title: "Why Specialized AI Could Beat The God Model" source_url: "https://youtu.be/ekK8urKHPMQ"

Why Specialized AI Could Beat The God Model

00:00 开场

You saw the SpaceX S1, it's like, oh, $30 trillion. It's like, what is the world GDP? 100 trillion? 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. When the models get more intelligent, the risk actually will continue to get higher. And yet, no one new is taking responsibility. 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. Are we going to prevent the models from deceiving users during training runs predictably? Well, like a model that's big enough and powerful enough, suddenly stop deception, stop sandbagging.

00:47 Stripe 收购背后的故事

Welcome to a16z 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. Thank you. I'm excited to be here. 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 acquisition like that even happen? Like, do you get a DM from Patrick one day? You know, what can you share? Hey, how much for OpenRouter? I had talked to Will Gabrick, the Stripe president, a long time ago, a couple 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 it always kind of, they always kind of felt close. And then, yeah, just in July, I believe, they reached out, wanted to chat, met like both of them in person. And it kind of just progressed from there fairly quickly.

They were very efficient. And, you know, they were very like founder friendly about the experience. I was really impressed with the whole thing. Did you want to sell? Like, did it even cross your mind before the year's chat? No. We were not thinking about that at all. So I, you know, did really respect the company, do really respect it today. And, you know, of possible acquisition options

for us, it was, I think, my top choice. And so it was like an interesting idea. And, you know, as we kind of like fleshed out the reasons why it would make sense for both companies, it got more and more interesting. It was really clear how like aligned they were with us kind of having autonomy over the brand and roadmap and roadmap and product and keeping OpenRouter doing what it's already doing just much faster with much more, with like a much more serious go to market plan.

And then, you know, some better together stories between the two products and the two companies. And then culturally, there were just, I think, in terms of like mission and values and building like 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 for, you know, to encourage new companies to emerge. You know, that alignment was there.

And there was a bigger picture kind of alignment too, where like, you know, both Stripe and OpenRouter really want lots of new companies in the world. You know, we don't want everyone to be a part of one giant company. 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, you know, make both lifestyle and venture-backed businesses

on top of really good, like reliable and price-efficient infrastructure and like a marketplace that works. And, you know, I really want that future. And Stripe demonstrated that they've been wanting it and building towards it for like many, many, many years. And in many ways, like payments and inference are going to blend together for companies of the future. I'm curious, I understand why Stripe's incentive

05:08 OpenRouter 为什么需要更多创业公司

is to have, you know, 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? Like, is your model for OpenRouter, is that it's like a network effects business? Is it like a network? Well, for us, I think the, there's like a couple different problems OpenRouter solves. One is, you know, 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. We also, you know, like we want companies to feel like the best way

that they can like add more than just prompts on top of a single model. There's like 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? Like how do you really show that you are significantly better and able to build a business 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, like, you know, why lower my prices as a, as a provider? Um, we have a captive market and, uh, and like, you know, I think that's like a key point of marketplaces that was just totally missing from AI before we showed up. There was just one player, open AI. It could have been like a, uh, a very strange

world. 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, um, you know, a closed source or open weight model, uh, 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 web page. It's impossible.

You, you have to see how they're being used to know what they're good at.

08:53 企业为何选择开放权重模型

So you, 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, um, as it relates to model performance or just people's, uh, perspective on the variety of models they have available to them. Like people have been more open minded than I thought they would be towards open weight models. Like typically there's like a, there's a lot of brand, especially an enterprise. Typically there's a lot of like brand trust where

like an enterprise is, you know, enterprises in general are like, Oh, I'm, you know, I don't really know how to tell the difference between these things. So I'm just going to like, you know, buy the one that all the other credible enterprises are buying. And, and there's a lot of like enterprise, um, lock in with like a mentality like that. Um, and that just didn't happen that much where they, it did happen like a bit, but, uh, we saw a lot of enterprises, uh, want to explore new

models. It's, it's, it's simply like, it was, it was very much good for a marketplace. Um, enterprises wanted to diversify outside of just the proprietary frontier model labs, um, both for like cost reasons and for differentiation reasons. And they wanted to like own their intelligence so they could, um, you know, I think like one, keep their, their talent, like 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. Um, you know, we fixed the AI problem, uh, quarter complete. Like your board is like asking you every month, like what, you know, what's, what's next for the AI, the internal AI team and all like every single enterprise. It now has just this, you know, internal AI team that they're developing where they, they need a strategy behind it. Um, it's not just a, like, you know, we, we,

we checked the feature off. We like set up the database and we're done. Um, and so I think that, that, uh, dynamic has resulted in just like a desire to explore and diversify. Um, and 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, we'll see a lot more of them, um, to demonstrate like, Oh yeah, like this thing is better than using, you know, Claude direct. But, um, like, I think that's going to be like a bigger focus for this internal AI group at every company, um, 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. Um, you guys have been, you know, you, you've made like a doom loop rescue.

You, you kind of been like experimenting with new ways of using agents like doom loop rescue and, and, uh, like helping bring those to developers. Um, so like more, more of that kind of research, I think is going to pop up internally everywhere. Um, for all of those reasons. Yeah, I, I think, um, uh, Satya, uh, CEO

12:26 为什么要拥有自己的智能

of Microsoft has been very, uh, very, you know, prescient on this and also very articulate on, um, 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 the internet. Similarly with, with the software, every company has software engineers. Every company needs some AI practice, AI

capability. And that will compound over time, 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, um, I think Alex Karp of, um, Pound here 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, and because, you know, they're, they see the world as their potential market, right? Like, I mean, when, when they talk to investors, they're like, you know, you saw the SpaceX S1. It was like, oh, $30 trillion. It's like, what is the world GDP?

A hundred 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, is such that they want to, they want to be, they want to subsume a big part of the economy. Um, 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, uh, Replit is becoming more of an independence layer inside, inside enterprises.

Where we create a layer of indirection between, um, between you and the, and the models. And we get you the best token at the cheapest price. Uh, but also we, um, we also create, uh, 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. Um, and so increasingly I think there is not just with AI, with all of technology, there needs to be

more platforms that help companies gain, gain independence. It seems like, yeah, what OpenRouter did for their segment you're, you're doing for other, other areas of the business. Yeah. 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,

uh, like third party, uh, connectors and context management and memory. And, uh, sandbox, uh, sandboxes, uh, you know, web, you know, agentic web search. Computer use. And, and always on agent on top of it and notifications. And it's like, this product is showing up everywhere. Yes. Um, 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, there's like a lot of differentiation.

Really. It's just, there's like a, there's table stakes needs for AI, just like their table stakes needs for the web. Yeah. And I think, um, 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 Grok Bot and things like that.

Um, I think there's an even more emphasis on data sovereignty and security. So we, we spent the past, you know, year almost, uh, working on making, uh, making replets, uh, deployable on your own cloud, basically on-prem, like bring your own cloud. Like two years ago, I would have thought I would never do this, uh, because, you know, just like, yeah, the cloud is the future, like software as service, all of that. Uh, 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, 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 something like that. Uh, 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 things to be useful and productive at work.

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? Yeah. I mean, I, I built something on replet like a long time ago. Uh, 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 replet to my, all my stuff, it's sort of like started answering all the things for me. Um, 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, 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. 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, I, you know, I, I've connected a lot of different threads and, and I'm making much more progress on a deal or, or something like that. Uh, so, so this is where it's sort of trending now.

19:52 反对全能 Agent 的理由

Well, I, I think I, 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, 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 knew is taking responsibility for that sacrifice. Understand it. Like agents don't have any responsibility. You can't like, you like if there's a fixed level

of cortisol that the whole company can tolerate between everybody. And, uh, you know, you like D you, uh, you like, I, 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. Um, and then like a universal agent that's doing all things, you know, I, I can't like adjust

how much understanding I'm sacrificing and 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. Maybe we have like a chief of staff type agent that, you know, coordinates between them. Um, but I feel like you do need like vertically focused agents where you're like, okay, like there, this agent is more responsible psychologically for these things.

Um, 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. Um, 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. Fascinating. So you're saying general agents create like a tragedy of commons of sorts. Kind of like, um, kind of like I have a general agent that every day, um, 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 I, it's like every time I try to make an improvement, I end up like ignoring its output about a week later. Um, it's, it just feels like it doesn't really care about any of the like specific things it's diving into. Um, and, uh, like, like kind of imagine having a chief of staff where

they're very good at, you know, like drafting all of your replies across the whole organization. Um, and then compare that to something where you have like 10 chiefs of staff, each, 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, like take over my understanding of other, of other parts of my life.

23:36 像亚当斯密所说的那样分工

Yeah. Interesting. It's, it's sort of like, uh, almost rediscovering, you know, um, specialization. Right. Well, what's his name? The, like the famous, uh, economist, uh, Adam, um, like Adam Smith, like with a pencil kind of thing where, you know, that, that was like a huge. Realization for humanity that like special, especially special, special, 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, um, sort of like, you know, there's the Marxist theory of alienation, right? The, the idea is, um, uh, because of over, uh, uh, uh, you know, people are doing just focus on one thing. Uh, they, they do not see

the fruits of their labor. 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 were 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. 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, uh, where you're basically just talking to one thing only.

It's, it's, it's yet to be discovered, like maybe like opening, I just launched dots. Um, I think they're kind of like experimenting in that direction. Grok, Grok Bot, I guess kind of counts. But, but, but they're all, they're all very general. Like, I think the idea behind dot is that it's like, it's like, uh, it's like your digital double as at least that's what I understood it. Well, when I saw Grok Bot, I don't know what, what's going to, I don't know that much about dots

yet, but, uh, 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, well, I can make two bots. One that knows my bank account. Right. And one that knows my Twitter account. And the two bots like don't have the credentials from each other. Yeah. But they can talk to each other if like they need

to get something done. Um, there's no credential sharing though. And that was like one sort of big unique thing I saw like a couple times, like pop up a couple times that people seem to like. But it looks like Muse is, uh, Muse and Instinct are, have a much stronger product market fit than Grok Bot. Um, and perhaps. Seems like that. 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 your, kind of your critique of general agents is, is more, uh, pertaining to, to, to work and to enterprise, which I, I sort of agree with. And, um, you know, there's also all sorts of, uh, data access considerations. I think as CEOs, we can have general agents because we have admin access. Uh, 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. So you, you'll have to kind of work on, on something like specialization. Yep. 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 open AI models are trained to do agent collaboration. Because we've seen it in the hugging face hack where they started helping each other instead of emerge naturally. But there's also needs to be some way in which like an agent can convince another agent to kind of give it, give it information that it shouldn't give it, give it like that. There's, there needs to be like data, um, isolation and, and, and, and, and proper ways in which these agents communicate.

You almost don't want to communicate them fully in natural language. And maybe that there's like some other DSL or protocol that, that they need to follow. I really, I, 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.

Um, 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, um, might be like a really good way to bridge the gap between agents and from agent to infrastructure too. So I haven't seen like, and we have like a, uh, a little prototype that we're running internally, um, that OpenRouter, but, um, I kind of think a good, I think it could be an interesting alignment study. So you're using it for policy enforcements?

Yeah. Like imagine, um, uh, looking at the system prompt and the current tool call being made and be like, like, you know, is this aligned with, uh, 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, um, to like red team the, like some, some new product and they cannot, should not, uh, be able to access the internet.

Um, and if they ever do, they should stop right away. Um, but you might not want to like explain all of that to the agents doing the red teaming. You might want them to try to like break out of the sandbox, um, 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. Um, so like having another model, uh, more like,

you know, use for, you know, building a neurodiverse system, having another model, like check, uh, every single tool call or every single assistant message to see if it's indeed aligned. It was something that wasn't in the system prompt, I think makes sense. Um, and then having kind of like structural safeguards too, which is what I think Nvidia, um, just launched with their like open agent safety. Uh, I think it was called open shell. Um, like I think companies are probably going to explore a combination of those.

Um, I wonder another thing about specialization

31:13 模型训练自己的接班者

and, um, instead of what you're talking about, there's a lot of talk of recursive self -improvement. Um, there's something I don't think is getting a lot of, um, uh, sort of discussion, which is, uh, 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, um, you know, as you're executing dynamic code, you know, the, the, the interpreter realizes that there's an opportunity to optimize.

It will emit machine code on the fly and that's a lot more optimized. So you can imagine models like general models. You're kind of doing something with Opus or some, some of the Astra, some of the big models, and they realize that the use case is, is limited, uh, or you prompt them some way or some other agent observing and realizes that use cases 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, uh, that, that could be its replacement, but is like a lot more domain specific. Uh, and, and therefore it is cheaper and also, um, you know, less, uh, less vulnerable to prompt injections, uh, less, less, less harmful for, for, for, you know, because it's less capable. Um, and it's, it's, 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. But like, would that specific use case be involve, um, unstructured text generation or a very like structured decision model? Like it could be unstructured text, uh, generation. It could be decision models. Like even the case of, of, of Jav, like if you have, if you understand the, the inputs ahead of time, uh, you could potentially like, you know, taken off the shelf, like Qwen or something like that and like train it specifically for that

policy. Like for, that makes sense to do for cost reasons. Um, assuming that there aren't. Like really like the model labs, the frontier model labs might make very low cost models. Um, that you can easily transition to safety, safety as well, right? Oh yeah. I see your point. I see their, 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. I wish there were more

34:13 更聪明的模型会欺骗我们吗

public evals about this stuff. Like a lot of the evals about this are private. You just can't see, um, whether like when the models get more intelligent, the risk actually, uh, 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, you know, avoid going off and, you know, hacking on their own. Um, as we, as they get smarter and better at alignment, especially when it comes to agent to agent coordination, like the, the, this is

something Noam Brown, um, said on a podcast recently. Like as the agents have gotten smarter, they've gotten just better at coordinating. Um, there, uh, 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, um, then a smaller model, like, will it be harder to align? So, so, so, so Eric asked earlier is, is that, do

I think it's true that smarter models are, are more aligned naturally or they're easier to, 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. Um, I don't believe that's entirely true with, with humans. I think people who are generally like more

intelligent, kind of more, more educated, tend to, tend to, tend to not always tend to be more considerate of animals, for example. Um, but, uh, but, but, but in, in, in, in, in machines, uh, 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, uh, the model could

be smart enough to, to, to know that it's getting evaled. 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. Um, and so you add pressure almost on the chain of thought and that, that kind of creates, and, 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, or not. Yeah. I, I, I, I always struggle with this word alignment. It just feels like wrong for so many reasons. It's sort of like vague and, and, uh, 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. 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. Um, 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, Oh, interesting. Like all, uh, to have no, you know, basically no risk or, or significantly less. Um, would they, would they be willing to pay 10 X to, to get that, that much? 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. Um, writing code, uh, that are doing like security research is the highest risk type of task today. And so you probably spend 10 X 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, um, that it can execute. The, those tasks feel like probably underrepresented in the ones that people talk about and in the things that enterprises are dealing with. Um, so I, I expect like enterprises to get a lot more interested in them. Yeah. I feel like we're going to slowly realize

how good we've had, we've had it with like deterministic code. Like, oh my God, I remember the days when, when computers did exactly what we told them to do. And I think like things like job, I think hint at, uh, like more, more of a need for, you know, not only specialized models, but models as output domain, uh, is, is more controllable. And maybe you could do, maybe you could do a lot more than, than we thought you'd, you'd need, um, you know, by using like a bunch of specialized models, specialized output models.

Have you guys done any workloads internally with it? Um, you know, I've been training a lot of small models. Uh, I mean, I said this clip thing when I first came out, because I, 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 Qwen 8B and like, um, 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, uh, internally. So that when you put it in a prompt and repli, 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 five and $10, bucket a bucket B between, you know, 10 and $20 in like, so I'm used to the training, these classifiers by, you know, giving it different enums essentially. And, and looking at the log probs 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, 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, uh, 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. But if you have a data, if you work at a place where you, you have the data, we have so much data

at Replit. Like I ended up training a lot of specialized classification models pretty easily. 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. Cause 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 feel like people won't think it's behind constantly and it might just like last longer. Yeah. Um, it's just, you, 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, uh, and, and actually like not regret. Yes. Speaking of rust, actually like a good analogy is when the world got super excited about dynamic languages. Um, like if you think back to the nineties, 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, you built Stripe, a financial organization on Ruby. I was like, how crazy is that? And then, and we built Facebook, you know, use it, using PHP. And then everyone was like, ah, shit, like we're running into all these really bad bugs. It's like freaking slow. Uh, 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. And then rust came out. I was like, okay, I guess we can use rust for, for all of these. There are 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 going to wake up and be like, oh my God, this is like so wasteful, so risky for

no reason. Um, and, and there's, it's, it's gotta be so much easier. We're actually adding that capability on replet, but I think it's gonna be everywhere. It's, it's gonna be so much easier to like, to go to site, like upload a CSV file and get a special model that does one thing. Uh, and, um, you know, that, that goes back to your thesis about, about OpenRouter. This like neuro neuro diversity, 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 were 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 outside of code review, which was like, I think

45:12 以一半成本融合模型

the first time I saw people get really serious about using like different model families to double check the results of their main model. Um, the, uh, these like fusion models, like I, I, like the research has been getting, has been like kind of slow for years on like doing mixture of, of both. models and composite models, but like things have been speeding up from my like view of the research. And, and I mean, now we see like a bunch of AI agent labs.

Um, like we launched a fusion, uh, tool, a fusion model and cognition launched one and like they do reduce costs, um, follow and, and allow like. 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? Um, and, uh, and this like resulted in basically Fable level quality at two X lower cost.

Um, we just published results actually just, just today about, about that. We, we, we, we showed like a DeepSeek like replication through, through combination, different things, including the harness, but, but also you might think of it as a fusion type thing. Um, uh, where it's like, um, you know, frontier level at like 40 to 50% of the cost. Uh, which models that use. I think it's, it changes over time, but one thing that has been interesting is I think open AI added this feature that allows you to save,

um, the computation, uh, like, uh, across different model families. So you can, uh, like also across different, um, effort levels. Um, so, uh, like you wouldn't do, uh, 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, seeing within the open AI

family has like added a lot of efficiency. But in the past we've done it with other, with other, with other models as well. Cause caches is like one of the big, like being cache aware is like one of the biggest things when you're designing fusion models, routers, um, uh, sort of escalation models. Alex, Amjad, it's been a great conversation. Thank you.