"You wrote somewhere that creating powerful  AI might be the last invention humanity ever   needs to make. How much time do we have, Ben?   I think 50th percentile chance of hitting some  kind of super..."

"What is it that you saw at OpenAI? What'd you  experience there that made you feel like, okay,   we got to go do our own thing? We felt like safety wasn't the top   priority there. The case for safety..."

"What are the odds that we align AI correctly? Once we get to superintelligence, it will be too   late to align the models. My best granularity  forecast for could we have an X-risk or extremely   bad ..."

"bad outcome is somewhere between 0 and 10%. Something that's in the news right now is this   whole Zuck coming after all  the top AI researchers,   We've been much less affected because people  here, ..."

"If you just think about 20 years in the  future where we're way past the singularity,   it's hard for me to imagine that even capitalism  will look at all like it looks today.   Do you have any advice..."

"Today, my guest is Benjamin Mann. Holy moly.  What a conversation. Ben is the co-founder of   Anthropic. He serves as tech lead for product  engineering. He focuses most of his time   and energy on al..."

"including his thoughts on the recruiting battle  for top AI researchers, why he left OpenAI to   start Anthropic, how soon he expects we'll see  AGI. Also, his economic touring test for knowing   when..."

"world and his own life and what he's encouraging  his kids to learn to succeed in an AI future.   A huge thank you to Steve Mnich, Danielle  Ghiglieri, Raph Lee, and my newsletter community   for sugg..."

"and follow it in your favorite podcasting app or  YouTube. Also, if you become an annual subscriber   of my newsletter, you get a year free of a  bunch of amazing products including Bolt,   Linear, Su..."

"The way teams turn feedback into product  impact is stuck in the past. Vague reports,   static taxonomies, unactionable insights that  don't move business metrics. The results churn,   lost deals, mis..."

"business impact and act faster. It listens to  your sales calls, support tickets, turn reasons,   and lost deals, surfacing the biggest product  issues and opportunities in real time.   It then routes..."

"caught a spiking issue and prevented millions in  churn. You can too at sauce.app/lenny. Sauce built   for AI product teams. Don't get left behind. This episode is brought to you by LucidLink,   the s..."

"videos, design assets, layer project  files, you know how painful it can be   to stay organized across locations, files live  in different places. You're constantly asking,   is this the latest versio..."

"LucidLink fixes this. It gives your team a shared  space in the cloud that works like a local drive.   Files are instantly accessible for anywhere, no  downloading, no syncing, and always up to date. ..."

"Shopify, and top creative agencies use LucidLink  to keep their content engine running fast and   smooth. Try it for free at lucidlink.com/lenny.  That's L-U-C-I-D-L-I-N-K dot com slash Lenny.   Ben, ..."

"guess there's this scary transition period from  where we are today where people have jobs and   capitalism works and the world of 20 years from  now where everything is completely different, but   pa..."

"my job seems fine. Nothing's changed. What  are you seeing just happening today already   that you think people don't see or misunderstand  in terms of the impact AI is having on jobs?   I think part ..."

"But I guess to cite a couple of areas where  I think things are changing quite quickly. In   customer service we're seeing with things like Fin  and Intercom, they're a great partner of ours, 82%   cu..."

"need to get ahead of and work on. Okay. I want to talk more about that,   but something that I also want to help people with  is how do they get a leg up in this future world?   They listen to this, t..."

"how you use the tools and being willing to learn  new tools. People who use the new tools as if they   were old tools tend to not succeed. As an example  of that, when you're coding, people are very  ..."

"Okay, so the advice here is use the tools.  That's something everyone's always saying,   just actually use these tools. It's like sit  in Claude Code. And your point about being more   ambitious than ..."

"and that's why we're hiring super aggressively. Let me take another approach to asking this   question something ask everyone that's at the very  cutting edge of where AI is going. You have kids,   kn..."

"20 years ago and I had a kid, maybe I would be  trying to line her up for going to a top tier   school and doing all the extracurriculars and all  that stuff. But at this point, I don't think any   of..."

"That doesn't come up as much just being creative. I want to go in a different direction. I want   to go back to the beginning of Anthropic.  Famously you and eight of you left OpenAI   back in the day..."

"safety wasn't the top priority there. And there  are good reasons that you might think that if you   thought safety was going to be easy to solve or if  you thought it wasn't going to have a big impac..."

"and now that exact technique is working and  many others that we have been thinking about   for a long time. Yeah, fundamentally it comes  down to is safety the number one priority? And   then somethi..."

"do you do a refusal that doesn't shut the person  down, but makes them feel like they understand   why the agent said, "I can't help you with that.  Maybe you should talk to a medical professional,   ..."

"That's right. That's right. And from a distance,  it might seem quite disconnected, like how is   this going to prevent X risk? But ultimately it's  about the AI understanding what people want and   n..."

"that we've decided are good. And this is also  not something that we think as a small group of   people in San Francisco should be figuring out.  This should be a society wide conversation. And   that..."

"like space operas where humanity is a multi  galactic civilization has extremely advanced   technology building Dyson spheres around the sun  with sentient robots to help them. And so for me,   coming..."

"It's interesting because you guys put out more  examples of your models doing bad things than   anyone else. There was I think a story of an agent  or a model trying to blackmail engineer. You guys   ..."

"I mean, I think part of why we publish these  things is we want other labs to be aware of   the risks. And yes, there could be a narrative  of we're doing it for attention, but honestly   from a atten..."

"You wrote somewhere that creating powerful AI  might be the last invention humanity ever needs to   make. If it goes poorly, it can mean a bad outcome  for humanity forever. If it goes well, the soone..."

"a forecast that's pulled out of thin air. It's  based on a lot of just hard details of the science   of how intelligence seems to have been improving,  the amount of low hanging fruit on model trainin..."

"about what that would mean from a individual story  standpoint. If the amount of goods and services in   the world is doubling every year, what does that  even mean for me as a person living in Califo..."

"prior on that is updating to be less likely. In the optimistic world, we're basically done,   and our main job is to accelerate progress and to  deliver the benefits to people. But again, I think   ac..."

"absolute best to make sure that that's true. Wow. What fulfilling work. For folks that are   inspired with this? I imagine you're hiring for  folks to help you with this. Maybe just share that   in ca..."

"Awesome. Even if you're not working  directly on the AI safety team,   you're having an impact on moving things in the  right direction. By the way, X risk is short for   existential risk. In case fol..."

"does it pass the linter? Things like that. That  also could be included in RLAIF. And the idea   here is that if models can self-improve, then  it's a lot more scalable than finding a lot   of humans...."

"an empirical company. We have a lot of physicists  like Jared, who's our chief research officer who   I've worked with a lot, was a professor of Black  Hole Physics at Johns Hopkins, and I guess he   ..."

"into and yet they're finding ways around it. We've got to start using parallel universes for   some of this stuff. I guess so.   Okay, I want to zoom out and talk about just  Ben, Ben as a human for a..."

"density. One of the things I love the most about  our culture here is that it's very egoless. People   just want the right thing to happen and I think  that's another big reason that the mega offers  ..."

"that time has been when I started the labs team  about a year ago, whose fundamental goal was to   do transfer from research to end user products  and experiences. Because fundamentally I think   the ..."

"see really cool stuff coming out soonish. Yeah,  just leading that team has been so fun. MCP came   out of that team and Claude Code came out of that  team. And the people who I hired are like combo, ..."