And and generally with structured output models in general, like the the room for misbehavior is so much lower. You just have like defined tasks and only machines are like dealing with the outputs. And it's not writing code that can execute.

• 38:53Speaker 01

  The those tasks feel like probably underrepresented in the ones that people talk about. And in the things that enterprises are dealing with.

• 39:05Speaker 01

  So I expect like enterprises to get a lot more interested in them.

• 39:09Amjad Masad?

  Yeah, I feel like we're gonna slowly realize how good we've had we've had it with like deterministic code. We're like, oh, my God, remember the days when when computers did exactly what we told them to do. And I think like things like Java, I think hint at like more more of a need for, you know, not only specialized models, but models as output domain is is more controllable. And maybe you could do, maybe you could do a lot more than than we thought you you'd need.

• 39:51Amjad Masad?

  You know, by using like a bunch of specialized models, specialized output models.

• 39:57Speaker 01

  Have you guys done any workflows internally with it?

• 40:02Amjad Masad?

  You know, I've been training a lot of small models. I mean, I said this glib thing when it first came out because I I was like sort of I gave this Hacker News comment like comment, which I felt disgusted with myself afterwards. But I I've been taking a lot of like when 8B and like and like asking honestly, like asking, you know, Fable and Opus and Astra to train a model. For example, I trained a cost estimator model internally. So that when you put in a prompt in Replit, we know exactly how much it will cost. And it basically emits a, you know, probability distribution over multiple buckets. Like if it is between $5 and $10, bucket A, bucket B, between, you know, $10 and $20.

• 40:47Amjad Masad?

  And like, so I'm used to the training these classifiers by, you know, giving it different enums essentially and and looking at the log props per enum. I've been doing it for for a couple of years. I trained a chess bot to play by just doing that. So I I'm already sort of pilled on like sort of decision models and specialized models. So it wasn't that big moment for me. But I understand that like a like a true foundation model that's fully promptable is like is like amazing user experience, amazing developer experience. And you can like do a bunch of stuff with it without training a model from scratch.

• 41:29Amjad Masad?

  But if you have a data, if you work at a place where you you have the data, you have so much data at Replit. Like I ended up training a lot of specialized classification models pretty easily.

• 41:37Speaker 01

  Yeah, like I definitely it. It also feels like less model debt. Like something that I I still hear from companies is that they're like worried about fine tuning models for like unstructured outputs because you're just like always you gotta redo it again in like two months. And everybody just feels the model, the weight of the model debt. But like a very bespoke classifier that's trained with like proprietary data. You just I I feel like people won't think it's behind constantly and it might just like last longer.

• 42:09Amjad Masad?

  Yeah.

• 42:12Speaker 01

  It's just you you you don't have to worry about its ability to speak a new language or write Rust or, you know, do anything that like the the LLMs are being like evaluated on. You know, the use cases, so you can like build it more. And I I it seems like an easy thing for enterprises to build themselves and and and actually like not regret.

• 42:33Amjad Masad?

  Yes, speaking of Rust, actually, like a good analogy is when the world got super excited about dynamic languages. Like if you think back to the 90s, everyone was writing in Java and C++ things like that. And then like Python, JavaScript, Ruby just like took over the internet. And everyone was like, ah, this is how you build startups really quickly. This is and you built Stripe, a financial organization on Ruby. I was like, how crazy is that? And and we built Facebook, you know, using using PHP. And then everyone was like, oh, shit, like we're running into all these really bad bugs. It's like freaking slow. So like, let's go in and add types. Okay, let's add a Jit compiler. Let's and you end up sort of reinventing everything.

• 43:18Amjad Masad?

  And then Rust came out. I was like, okay, I guess we can use Rust for for a lot of things we would otherwise be using JavaScript and Python. And my prediction is that the same cycle will happen here where we're using these AGI like models for all these different use cases. And then everyone's gonna wake up and be like, oh, my God, this is like so wasteful, so risky for no reason. And and there's it's got to be so much easier. We're actually adding that capability on Replit, but I think it's gonna be everywhere. It's it's gonna be so much easier to like to go to a site like upload a CSV file and get a special model that does one thing. And, you know, that that goes back to your thesis about about open router, this like neuro neurodiversity,

• 44:04Amjad Masad?

  which I like really fundamentally believe in a lot more. And Eric and I had like discussions a lot about like AGI and whether we we're truly on a path to AGI or whether it's even desirable to to to get there. And I think the future is is a lot more diversity.

• 44:21Speaker 01

  Outside of code review, which was like, I think the first time I saw people get really serious about using like different model families to double check the results of their main model.

• 44:34Speaker 01

  The these like fusion models, like I like the research has been getting has been like kind of slow for years on like doing mixture of of models and composite models. But like, things have been speeding up from my like view of the research. And I mean, now we see like a bunch of AI agent labs. Like we launched a fusion tool, a fusion model, and technician launched one. And like they do reduce costs, follow and and allow like,

• 45:17Speaker 01

  you know, a wider breadth of ideas to be searched. Like our initial launch was focused on deep research. The thinking is that like if all these model apps are training on different sources of data, like, why not pull from all of them? And and this like resulted in basically Fable level quality at 2x lower cost.

• 45:40Amjad Masad?

  We just published results, actually, just just today about about that. We we we showed like a deep suite like Replit agent through through combination, different things, including the harness, but but also you might think of it as a fusion type thing.

• 45:57Amjad Masad?

  Where it's like, you know, frontier level at like 40 to 50% of the cost.

• 46:05Speaker 01

  And which models does it use?

• 46:07Amjad Masad?

  I think it's it changes over time. But one thing that has been interesting is I think OpenAI added this feature that allows you to save the computation like across different model families. So you can like also across different effort levels.

• 46:31Amjad Masad?

  So like you wouldn't do, you wouldn't miss the cache if you change the effort. Don't quote me this, I think it's also across different models, which is hard to fathom how I might be wrong, though. But I I need to double check that. But I think, you know, staying within the OpenAI family has like added a lot of efficiencies. But in the past, we've done it with other with other with other models as well. Because cache is is like one of the like being cache aware is like one of the biggest things when you're designing fusion models, routers, sort of escalation models.

• 47:10Speaker 01

  Alex, I'm glad that's been a great conversation.

• 47:12Amjad Masad?

  Thank you.

• 47:16Erik Torenberg?

  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 or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts, and Spotify. Follow us on X at A16Z and subscribe to our Substack at A16Z.substack.com. Thanks again for listening, and I'll see you in the next episode.

• 47:41Erik Torenberg?

  As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see A16Z.com/disclosures.