The Moments That Matter from Cerebral Valley London

0:00This episode is brought to you by Forethought. Most companies build their customer experience in pieces. Sales in one system, support in another, and onboarding somewhere else. Forethought brings it all together. Forethought is an AI system made up of advanced agents that handle sales, onboarding, support, and retention. Each team manages its own agents. The customer sees one unified experience. Forethought powers over a billion interactions every month for brands like Scale AI, Cohere, Airtable, and Upwork. Learn more at forethought.ai.

0:39Welcome back to America. I'm here with Max Child and James Wilsterman. Hey, guys. Hello, Eric. Hey, Eric. We had a fun European adventure, what, a couple weeks ago with the Cerebral Valley AI Summit in London. Yeah, we had a blast. I don't know. What did you guys think about it? I thought it was a tremendous experience. I mean, it really drove home how exciting the AI scene is in Europe, which I think, you know, in the sort of mobile and web eras, maybe a scene a little bit beleaguered compared to America. But I felt like AI startups and big companies are totally popping off there. I mean, we had Granola, we had Lovable, we had Synthesia, PhotoRoom.

1:21And then, you know, on the big company side, obviously, we had Wave and Uber, which was super exciting. And so, well, in classic Cerebral Valley, I feel like the buzzy startup founders are the real celebrities. You know, it's like, I mean, people are excited about Dara and we had Dylan Field at Figma, which, you know, what was like a week away from filing its S1. But yeah, I feel like the lovable guy and the granola guy, you know, were definitely sort of people of interest at the event. My spidey sense for startup heat was very tingly around the lovable guy. He was Anton. I interviewed him. But even just walking in, he was being surrounded by investors and partners and other startups.

2:02And you could just get the vibe. He's a little mini celebrity in the AI scene. He's a unicorn now. He's a unicorn now as of today or yesterday. How many coding AI-adjacent companies are unicorns now? We've got Lovable, Replit, Cursor, Windsurf. It just goes on and on. And the narrative moved so fast. Is it good to be a coding startup now? Or are we already like, oh, man, cost of goods sold and everything are out of control? When Windsurf was falling apart and luckily they were able to save it. I don't know. Since Cerebral Valley, though, we've had people like bouncing between Anthropic and back and to Cursor and back.

2:45And we've had this whole saga of Windsurf getting chopped up. I don't know if I had to stick my finger in the air right now. I feel like the mood is it's good to be a model again. I mean, you've got a big sentiment that Anthropic and Claude are key to coding and you sort of rather be Anthropic than Cursor. And obviously, we're recording this on Thursday. It's coming out on Friday. I mean, OpenAI just did a live stream about agents and seems like, oh, we're figuring out this agent thing. I don't know. You know, they're once I think they're trying to be supportive of MCP, you know, the sort of generic open source interface across agents.

3:25But then they're making the announcement that like, oh, that thing you were trying to build a company around. Yeah, we do. We do that, too. So I guess my take at the moment is models back in ascendancy, which you'd hope so, given their valuations. yeah i mean i think obviously we've had this debate a million times but i think ultimately it is good to be the company that has put you know five to ten billions of dollars to work on something rather than the company so yeah yeah yeah then the company that is slightly repackaging the thing the other guy did for five to ten billion dollars like i mean there are obviously i think application use cases here but you know we've been bullish on applications yeah oh no i I am bullish on applications, but I do think, you know, anything that's sort of close proximity to the model itself, which I think a lot of these coding startups are very close proximity to the model itself in many ways, is a dangerous place to be because, you know, Dario and Sam have got their eyes on that space.

4:23It's like you don't want to be the most popular application that makes the most money because then you're squaring the targets of the foundation model labs. it's like you kind of want to be like making good money but not like the number one use case of LLMs we're going to get into clips so this episode is really we're going to engage with the key moments at the cerebral valley AI summit which I think will be fun well you can go watch them all on YouTube and I think they're all still pretty current even though AI moves fast but we're bringing you our favorite clips so that if you don't have the time you can get the best of what happened there i wanted to say two different things one on london i think one thing i really liked about the event is that we took a sort of like ai is global we're not getting dragged into like which city is the best i mean obviously london has deep mind and there are like good companies in europe but it's like yeah it's a global world all connected on the internet there are great people everywhere like i think like half the audience was london and half came from elsewhere probably like i don't know a quarter u.s and a quarter rest of europe i don't know Yeah, a decent amount of like France and Sweden and things like that, for sure.

5:29We had a train. Yeah. Public transportation infrastructure, despite being like much better on its surface, consistently we had transportation problems across Europe. Eurostar was in trouble from France. And then I had my flight deflated out of France from an air traffic controller strike. So I don't know. They brag about public transit. There are transit and they can't keep it working. It's like like their health care system or something. It works works great for most people most of the time and then falls apart at the at the limits. The other thing is we have not officially committed to going back to London.

6:10This is it's like we need we need to sponsor bidding war between New York and London between which one we head for first. So if you have a strong horse in the New York versus London race, reach out about sponsoring. I would just say in terms of vibes at the event, the London audience was dramatically more engaged than the New York audience. I would say they were excited to be there. They stayed for the whole event. They engaged with the speakers. They wanted to chat and meet people afterwards. Well, those New Yorkers, they're grinding. They're like, I have a fucking job. like new yorkers are like i've got a happy hour at 3 45 that i've got to be at uh so the 4 30 speakers will not be fully attended before i go back to work until yeah you live in new york you're a bit of a booster it did have a little bit of like you know we had goldman we had snowflake it had a little bit of like big company tries to like embrace ai whereas i do think it turned out that Europe had some of the cooler up and coming startups.

7:12I mean, I think Rogo, you know, which we had at the FinTech event is Buzzy New York application company. I mean, there's certainly, I'm sure if we dug into like finance AI startups, we'd get more of the like application excitement in New York. I flew back from San Francisco a couple of weeks ago to be on a New York Tech week panel. And of course we, you know, from one of the New York boosters, Julie Samuels, I don't know who it is. I forget what her group is these days, but you know, it is advocacy for New York's tech scene. It was like, oh, what do you think about the New York's tech scene? And I gave the most like, oh, pro San Francisco, just like, you know, the beauty of New York to me is that people come through.

7:54It's everybody's like second favorite city, no matter where you are. But I'm not sitting here claiming that New York is the center of the tech universe. And they're like, you are not invited to next year's event. I know. I just believe I've said this a million times, but I agree with Ryan Hoover, the product hunt guy who said to me during the pandemic, Silicon Valley is online. Come on. It's on the internet. Who cares where people are? All right. Back to the present moment. A lot's happened. We're in now this AI spending war for employees, obviously with Meta, which is, I think, the most present happening.

8:32Obviously, scale deal just happened when we started the London event. Yeah, let's get into clips and let's react to some specific lines from the most recent Cerebral Valley AI Summit. In the Uber case, I mean, you came in, you did spin out Aurora. How did you... We merged ATG into Aurora. Yeah, and then became its own company. How did you decide, I guess, that self-driving wasn't there in 2017 or they this shouldn't be something to invest in or how did you go into sort of all your technologists saying oh we're just tomorrow just tomorrow and say uh i'm not convinced listen i i think um eric was a really tough decision at the time and one is we were deeply marked by the fact that we had an accident uh lost life there uh and so the responsibility that you hold in your hands when you're operating in the real world hit us deeply.

9:27I think the other circumstance that affected the ultimate judgment was that we ran ATG very separately from mainline Uber because we wanted the opportunity to partner with the rest of the ecosystem again. We don't just have Priuses on our platform. We got Ford. We want to work with the entire ecosystem. And when I went and talked with some of the other AV developers, I'd say, you know, ATG, it's in the family, but it's separate. We're going to treat you fairly. We're not going to share data, et cetera. You know, when you're having a conversation, actually, like I usually do with you and I'm saying stuff and you don't believe a fucking thing I say, pardon me.

10:09That was the response. Like, we were having conversations, but I think people didn't trust us because they also viewed us rightly so as a competitor. Right. So we had a decision to make. do we go with a proprietary kind of vertical strategy or do we truly partner with the industry we couldn't have the best of both worlds and ultimately we decided that a partnership model was the right model going forward and based on you know basically all of the leading players sans tesla but everyone else is open working with us etc i think it was the right decision to make I guess I don't really buy this explanation, honestly, because like I if you think of like Amazon today, right, they sell a lot of their own, you know, Amazon branded products on their marketplace.

11:00It pisses off their suppliers and they do it anyway because it's a good business idea. Like, I just don't really buy that this is a good enough reason not to invest in your own self-driving technology just because other providers on the marketplace won't be happy about it. Right. I mean, the premise of my question was that I thought it was skepticism on the imminence of the technology working and Uber's team particularly working and Aurora, which is what they spun off. has not really delivered on a major successful. That might be the real answer. So to me, I did think like a call on the quality of the technology was the reason.

11:41But I mean, I don't know. I take him at his face. I mean, it's interesting that his actual first answer was that they had such a negative PR situation that could they really do it? And I think it's easy to forget how much Dara was brought in to like clean up the PR situation. It's like we had the hack. And they had literally launched self-driving cars in the firestorm of their PR nightmare without the approval of like San Francisco. It was like just brazen. I mean, I do ultimately think that if Uber can sort of maintain its marketplace middleman position between riders and drivers, even if there are no drivers and it's just Waymos or in the case of Cerebral Valley London, it was Wave they were partnering with.

12:29And it's plausible that for, you know, 120 different countries around the world, there's going to be, you know, 60 different self-driving companies you need to ultimately partner with, you know, to serve self-driving cars to people in all these different locations that Uber may ultimately be the sort of aggregator of this need for a car, whether it has a person behind the wheel or not. And I think if that strategy works, I think they're in an awesome position because kind of being the middleman in this massive market of personal transportation is really great. I think what's the challenge is in San Francisco, there's this sort of Waymo brand loyalty building among the tech community.

13:08I think people like James are like, I basically only take Waymos even if they take longer and they're more expensive. And I use the Waymo app. Right. And so it's an interesting question, like for someone like James, I would pose like if Waymo's were in Uber in any city you wanted to visit and you could just instead of clicking Uber X, you click Waymo. Like, would you just go back to using the Uber app? Because like sometimes you want to use the Uber X because it comes faster and sometimes you want to use Waymo because you like it more. Or would you be like, no, I'm always using the Waymo app? I do think a lot of Uber success here depends on their ability to specifically convince Waymo that they should not deploy in every city and need to participate in the platform.

13:51I mean, you know, and they're running those experiments now. So it'll be interesting to see where Waymo thinks the power resides. Yeah, I think I would. To that point, if I know there's a Waymos in certain cities that I can only access through Uber, it'll probably switch the habit back to just using Uber everywhere, including in San Francisco, if I can access Waymos through the Uber app. All of this, though, is just, you know, pointing back to my original point, which I just think Uber would still be better off if they owned something approximating Waymo right now. Like, I just don't see a world where they are worse off.

14:33If it's Waymo quality. If it's Waymo quality. Yeah. Yeah. And who knows if they could have gotten there. Right. Amazon knows that they can deliver, you know, batteries just as well as Duracell or whatever. Yeah. And Costco and places, they pick lanes where they see where they can offer something. I'm not saying it was like a bad decision to spin this off originally. I just think disingenuous to say that he's only doing this for platform neutrality. Well, literally like a day or two after the event, there was reporting in the New York Times that Uber might be working with Travis Kalanick to like acquire a Chinese self-driving car startup.

15:10So it's very possible this is on Dara's mind that we there are all these Chinese startups that have actually caught up really well in self-driving. And we spun off the one that wasn't succeeding and cost us a ton of money. Now everybody's ripping each other off and they're all going to catch up to each other, just like the language models. And so then we have our white label version so nobody can push us around too much. Like, yeah, I think that's very plausible. and they skip a lot of the needless money losing that's hard to do on the public markets. It's like Google Pixel or something, right?

15:45Yeah. You want the flagship that just kind of shows how to do this. And you want to learn even if you're also distributing Android to other phone makers. I do think, yeah, to Eric's point, Waymo is so head and shoulders above the field right now that you would just be terrified that you're going to lose Waymo. And so you've got to balance these considerations of like uh is it gonna piss off waymo if we work with this chinese startup like what if waymo just walks away tomorrow like we better have a backup plan you know so it's it's a challenging place to be and i feel like the waymo brand is so good right now you got started in like jaguars you have none of the baggage of like bad uber rides it's just like it's such a good the mps on waymo has got to be like great relative for actual like riders yeah anyway it's I mean, it's like a quintessential technology debate question, which is just like, what are the platform dynamics?

16:38Who actually has leverage? And it'll be interesting to see it play out, but it sounds like we all agree that it would serve Uber to have some credible self-driving car technology of its own, at least to understand the fundamentals and have a little bit of a competitive edge against potentially bullying partners. All right. Well, we did Uber. Let's do the other big headliner from the event, Dylan Field, the CEO of Figma, who literally like, he seemed very busy when we saw him. And I can see why. I think they literally flipped their S1 filing the next week. So very appreciative that he made it to London at sort of the busiest possible moment for a pre-IPO company.

17:21I wanted to talk about sort of the company building piece of this for a second. I mean, you're clearly sort of moving towards going public. I'm curious, like, what your decision-making process has been. We're in this moment where Stripe, SpaceX, like, it feels like they can stay perpetually private. What's sort of the impetus to potentially move Figma towards the public markets? Yeah, I mean, I can't talk, like I told you about anything regarding that process. But I think that just the important thing that I've been making sure that I'm grounded every day, whatever happens with going public, not going public, is this is the moment to make sure that we are all in on product, whether that's me, our team.

18:13There is so much to build, so many workflows changing, so much happening. and we have such opportunity to help our customers. So that's what we're pushing for every day. It's funny to hear that played back. Like as an interviewer, him being like, I told you I couldn't talk about this, like seared in my mind in a way that watching it was like, okay, he said that and he moved on. I do think he could have been slightly less defensive. To me, it's always better just to be like, they ask questions they feel like they have to ask and I can say whatever the fuck I want. He could say something totally different.

18:46And he could have pretended like I didn't even ask the question, you know. Anyway, besides that little like bristle at my just doing my job and asking a question, it's clear that they they feel the need to have like a great sort of pipeline of products heading into an IPO. And they literally are presenting themselves as having launched what four fresh products after having four products just this year. My take is that he actually believes that they're in a disruptive moment, which we've talked about on the podcast before, where technology is changing and will affect them in many ways through AI.

19:24And then also, you know, it's a message to his team to say, like, you know, even though we are going public and this is an amazing milestone for the company, like, don't get distracted because we actually have to fight those battles and build product. yeah i think it's interesting given our experience with these um vibe coding tools over the last couple months james that like to me it increasingly seems like the bottleneck for vibe coding or ai assisted coding or whatever you want to call it is the quality of the design i would say that like the designs you get out of these tools whether it's like cursor or lovable or clawed code or whatever are just garbage like really bad and it's a sort of interesting question if like figma is extremely well positioned in that case because they have all the designers like all the good designers and you know in software essentially use figma and they can actually convert this sort of group of people who actually know how to make things that look and feel really great and polished into becoming creators and developers or if you go the other way which is say like oh no well Everyone else is going to like actually grab these designers and they're going to start understanding how to use cursor and cloud code.

20:40And then once cursor and cloud code level up on their quality of design by either training on new data or just something else, then all these designers are just going to be using cloud code or cursor, which is kind of what we experienced in the company in the last few weeks when we trained everyone on cloud code. It's like all of a sudden every, every person in the company, whether in design or product or data or marketing can, can now build apps. they all just look like garbage because the models aren't there yet. I think Figma is in a funny situation, launching Figma Make, which is their sort of lovable, you know, what have you.

21:13Vibe coding platform. Vibe coding platform is what we're calling the category. I mean, I don't know what they're calling it, but that's what I'm calling it. I mean, it puts them in a weird position where they have to sort of defend the category, which I think is still like messy and broken. I think they created like Figma. I mean, a lot of what Figma is doing is like sort of you're a marketer or you're a website builder. We're going to give you sort of a lightweight version of Figma. So it's easier for you to get in and do what you want. And so they created Figma sites for web creation. And that has like a Figma make integration.

21:48And in some ways, I think that makes more sense with Figma's like long term plan where it's like deeply integrated with their design elements. And Figma make feels a little bit like, all right, there's this sexy thing that everybody has. It's just sort of like a wrapper over models. If there's going to be so much zeitgeist around it, let's have it too. But I agree that the real opportunity for Figma is that like they have all your design assets. They have all your design assets. It's a moment when context is seen as like extremely important. And Figma has all your context on how you design things.

22:23And so clearly the more they can make this make thing much more like related to the actual design tools and assets that Figma has, it'll be much more valuable and then it'll be truly differentiated from lovable and everybody else. I mean, for the record, like, I don't know if they're going to do this, if they have the talent to do this, or if there's, it's a focus of theirs, but they have this tremendous opportunity to essentially like trained as a you know the first good model in terms of design um for you know for vibe coding or for creating software right i mean they have again basically every like good design of software you know is on figma for the last like i don't know if they can actually use that data to train but yeah i mean they should i don't know they should their lawyers should try to find some loopholes on this yeah i guess what would you be outputting you'd just be outputting the layers themselves, not the code you're saying?

23:18I don't know how it would work exactly. Maybe you create your own version of Cloud Code. Maybe you create your own version of, you know, a super enhanced version of Lovable that just makes really nice looking websites instead of pretty crappy looking websites. But I'm just saying, you know, to Eric's point about like context is king, you know, i.e. like the data on how people, you know, use these tools or create things with these tools like Figma has the greatest goldmine of design context like in the world. And I don't know how they capture that or if they're thinking about it. But to me, it feels like this this massive upside opportunity to be the the creation tool that actually makes things that look nice and work well.

23:53The other thing Dylan bristled at me about was that I said, you know, oh, when I use these like tools, I just like fucking copy and paste, you know, sub stack or whatever. And so like build me sub stack. And he's like, well, you're like, I don't know. It's sort of like you're lazy. Not everybody else does that. I don't know, but I do think like a lot of what works about these is you start with some visual design that you like and sort of replicate it. I don't know. No, I'm fully on your team here. I feel like most design is most design is is remixing, you know, to sort of the remixing is the kind of like generous version.

24:30And I think the ungenerous version would be a lot of design. I think Figma is a tool where you're trying to build it up on the pixel level. But a lot of these like text space almost it's either you're asking the thing to just arbitrarily come up with something you don't know how or you're sort of like copying and pasting, putting things in. I realize you can put in your own stuff, but I do think there's a lot of like copying that's going on. I think a lot of this is take right now is going to age like milk. And in a year, we're going to be looking back on this in a world where the foundation models have just chosen to train more on on good design.

25:07You know, it won't be like Eric having to copy paste pictures of Substack to like copy it. It'll just come out with better looking designs when you vibe code. And I think that that's the future. You can feel the AGI today, James. Yeah, yeah. I just I just think the models haven't like, you know, trained like the coding models haven't focused on this yet. And they will like it's just a matter of, you know, actually like a super short Figma case then. I mean, I'm just saying this is why I'm saying they should be in that race. Like I'm saying, like, they've got this gold mine. They've got to get in here with these other models.

25:43Like, you know, they could they could be the leader. I don't agree with the gold mine because they don't actually have the connectivity to the code. Like at some point, like the way that cloud code works is it like generates actual front end, you know, CSS. Yeah. So I just think, you know, they they have like a goldmine of great looking imagery. All right. Well, breaking news. James is short Figma. James James is shorting Figma. I predicted at the start of this episode that we would be high on models today. And that's bearing out. All right. Next up, we've got Harry Stebbings, the only VC we trusted to interview somebody.

26:25It helps that he has a podcast of his own, so he knows what he's doing. He interviewed the Granola CEO, Christopher Pedregal. Here's the clip. My favorite quote is like, will the incumbent gain innovation before the innovator or the startup gains distribution? How do you think about that today? if you think about kind of an open AI's ability to do what you do in some respects and have distribution versus your ability to move fast, be very nimble, innovate? Yeah, I think my view here is that we are in, you know, when the iPhone came out and you had like the beer drinking apps and the lighter apps, like I still think we're in that stage of AI, which might sound ridiculous because whatever, ChatGPT has 500 million users and is clearly useful, But I really think that is the case.

27:13I don't think AI is that useful. I think it's today. And I think it's mostly masquerading as useful, but you end up spending a lot of time rewriting or redoing it. Another way to put that is it's going to be so much more useful in one to two years. And I think we have yet to see truly AI-native interfaces and products come out. And I think that's the whole question. I think it's not about the products that exist today. It's still like, you know, it's like the caveman days now. It's like who can get to the truly AI native experiences and whoever builds those first and gets distribution on that.

27:55That's my view. And I think we're all, everybody in the industry is pre that right now. I basically agree with that, but I'm not sure that means like bet on granola. I don't know. It's like a little bit like I sort of philosophically agree with him that there's a lot, you know, of kind of latching AI onto existing products and workflows. And even Granola, arguably, you know, is essentially just latching voice AI onto a Zoom call, essentially. And the interface is, you know, not innovative. It's like a notes app, essentially, with some folders. It's a great product for the record. I freaking love Granola.

28:29But I don't know if the the idea that like, oh, we're in the beer drinking iPhone phase. So like, ergo, just wait one or two years and then we'll see where the chips fall necessarily makes me believe in someone other than OpenAI, I guess. I mean, it's a great fundraising posture because it's like, oh, you think we're cool now? Like, think about what we'll wait till next year. I mean, I feel like, yeah, the missing piece for me was the explanation of what the evolved version of his product looks like, which maybe he doesn't want to say would have made to me that argument a lot more. Right. Yeah.