01:51:54 · John Coogan, Luca Ferrari, Jordi Hays
[01:51:54] S1: What And by the way, [01:51:56] S4: on that one quickly, I I I agree. We have some surprises in store for that community. I think they've been a bit frustrated over time, rightly or wrongly, but I can empathize with the direction Pimena has taken, which has been more like enterprise. And it makes sense. I mean, we'll see whether we can do something for them that they like. I don't know if that will turn out to be a success, but, yes, I do think there's potential there for sure. [01:52:28] S1: Yeah. A lot of people have fond memories. On on AI, we've seen a lot of engineering organizations sort of like take out the biggest hammer Wait. [01:52:38] S0: You talking about Artificial advanced Italians or artificial intelligence? [01:52:42] S1: Artificial intelligence. I know that you deploy a lot of advanced Italians, but I want to know how you're deploying artificial intelligence. Obviously, your operating team is fantastic. But I imagine that your operating team is also fantastic at making sure that you're not token maxing and overspending that you're if you are using a coding agent, you're using it judiciously. What have the learnings been? Have you been on a as much of a roller coaster ride as some of the bigger tech companies? We've seen the headlines out of Uber where they spent so much token max and they spent their whole budget then they had to pull back. There's been these back and forths. What's your journey been like at Bending Spoons? [01:53:24] S4: Yeah. Totally. So but by the way, we have more people in The States than in Italy.
01:53:24 · Luca Ferrari, John Coogan, Nick Grossman, Jordi Hays, Shubh Sinha
[01:53:24] S4: Yeah. Totally. So but by the way, we have more people in The States than in Italy. [01:53:28] S1: Think Okay. [01:53:28] S4: Because we're based in Italy. It's a little bit like it's it's unusual and so it's cool that in the Italians and whatnot. And I'm I'm Italian, but Anyway, we've got a bunch of Italians. That's true. But not just Italians. [01:53:37] S1: Sure. Just as many Americans. That's true. Fair. [01:53:40] S4: In any case, yeah, AI I mean, almost nobody knows that my cofounders and I had an AI startup in 2010. Wow. Massive failure. And we raised a million dollars, which at the time felt like a lot today. It's basically, you know, like [01:53:54] S7: nothing. No. [01:53:55] S0: That was a lot. So Demis Yeah. In 2010 raised 2,300,000 at a five cap. Yeah. Like that's and that's like the most elite AI scientists in the world. That's at the time. [01:54:05] S1: Yeah. That was the market. It [01:54:07] S8: did much better than we
01:54:07 · Shubh Sinha, Luca Ferrari, John Coogan, Jeffery Liu
[01:54:07] S8: did much better than we [01:54:08] S4: than we did at the time on AI. I'm sure he's more confident on AI. We I mentioned it because it's certainly something Michael Funders and I have been passionate about and try to use as much as possible in the business. Yeah. We've had it as part of our proprietary technologies for almost a decade. I think 2018, certainly, we were using machine learning. Yeah. Yeah, we push it pretty hard. We actually don't have any limitations on using AI, the company. We encourage everybody to push the envelope there, but we have very limited cost. And the reason why we have very limited cost is that we have developed a ton of models in house. It turns out that if you have very good people, can build narrow purpose models that do one thing really well. They're pretty dumb in general, but they can do that one thing pretty well. Mhmm. And they're very cheap. They're self hosted, basically. [01:54:51] S1: Mean Sure. [01:54:51] S4: Almost free. [01:54:52] S2: Yeah. Just
01:54:52 · Jeffery Liu, Luca Ferrari, John Coogan
[01:54:52] S2: Yeah. Just [01:54:52] S4: a little. And then, you know, there are good open source models you can fine tune, combine. Again, they may not be as good as the frontier models, but if you use them intelligently, we have built our own AI orchestrator that will help us make sure that for each task we're using the models that are optimal in terms of cost quality. We don't end up using the frontier models for a whole lot. Yeah. Just either the more complex stuff or supervision of more dumb models. With that, we we are pushing pretty hard on AI where it lost, you know, as per our disclosures, are over 90% of our code being written by AI and pretty modest expenditures. So we're pretty happy there. [01:55:31] S1: That's yeah. Yeah. Yeah. That's fascinating. What how do you see the the footprint expanding for the company post IPO? You mentioned that you have a lot of employees in The United States. Do you expect to spin up more offices, go more regional, centralized, decentralized? What's the thesis there?
01:55:50 · Luca Ferrari, John Coogan
[01:55:50] S4: I think we'll I mean, the the the trend is one where we open more offices. And we we also ultimately are are basically a key way we we do well is by by creating this core team of exceptional people, giving them a lot of responsibility, powerful tools, and a culture that helps them be a little bit better than they would otherwise be. And and sometimes we we acquire companies that have failed to attract some of the best for a long time. Not for anybody's fault. Sometimes management teams are great. It's just that if the brand is not as appealing, you just don't want attract the right applications. By injecting some of that talent, we help reinvigorate those organizations, bring back that, we call it, startup mode with high talent density, more intensity, ambition, innovation. Basically, we look to hire as many great people as we can from wherever they are. Historically, just out of, you know, like, simply where we were based and are based, it was like Milan, Italy, and then London, and now again The States, and and Madrid in Spain, Poland, anywhere we can find great people who are excited about what we do using tech to reinvent how businesses are run and take a lot of responsibility much earlier in their careers than would appear reasonable, I think that's us. Mhmm. So yeah. Probably more locations, not fewer locations going forward. [01:57:09] S1: Makes sense.
01:57:09 · John Coogan, Jordi Hays
[01:57:09] S1: Makes sense. [01:57:09] S0: What do you track average timelines from first conversation with the management team to to close? Like, I imagine there's some companies that you've acquired where you're talking for years and they're not really ready yet, but then the moment Yeah. Comes and then you can move really quickly, and then others where it's like you have an opportunity to do a deal in, you know, six weeks.