The Exchanges

Q The naming schemes in AI are, are something else. So you, you folks have a, a simplified version in some sense. Do you want to mention any of the highlights from four that you think are especially interesting or, you know, those things around coding and other areas? We'd just love to hear your perspective on that.

A By the benchmarks, four is just dramatically better than any other models that we've had. Even four Sonnet is dramatically better than three seven Sonnet, which was our prior best model. Some of the things that are dramatically better are, for example, in coding, it is able to Uh, not do it, uh, sort of off target mutations or over eagerness or reward hacking. Those are two things that people were really unhappy with in, in the last model where they were like, wow, it's so good at coding, but it also makes all these changes that I definitely didn't ask for. It's like, do you want fries and a milkshake with that change? And you're like, no, just do the thing I asked for. And then you have to spend a bunch of time cleaning up after it. The new models, they just do the thing. And, uh, and, and so that's really useful for professional software engineering where you need it to be maintainable and reliable.

AI assessment note: “Some of the things that are dramatically better are, for example, in coding”

Q earlier question about, let's say like model self-improvement and tell me if I'm just like missing options here, but if you were to stack rank the potential ways Models could have impact on, you know, the acceleration of model development. Do you think it will be on the data side, on infrastructure, on like architectural search, on just engineering velocity? Like where do you think we'll see the impact first?

A It's a good question. I think it's changing a bit over time where today the models are really good at coding and the bulk of the coding for Making models better is in sort of the systems engineering side of things. As researchers, there's not necessarily that much raw code that you need to write, but it's more in the validation coming up with what surgical intervention do you make and then validating that. That said, Claude is really good at data analysis. And so once you run your experiments or watching the experiments over time and seeing if something weird happens, we found that Claude code can be a really powerful tool there. In terms of driving Jupyter notebooks or tailing logs for you and seeing if something happens. So it's, it's starting to pick up more of the research side of things. And then we recently launched our, uh, advanced research product, and that can not only look at external data sources like crawling archive and whatever, uh, but also internal data sources, like all of your Google drive. And that's been pretty useful for our researchers figuring out is there prior art? Has somebody already tried this? And if they did, what, what did they try? Cause you know, no negative results are final in research. So trying to figure out like, oh, maybe there's a different angle that I could use on this, or maybe there is some like doing some comparative analysis betwee…

AI assessment note: “today the models are really good at coding... in sort of the systems engineering side”

Q you can optimize against, or an environment that you can optimize against. In the context of a lot of other aspects of human endeavor, that, that seems more challenging. And you folks have thought about this so deeply and so nicely. I'm just sort of curious, you know, how do you extrapolate into these other areas where the, the ability to actually measure correctness in some sense is more challenging?

A For, for areas where we can't measure correctness and the model doesn't have more taste, Then it's an execution ability. Like I think Ira Glass said that your vision will always exceed your execution if you're doing things right as a person, but for the models, maybe not. So I guess first figuring out where you are in that turning point in that, in that trade off and see if you can go all the way up to that boundary. And then second preference models are the way that we get beyond that. So having a small amount Of human feedback that we really trust from human experts who are not just making a staff judgment, but really going deep on why is this better than that one? And did I do the research to figure it out? Or in like a human model, centaur model of like, can I use the model to help me come to the best conclusion here? And then it'll hide all the middle stuff. I think that's one way. And then During reinforcement learning, that preference model represents the sort of aggregated human judgment.

AI assessment note: “preference models are the way that we get beyond that. So having a small amount Of human feedback”

Q That, that sounds expensive, right? In terms of both scaling compute, like reasoning tokens here, and then also just like, you know, all the tool use you might want to constrain in certain ways. Does CloudFour make decisions about how hard problems are and how much compute to spend on them?

A If you give Opus a tool, which is Sonnet, It can use that tool effectively as a sub-agent. And we do this a lot in our agentic coding harness called Cloud Code. So if you ask it to like look through the code base for blah, blah, blah, then it will Delegate out to a bunch of sub agents to go look for that stuff and report back with the details. And that has benefits besides cost control. Like latency is much better. Um, and it doesn't fill up the context. So models are pretty good at that. But I, I think at a high level, when I think about cost, it's always in relation to how much it would have cost the human to do that. And almost always it's like a no brainer, right? Like software engineers cost A lot these days. And so to be able to say like, oh, now I'm getting like two or three X the amount of productivity out of this engineer who was really hard for me to hire and retain. They're happy and I'm happy. And yeah, it works well.

AI assessment note: “If you give Opus a tool, which is Sonnet, It can use that tool effectively”

Q Maybe a meta question of how do you think about competition and the provider landscape and how that turns out?

A I think our company philosophy is very aligned with enterprises. And if you look at like Stripe versus Adyen, for example, like nobody knows about Adyen. But at least most people in Silicon Valley know about Stripe. And so it's this like business oriented versus more consumer and user oriented platform. And I think we're, we're much more like Adyen that we have much less mind share in the world and yet we can be equally or more successful. So yeah, I think our, our API business is extremely strong, but in terms of what we do next and our positioning, I think it's going to be Very important for us to stay out there. And because if, if people can't easily kick the tires on our models and our experiences, then they won't know what to use the models for. Like we're, we're the best experts on our models sort of by nature. And so I think we're going to need to continue to be out there with things like cloud code, but we're thinking about how do we really let the ecosystem bloom? And I think MCP is a good example of, of that working well, where a different world that sort of like the default path would have been for every Model provider to do its own bespoke integrations with only the companies that it was able to like get bespoke partnerships with.

AI assessment note: “we're much more like Adyen that we have much less mind share”

Q what is considered an acceptable question, right? So examples of that, that I'm not saying are model specific, but societally sometimes cost flare ups is asking about human IQ or other topics where there is a factual basis for discussion. And then often those sorts of things tend to be censored, right? And so the question is why, why would a foundation model company Delve into some of those areas.

A On things like questions about IQ, I'm not up on the, on the details of that enough to comment, but I can talk about our RSP. So RSP stands for responsible scaling policy, and it talks about how do we make sure that as the models get more intelligent, that we are continuing to do our due diligence and making sure that we're not deploying something that we don't have the correct safeguards in place for. Um, and Initially, our, our RSP talked about CVRN, which is chemical, radiological, nuclear, and biological risks, which are different areas that could cause severe loss of life in the world, and that's how we thought about the harms, but now we're much more focused on biology, because if you think about, like, the amount of resources that you would need to cause a nuclear harm, you'd probably have to be Like a state actor to get those resources and, and be able to use them in a harmful way. Whereas a much smaller group of random people could get their hands on the reagents necessary for biological harm.

AI assessment note: “I'm not up on the, on the details of that enough to comment, but”