Lenny's Podcast: Product | Career | Growth
00:30:54 · Dianne Penn, Lenny Rachitsky
On the second case, it would've been a failure on, let's say, search or knowledge insertion, search synthesis, or it could be something around alignment. And so bring that level of detail to researchers coming up with, is this a big enough problem? Figure out things like evals to then describe what we've improved it. Those are the levels of actionability, and it's the day-to-day language of their researchers. And so we try to stay very close to how to bring that in an actionable manner between users to the core model training and the research development loop. [00:31:36] S1: I was talking to someone the other day about how it feels like AI research is the place to be now if you want to be very successful in life. What does it take to become a really successful researcher from which you can tell? Not everyone's brain is going to work this way, but just say people are like, "Hey, I want to explore this career path." From what you've seen, what does it take to make it there? [00:31:59] S0: Researchers generally, or research and product managers working with research, or both? [00:32:04] S1: Let's do both. [00:32:04] S0: Yeah. [00:32:05] S1: But the researchers, PMs working with researchers also going to be very successful, but it feels like everyone's trying to poach all the top researchers across every company. So I know you're not an AI researcher, but just from what you've seen, just what does it take to make it in that career path?
00:32:05 · Lenny Rachitsky, Dianne Penn
[00:32:05] S1: But the researchers, PMs working with researchers also going to be very successful, but it feels like everyone's trying to poach all the top researchers across every company. So I know you're not an AI researcher, but just from what you've seen, just what does it take to make it in that career path? [00:32:22] S0: Yeah. I think a lot of the most successful researchers and research leadership at Anthropic are folks who are really strong first principles thinkers about problems. They reason through problems really well, who are just passionate about their research area and have a bold description of what that could look like. And then who are actually close to the details. And so our leadership, our chief scientists, our heads of fine-tuning and RL, folks are actually really close to the training runs and actually look at things like how the training run is going, evals, looking at the underlying data. So actually staying really close and be excited to be in the details, I think have been a sign of really strong researchers and developing taste. And I think another piece is just their ability to think big over time and be very ambitious. Like the Dario, like we can transform software engineering. And I think going in that direction, you learn so much. You have to shoot for the stars in many ways across your ideas, I think, in order to be a successful researcher.
00:32:22 · Dianne Penn, Lenny Rachitsky
[00:32:22] S0: Yeah. I think a lot of the most successful researchers and research leadership at Anthropic are folks who are really strong first principles thinkers about problems. They reason through problems really well, who are just passionate about their research area and have a bold description of what that could look like. And then who are actually close to the details. And so our leadership, our chief scientists, our heads of fine-tuning and RL, folks are actually really close to the training runs and actually look at things like how the training run is going, evals, looking at the underlying data. So actually staying really close and be excited to be in the details, I think have been a sign of really strong researchers and developing taste. And I think another piece is just their ability to think big over time and be very ambitious. Like the Dario, like we can transform software engineering. And I think going in that direction, you learn so much. You have to shoot for the stars in many ways across your ideas, I think, in order to be a successful researcher. [00:33:49] S1: I love just this meme of just be more ambitious comes up so often now, which is so hard. It's easy to say that it's hard to actually just like how big can you think, and how that's so much of what AI now unlocks, just be more ambitious. [00:34:01] S0: Yeah. [00:34:02] S1: Yeah.
00:34:02 · Lenny Rachitsky, Dianne Penn
[00:34:02] S1: Yeah. [00:34:03] S0: I think it's thinking through it once or twice end-to-end. And then being, I think, stubborn about the area and maybe more loose around the exact approach. It is a question we challenge ourselves with, but the technology is moving so quickly, and so how do you make sure what you're building is actually forward compatible? And so it's also actually part of, I think, the core product development loop to think bigger. One thing I ask the team frequently or how I think about when we're building a product is let's say Claude 8 comes around. What changes in what users do? And then what does that mean for how you're building today? Is it going to be forward compatible to that experience? So just grounding, I think being ambitious is very broad. And so trying to ground it in some ways of describing that.
00:35:13 · Lenny Rachitsky
[00:35:13] S1: And also, yeah, everything heading in a direction that all is cohesive and makes sense versus just ambitious in a completely different direction. Speaking of ambition and Claude 8, Fable/Mythos recently feels like hit this very new tipping point with models where it used to be you have an awesome model, release it. Hey everyone, welcome. Opus 4.5 is out. Everyone can use it. Mythos went in a very different direction. We got blocked. There was a lot of scrutiny, a lot of concern about what it was capable of. All the companies had to go make sure it wasn't going to hack into all their systems. And it feels like now every model, because they continue to get better, will now have a lot more scrutiny and there'll be more restrictions on who can use them, which feels like a big deal. How do you think about that? How does that change the way you operate?
00:36:02 · Dianne Penn
[00:36:02] S0: I'm going to maybe leave the policy and the export control side to folks on that and work on that. I think the product question and how we interact with these internally is, I think as you mentioned, as frontier models become more capable, the safeguards and the ways of red teaming and testing and the pre-release process also needs to evolve and adapt quickly to address that. And so one example is before Fable models, we didn't have as strong of, let's say, fallback UXs and systems because our goal was to make sure that there is asymmetrical benefit for this technology and to minimize the downside or a severe risk of it. And so we ended up building fallback systems so that users will still get a great response from Opus 4.A immediately. And so I think there's a piece around as we evolve and improve safety systems, how do we continue to develop and deliver great user experiences? I think there's more that we can do on both sides. And so you'll see us innovating, improving on what we call now the model safeguards package more and more in the coming weeks and month.