Latest / Newcomer Pod / Cerebral Valley: AI Agents Are Already Here
Transcript
- 0:00This episode is brought to you by Forethought.
- 0:03Most companies build their customer experience in pieces.
- 0:06Sales in one system, support in another, and onboarding
- 0:09somewhere else. Forethought brings it all
- 0:11together. Forethought is an AI system
- 0:14made-up of advanced agents that handle sales, onboarding,
- 0:17support, and retention. Each team manages its own
- 0:21agents. The customer sees one unified
- 0:23experience. Forethought powers over a
- 0:26billion interactions every month for brands like Scale AI,
- 0:29Cohere, Air Table, and Upwork. Learn more at Forethought dot
- 0:34AI. Hey, it's Eric newcomer.
- 0:41Welcome to the Cerebral Valley podcast.
- 0:44Our occasional detour from the newcomer podcast with me are are
- 0:50now three time hosts Max Child and James Wilsterman, the Co
- 0:55founders of Volley and the Co hosts of the Struble Valley AI
- 0:58Summit. Welcome to the podcast, guys.
- 1:00Thank you, glad to be here. Eric, happy to be back, excited
- 1:04about the conference coming up in London.
- 1:06Yeah. So we, you know, always sort of
- 1:08jump into the Struble Valley podcast ahead of our Tribal
- 1:11Valley AI summits. And we've got one June 25th in
- 1:15London. People are like, oh, isn't it
- 1:17hard to do an international conference?
- 1:19And like, like everything in startups, it's easiest to do
- 1:22hard things when you underestimate how difficult it
- 1:25is until you're like, oh, yeah, we're just doing it.
- 1:27Someone is figuring out taxes. We have an event team there.
- 1:31I saw we have the ticket prices in pounds on the website so I.
- 1:34Think exactly. Like someone was on top of that.
- 1:37So a couple fun things for this episode for the long time
- 1:40listeners. At the end of the podcast, we
- 1:43will return to the startup draft, perhaps my favorite part
- 1:48of the show. We have overtime accumulated
- 1:51startups in our imaginary portfolio and get to score keep
- 1:55how we're doing working backwards.
- 1:58Before that, we're going to dig into some of our predictions
- 2:01from the last series. This is sort of a mid year
- 2:04snapshot, right guys, in terms of the predictions that we made?
- 2:07Yeah. We made about 10 predictions
- 2:10last year in November with the idea that they would come to
- 2:14fruition within a year. So this is the mid year check
- 2:17in, no official scorekeeping needed, but we will obviously be
- 2:22competitive on our mid year check insurance as well.
- 2:25In the latter half of the show, we'll be doing our, our games.
- 2:28But I just wanted to start off, you know, we've been doing this.
- 2:32We launched the first Cerebral Valley in March 2023.
- 2:36ChatGPT had just come out and everyone was getting their heads
- 2:40around models Max and James, if you want to sort of walk me
- 2:42through, how do you think of sort of the thematic evolution
- 2:46of AI in that period? And where?
- 2:48Where are we now as we're programming for Cerebral Valley
- 2:51London? Last time we chatted we it was
- 2:55the end of last year, which feels like a long time ago in
- 2:57the AI world. But we were all having this
- 2:59discussion of if we were hitting a scaling wall and if the models
- 3:02would stop getting better. And I would say it feels like
- 3:05every two weeks since then, something amazing has happened
- 3:08in AI. Is pretty trivial.
- 3:11Has been a great leap forward. I actually think the theme of
- 3:13hitting the scaling wall aged really poorly, which we somewhat
- 3:17predicted at the time. But you know, maybe we didn't
- 3:19know how wrong we were. I mean, just some highlights.
- 3:21We had, you know, GBT 01, the sort of first thinking model.
- 3:25We had O3, which has brought thinking models into the tool
- 3:29world. We've had Gemini really take
- 3:31great leaps forward and kind of become a state-of-the-art model
- 3:33system. We just had Claude 4-3 weeks
- 3:36ago, which a lot of people consider the greatest model
- 3:37built so far, maybe the best coding model ever made.
- 3:41We've had incredible Leafs in image generation from mid
- 3:44journey, Gemini Flux, a bunch of other folks, and then of course
- 3:49we're going to this episode 2. Video generation has also been
- 3:52unbelievable. VO3 in particular, which is
- 3:55Googles new video model I think is that the first sort of truly
- 3:59realistic seeming video generation model in my opinion
- 4:02and and kind of a great leap forward there.
- 4:04You know, Alexander Wang at our conference in November was
- 4:08probably the most prominent person arguing for a scaling
- 4:12wall or a potential issue. I mean, he was talking his own
- 4:16book. You know, he's in the post
- 4:17training business. I do think there's an argument
- 4:20that a lot of the progress, some of the progress, some of the
- 4:23progress we've seen has been about post training.
- 4:26The models are a certain level of smart, but then they sort of
- 4:30talk to each other, they're corrected in certain ways and
- 4:33that's where they get more intelligence.
- 4:35That seems to be the improvement we're seeing with O3 and the
- 4:40Chain of Thought models. Yeah.
- 4:41I mean, I think, I think Alexander Wang, I give him a lot
- 4:44of credit for what he was saying at our conference in November
- 4:47because he was arguing that maybe we would see a scaling
- 4:51wall to some degree in pre training, but we wouldn't see
- 4:55performance levelling at all. And I think that's exactly
- 4:59what's happened. And to your point, like a lot of
- 5:00the gains from thinking models have come from new post training
- 5:05techniques, reinforcement learning and new a whole new
- 5:08scaling paradigm I guess. Yeah, I mean, in the end, the
- 5:12models have gotten really, really effing good.
- 5:14And whether it's pre training or post training, they go good.
- 5:16Yeah. Yeah.
- 5:17I feel like the vibe was a little more like maybe this is
- 5:20the end of AI progress, not just like pre training is over.
- 5:22And to your point, there was a distinction drawn by some of the
- 5:25people at the conference, but I think there was a maybe a more
- 5:27negative tenor going forward from a lot of folks on stage.
- 5:30And instead, I think it's been maybe the craziest 6 months in
- 5:33the history of AI. I don't know.
- 5:34It's felt like that to me. The interesting thing is, I
- 5:37guess like in January, we had that DeepSeek moment, right?
- 5:40We haven't talked about that. I forgot.
- 5:42About DeepSeek. Too.
- 5:42Yeah, yeah. Wow.
- 5:44So to some degree that was maybe echoing some of these scaling
- 5:48wall concerns because if you can get such high performance out
- 5:54of, you know, open source models that are effectively competing
- 5:58with the frontier U.S. companies, like maybe there is
- 6:01an argument there. But then I guess, you know,
- 6:05that's kind of died down a bit as we've seen both anthropic and
- 6:09open AI come out with superior models to some degree.
- 6:12What models are you guys using? Applaud for Gemini 2/5 for
- 6:16coding and then for my own personal sort of research and
- 6:19note taking and stuff. Probably O3.
- 6:21That's the same for me. I've been using a lot of deep
- 6:23research through ChatGPT. I think that's my favorite
- 6:27product maybe of the last few months.
- 6:29I don't have time. I use O3 a lot.
- 6:32I use deep research some. I don't want to blow anyone up
- 6:34here, but I I had deep research, right?
- 6:37A whole like political consulting memo for somebody I
- 6:40was trying to get to run for office.
- 6:42You know, it's like it's amazing.
- 6:44I mean, there is a world where I would have gone and paid a
- 6:47consultant to be like draw out for me, like when races will be
- 6:50available and when, when they could run.
- 6:51And it's like you just like, oh, in the morning you're having a
- 6:54manic fit and you're like, oh, let's, let's see what Chachi BT
- 6:57can do. It's it's insane.
- 6:59It's crazy. This is a very nerdy, lame use
- 7:02case, but I like to buy cheap wine that is still good.
- 7:06And there was a secret wine from a local provider and they said
- 7:10it's this anonymous secret wine that we're selling for 1/5 of
- 7:14market price. But you know, if you bought it
- 7:16at market price, it would be a 250, three, $100 wine.
- 7:19And they were like, it comes from these amazing vineyards and
- 7:22the West side of Napa and the foothills, blah, blah, blah,
- 7:24blah, blah. And I just pasted the
- 7:26description into deep research and I was like, figure out what
- 7:28the secret wine is like 20 and 20 minutes later comes back.
- 7:32It was like, obviously this is like this BV Latour 2022 cab or
- 7:37whatever, and I was like, what? I've been super impressed with
- 7:40ChatGPT multimodal for shopping. The three of us are all going to
- 7:45the same wedding in France in a month or so, and I don't know if
- 7:49you guys have looked at the required attire, but they're
- 7:53half of. It Oh my God, my wife is very
- 7:55concerned. Yeah, this hot tip is paste that
- 7:58whole thing in a ChatGPT, ask it to shop for you.
- 8:00It'll be great. My wife and I.
- 8:03This was a good idea. I mean, it's not perfect, I
- 8:05won't lie. Like how like maybe one out of
- 8:0710 suggestions are like completely off base, but the
- 8:11shopping integration, it just, it's kind of showing where
- 8:14things are headed like, you know?
- 8:15What do you mean integration? Well, because it's actually like
- 8:19searching the web, you know, doing an agentic workflow of, of
- 8:23looking for these items and then it's pulling that information
- 8:26back into Chachi BT in line in the chat.
- 8:29There's links that link out, right?
- 8:31Eventually I'm sure you'll be able to just like add that to
- 8:34your cart within Chachi BT It's a full shopping experience.
- 8:37It's not just researching. Yeah, interesting.
- 8:40He used the word. He used the word agentic.
- 8:43Yeah. OK.
- 8:44I mean, yeah, we're going to talk a lot, you know, over the
- 8:46next couple episodes in terms of how AI can get put to use.
- 8:51The topic I really wanted to get into this week before reviewing
- 8:55our games and our scorekeeping is agents.
- 8:58Like I feel like agents have been at once sort of the
- 9:02buzziest thing in the backdrop of a couple events, but and I
- 9:06never quite here. And so I guess the first direct
- 9:10question I want to ask is, is deep research an agent?
- 9:15Is O3 an agent? Like what, what?
- 9:17What is an agent these days if it's just delivering you a
- 9:21report? I buy the agent definition that
- 9:25an agent is, you know, an AI tool that can actually do stuff
- 9:29for you, right? That can go through some sort of
- 9:31series of steps involving actions and you know, quote UN
- 9:36quote tool use is sort of one of the popular phrases these days
- 9:39of, you know, using different tools.
- 9:42Currently the only tools these agents can use really are
- 9:45essentially like web search and you know, maybe pulling shopping
- 9:49links and showing pictures to James's discussion.
- 9:53But I still fundamentally think like there is a big difference
- 9:55between, you know, ChatGPT of 6 to 12 months ago where right,
- 10:00you know, you ask it a question, it gives you an answer.
- 10:01Essentially it's you know it's a text in text out engine, right?
- 10:05You put text in one side, it gives you text out the other
- 10:07side, but it doesn't go do stuff that's.
- 10:09Part of that kind of web it's. Like, yeah, yeah, yeah, exactly
- 10:12right. Even the deep research that we
- 10:14were talking about, I mean, I used to be a management
- 10:16consultant for two years, 2 horrible years, and it can do
- 10:21much better versions of what I did as a management consultant,
- 10:26you know, in a matter of minutes, it could do 5 days of
- 10:29world class management consultant level research on a
- 10:32topic, right? And I guess if you don't think
- 10:35that's agentic, I think you're a little crazy, like you're too
- 10:40high, right? It's clearly running around.
- 10:42I I've seen some people use like time is it's like if it takes
- 10:46time, how long it takes, you know, if it's all going and
- 10:48doing things and interacting with the world and coming back,
- 10:51I want to throw down a gauntlet to me.
- 10:55And this could come soon. We'll we'll be in the world of
- 10:58agents once people are letting them run wild with their own
- 11:02credit cards once, once agents are spending money without a
- 11:06human check in, that's when we've got sort of real agents.
- 11:10What do you think? So does it.
- 11:11Not count in your mind if like my agent finds me a dope pair of
- 11:15shoes and, you know, text me, hey, can I buy this?
- 11:18And I'm like, yeah, go for it. No, it needs to transact.
- 11:20It needs. It needs to do without you.
- 11:22Oh, truly? It'll be like this.
- 11:25My great. Thing that's like that's an
- 11:28agent. I mean, I understand like you
- 11:30know, you, you book flights and an agent would have like asked
- 11:33you before that. Definition I I feel like if it
- 11:36does a restaurant reservation, maybe no money changes hands.
- 11:38That's still an agent. I mean, I I just think this.
- 11:40Is it's definitely an agent. I I'm just saying that's like a
- 11:43great employee solves the problem, right?
- 11:46It's not like, oh, they come back to you and want all this
- 11:48feedback. It's like do the thing like pull
- 11:50the trigger like once we can trust to do that.
- 11:53I'm just saying that would be a landmark moment that I don't
- 11:56think is so far away. Or do you think that is far
- 11:58away? No, I, I honestly think that if,
- 12:01if you had the capability to do that already in ChatGPT, like
- 12:05James did his clothing research, probably some people would be
- 12:07like, yeah, fine, go do it. Like don't buy too many things
- 12:09without asking me. But like, I think people would
- 12:11already be down. I honestly think the company's
- 12:13reluctance is probably just mostly like they don't want it
- 12:16to go off the rails and spend, you know, thousands of dollars
- 12:18of people's money or whatever. But it's it's possible already
- 12:21in my opinion. I don't know if it would work
- 12:23yet. Like, and I think a lot of this
- 12:25ties into context how much Chachi BT knows about me.
- 12:29I mean it it's obviously, you know, starting to build that
- 12:32memory, but I don't think it knows enough without me
- 12:34prompting it or, you know, having very targeted list of
- 12:38shopping ideas for this wedding to go just start buying me
- 12:41stuff. I that'd be super interesting
- 12:43once we get there, but I don't think it's ready for that yet.
- 12:47Max just was sort of getting at this.
- 12:49I mean, stop limits I think are key, right?
- 12:51I mean, there's a degree to which algos, you know, I was
- 12:54talking to a former banker about this the other day.
- 12:55Like, you know, it's not crazy that we would let a machine, a
- 12:59computer make payment decisions on its own.
- 13:02Traders do it all the time. You just create some limits and
- 13:05checks and hopefully have people hovering.
- 13:07But like the algorithms have to move before a person could
- 13:10react. And so that's happening there.
- 13:13And so you can see with LLMS, it's like, OK, you build up a
- 13:17trust up to the, you know, $500 limit.
- 13:19And you're like, you can, yeah, it'll be interesting.
- 13:22I mean, how many chargebacks are our chargebacks can go up with
- 13:25everybody releasing power to agents?
- 13:27And you're like, oh man, we need to do more refunds when the
- 13:29agents do something crazy. I mean, I think that kind of
- 13:33gets into some of this like quote UN quote agentic web
- 13:36discussion, which is like, can we redesign the web in a way
- 13:38that enables more of this behavior?
- 13:39Because I do actually think if you're a clothing retailer,
- 13:43there probably is some kind of business model in which you let
- 13:47people buy way too many clothes and then they send most of them
- 13:50back or they cancel most of them before they ship or whatever.
- 13:52Right? Like, which obviously is along
- 13:54the lines of like a Stitch Fix or Trunk Club or some of these
- 13:57other companies which, you know, weren't wildly successful
- 13:59because I think the return fees are pretty punitive or, you
- 14:03know, and you end up with a lot of fraud and, you know, damaged
- 14:05clothes and stuff like that. But I do think there's an
- 14:07interesting question of like, could you build some sort of
- 14:10consumer commerce website where people's agents can like buy way
- 14:13too much stuff, like basically and then cancel it or return it
- 14:17or limit it or some in some way because people buying too much
- 14:20stuff by accident. And it's it's probably good if
- 14:22you're selling things like even if you have to find a way to let
- 14:25them cancel it or return it. My first reaction to what you
- 14:27were saying is, oh, this is like a workaround to allow agents to
- 14:31spend when we're really going to unwind it, but it makes the
- 14:34default spending versus not. It's like, oh, the sugar is
- 14:38good, which is which is really good for if.
- 14:40You're selling stuff. Yeah, yeah, yeah, yeah.
- 14:42James, what's the agent use case you're most excited about?
- 14:45I can't ignore coding. We haven't talked about coding,
- 14:47which I think is like the most actually valuable the.
- 14:51Real one. That's why we want to talk about
- 14:52it's boring. It's happening.
- 14:54And that's. Something actually works?
- 14:56Oh yeah, yeah, yeah, OK. AI podcasts are only about the
- 15:00future, you know. Yeah, I think, well, I just
- 15:04think that there's a lot to unpack about the future of the
- 15:07of coding. I mean, I am a CTOI code when I
- 15:11can and this agentic world has dramatically changed what I can
- 15:16do in terms of prototyping and participating in the coding at
- 15:21Volley and learning faster, right?
- 15:24I mean, you just learn so much faster about different
- 15:27technologies and tech stacks. And yeah, I think it's like if
- 15:31you're not an engineer, you maybe understand this a little
- 15:33bit or you've played around with things that are a little bit
- 15:36more. No code like lovable.
- 15:37Or figma or. Figma right?
- 15:39I mean I. Think both Lovable and Figma are
- 15:41speaking our event and they are now competitors in the the no
- 15:45code world. Yeah, both non coders and
- 15:48coders, if you've dabbled with any of this, you are, you know,
- 15:52receiving the future of like what could happen to all types
- 15:55of computer work, white collar work, if you want to call it,
- 15:59but is happening first in in the engineering space and it's it's
- 16:04pretty remarkable. I think Nat Friedman had a
- 16:06really good analogy that I can't get out of my head about agent
- 16:09tick coding. And he talks about the idea that
- 16:11you have, you know, you have this room full of interns who
- 16:16are all like junior engineers basically, right?
- 16:18And you assign each of them a task and they go off and try to
- 16:22do it. And then when they get stuck,
- 16:24they raise their hand and they say, hey, I need help here,
- 16:27come, come help me. I'm, I'm stuck with this bug or
- 16:29or this this, you know, issue the, the app's not working,
- 16:31whatever. And he podcasted, I think about
- 16:35six months ago and he was like, right now basically these
- 16:37interns raise their hands like every 5 minutes in like human
- 16:40time. Like so they do like 5 minutes
- 16:42of work and they raise their hand and you've got to go help
- 16:43them. And then they raise their hand 5
- 16:44minutes later and you know, over and over and over again, he's
- 16:46like, so you can't really have met that many of these, you
- 16:48know, imaginary interns going because you can just become, be
- 16:50running around fixing their problems all the time.
- 16:52You know, maybe only one really like because you're just
- 16:54constantly having to give them feedback.
- 16:56I do think like the frontier for like how much quote UN quote
- 17:00human work these agentic coding tools can do now without you
- 17:04having to run over and help them when they raise their hand is
- 17:06like probably somewhere in the like 15 to 30 minute range now
- 17:10where like, you know, they obviously come back to you very
- 17:13quickly. Like they iterate through their
- 17:14work very fast because they can type at a superhuman speed,
- 17:17right? They can put out hundreds of
- 17:18lines of code. But the sort of amount of human
- 17:21level work I would say in my testing that they get stuck is
- 17:24probably like, you know, some of that 15 to 30 minute range.
- 17:26I'm confused, like are you saying 15 to 30 minutes of like
- 17:29what it would take for a human intern?
- 17:31What it would take like a good human engineer like, you know,
- 17:34Yeah, I got it. You know, mid to senior software
- 17:36engineer just just hammering away at code, right?
- 17:38Yeah. Like how much code do you get
- 17:39out between them getting stuck? Right.
- 17:42Like, yeah. I would say yeah, maybe half an
- 17:43hour of like human, human work, but.
- 17:45You get it. You get it in 3 minutes or two
- 17:47minutes you get. It in a minute, you know, or 30
- 17:49seconds usually, which which is which is mind boggling.
- 17:51But like the dream is you get, you know, 4 hours of human human
- 17:55work or 8 hours of human work or eventually, you know, weeks or
- 17:58just starts fixing itself and you never go in there, right?
- 18:01I mean, you know, I spent a fair bit of time vibe coding a month
- 18:05or so ago. I do think there are some self
- 18:09driving car aspects in the sense that like the last 5% or 10% of
- 18:15a problem is very important. And like it's like, oh, it looks
- 18:19close. It looks close, but it's like,
- 18:21sure, it's really good when I'm like copying the sub stack
- 18:25design, dropping a lovable like, oh, rebuild that.
- 18:28But like at some point I feel like you've just been working
- 18:31long enough and you have some minor tweak you want to make and
- 18:35it just starts getting stuck and has no idea.
- 18:37I mean, partially that's why I'm trying to do no code and no code
- 18:41knowledge. But I do think these programs
- 18:44are really good at being enticing in the beginning, and
- 18:47then they sort of get overwhelmed as the project
- 18:50starts to expand. Maybe this project that you were
- 18:53doing, you won't be able to ship because of that last five mile
- 18:56problem or whatever. The fact that you even started
- 18:59it is kind of amazing that you did that.
- 19:02Like you wouldn't have been able to even start the project three
- 19:05months ago or six months ago or something.
- 19:06Sure. Yeah.
- 19:07So I don't know, it's just kind of interesting, like where that
- 19:10heads is. You know, I think eventually you
- 19:12get to the point where it's a finished product and then
- 19:15there's just like way more coding projects happening in the
- 19:17world, right? And then, you know, I think
- 19:19within companies that already have an engineering workforce
- 19:22like it actually can get across that that last mile.
- 19:25I think another important theme that always comes up is like if
- 19:30we froze this moment in time, how much value is there that
- 19:34people are still understanding versus depending on continued
- 19:40progress in the models that's that's revolutionary.
- 19:42Where are you guys on how much harvesting could be done with
- 19:46what would already exist versus what we're waiting on?
- 19:50I would say a lot like a lot of harvested progress.
- 19:54Like I can't even think like trillions.
- 19:56I don't know. I haven't decided of 10s of
- 19:58trillions, but let's say trillions.
- 20:00What's the measures? In value.
- 20:01Oh, like GDP. GDP value of just harvesting the
- 20:05stuff that we've done so far. But it's amazing you still see
- 20:09people who are like, skeptic. I mean, I don't know.
- 20:12I just don't want to be, you know, deluding myself here.
- 20:15Yeah. I mean, I think like
- 20:16fundamentally we've all had the experience, right?
- 20:19Where something that used to take 8 hours now takes less than
- 20:25a minute, right? I mean, like we've all had that
- 20:27experience, right? And so if you just do the math
- 20:29on that, you say, OK, I took something that used to be a 500
- 20:31minute problem and now it's a one minute problem.
- 20:34Unless you believe that thing had no value whatsoever, how
- 20:38could you not see a 500X in knowledge work in front of you
- 20:41and say, hey, maybe we haven't harvested all the value on this
- 20:45500X improvement, though this thing I just did.
- 20:48How could you not believe that that's going to have insane
- 20:50ramifications throughout our world and and insane amounts of
- 20:53value that eventually can be created.
- 20:55You know, so I don't know. I just to me, I just look at
- 20:58that, you know, 2 orders of magnitude differential we're
- 21:00already seeing on labor and think it's got to be huge.
- 21:04Here's a question for you guys. Like, I have a friend who's in
- 21:06PE and he, you know, doesn't really, he's not really in the
- 21:09tech world that much, but he, you know, tried chat GPD didn't
- 21:13really, you know, find it super valuable.
- 21:15He had never tried O3. And I just showed him what was
- 21:19possible and he, like, you know, kind of changed his whole
- 21:21opinion of what types of companies he should buy in the
- 21:24future. Wow.
- 21:26Well, I do think that's a real danger that there are, there are
- 21:29things that I'm changing about my life and medical decisions
- 21:33and lots of stuff off O3 BS. It's so persuasive that we'll
- 21:37never really be able to back out the psychological.
- 21:39And, you know, it's all right. It's like, is it having an
- 21:41effect? It's like, yeah, it's deeply
- 21:43affecting big decisions in my life.
- 21:46You know, just because it's like the thought partner, it's there
- 21:49just like, you know, if you were to have the friend that's there
- 21:51while you're soundboarding an idea, like that's going to be
- 21:54dramatic. And I think just the fact that
- 21:56it's in the loop. Wouldn't you say overall that
- 21:58it's in person giving good advice?
- 21:59That's why I keep it in the loop.
- 22:01Yeah, right. It's not like, but but it will
- 22:03be hard to know if it if it hallucinated.
- 22:06I don't know. It's like, is it gonna?
- 22:08Yeah, hopefully in a couple of years I'll chase after me in
- 22:10this shot. GPT will be.
- 22:13I told you a couple of key things that I think you took to
- 22:15heart and and now that I'm a little smarter, I'm gonna go
- 22:18back on that. Yeah, we'll see.
- 22:20Having lived through the iPhone experience of of having the
- 22:23first iPhone in 2007 and having to spend the next four years
- 22:27explaining to my friends why they had to get an iPhone and
- 22:30why they. Should probably.
- 22:31Consider getting an iPhone. I mean James, you had a
- 22:34BlackBerry until what, 2011 or something like that?
- 22:36Like 10/20/11 like. James, I'm sorry, I was the
- 22:42lager to. I had a BlackBerry for a long
- 22:44time. Yeah, and people would be like,
- 22:45well, Max, what? Why?
- 22:47Why do you, I need an iPhone. What do you do with an iPhone?
- 22:49And I'm like, well, you can like browse the Internet.
- 22:51And they're like, well, you know, I browse the Internet on
- 22:53my phone. It's fine.
- 22:54And I'd be like, well, these apps are pretty cool.
- 22:56And they'd be like what? Like the the app that makes it
- 22:58look like you're drinking a beer and I'm.
- 22:59Like, well, I'm not. Using that every day, but just
- 23:04the last point I have to make on this BlackBerry thing.
- 23:05I'm sorry, BlackBerry, which obviously is a dead company
- 23:09today. No one you know has a
- 23:10BlackBerry, right? Their sales continue to grow for
- 23:14four years after the iPhone was released right?
- 23:17So O 708-0910 eleven BlackBerry sales continue to increase even
- 23:21though we all today look back and even in the movie about
- 23:25BlackBerry, they pretended that the day the iPhone came out it
- 23:28was like P BlackBerry. Sorry guys, you missed the
- 23:31future. And I'm just pointing out that
- 23:34in the real world, sales continue to grow for the next 4
- 23:37years. And so I'm just saying that when
- 23:38you're in one of these sort of Roadrunner, Wiley Coyote chases
- 23:42the Roadrunner over the ledge moments where you take a second
- 23:45before you look down and realize that gravity is pulling you
- 23:47there. I think similarly with AI, we're
- 23:50all like, holy crap, we just went off the ledge.
- 23:52Like shit is about to get really dramatically different here.
- 23:56And you can still stand there in the air for a year, two years or
- 24:00three years or four years before gravity really hits, right.
- 24:03And and I think to your initial question, like we are in that
- 24:07period where even if nothing changed, like we're already off
- 24:11the ledge, you know, stuff, stuff is going to be
- 24:14dramatically different. No matter what happens from
- 24:16here, which I know we all believe, you know, the models
- 24:18are going to continue to get better.
- 24:20So the pace of acceleration is going to be even higher.
- 24:22It's a perfect endnote. Let's move the conversation to
- 24:27our predictions from six months ago.
- 24:30Basically, we're not going to get 2 in the weeds.
- 24:33It was a fun discussion. James, do you want to take us
- 24:37through them, What the question was, where we each landed and
- 24:40then we'll give a quick reaction and then go to the next one.
- 24:43Sure. Sounds good.
- 24:45Just to clarify also we back in November, we asked Claude and
- 24:49Chachi BT to generate these predictions for us including
- 24:53providing probability estimates of how likely they were.
- 24:57And then so our job was to take the over or the under on each
- 24:59prediction. The first one was open AI shifts
- 25:02GPT 5 with a greater than 10 trillion parameter model Max was
- 25:07the over, Eric under and I took the over.
- 25:10Thoughts. Where is it?
- 25:12GPT 5 here. Yeah, so far.
- 25:14So far I'm correct. Six months.
- 25:16I mean, I think we much hung up on they wouldn't use the name or
- 25:20maybe there was a high chance that they would abandoned the
- 25:22name. It seems like they're going to
- 25:24do it honestly right now. What the vibes are that they're
- 25:27going to kill the O series and just make 5 the overall?
- 25:32I feel super good about the name and it's shipping this year.
- 25:35At this point. I think the only thing that
- 25:37could hit the under would be the 10 trillion parameters because
- 25:40didn't Brad, like at the CEO literally say they were going to
- 25:42ship GPT 5 this year and it was going to be called GPT 5?
- 25:45Like I'm pretty sure he said that like last.
- 25:47Seems like a. So I would still smash the over
- 25:50the 10 trillion parameter. Thing is, I guess the one.
- 25:52Certainly I would buy the over now.
- 25:55I might take the under because of that 10 trillion like I oh.
- 25:59You're I'm sticking to my bet. Whatever I we're the best locked
- 26:01in. Yeah, I can't be wrong.
- 26:03I was trying to figure out like so Claude 4 Opus just launched
- 26:06and there's no comment on the parameters that I can find, but
- 26:11the best estimate I could get from Claude was 2 to 3 trillion,
- 26:15so I don't know if that's nice. Nice.
- 26:18I mean a lot of the improvements seem to be post training and
- 26:22chain. Of thought and if they're.
- 26:24Going to merge, you know, the O series with GPD 5.
- 26:28It's partially because they think they need part of the
- 26:30improvement off these reasoning models, and so maybe it's a sign
- 26:33that the parameters aren't getting as big as we thought.
- 26:36Also also it seems like adding that many parameters you know
- 26:39creates huge issues with inference cost and just serving
- 26:42them it's. Expensive.
- 26:44Yeah. Anyway, the next one was three
- 26:48or more countries enact national rules regarding AI medical
- 26:52diagnosis. Max, you took the under, Eric
- 26:56took the over, and I took the over.
- 26:58We had it at a 70% probability. You guys are crazy.
- 27:02You guys are crazy. There's no way you think in the
- 27:04next 6 months three countries are gonna enact regulations.
- 27:08That happened so far, medical diagnosis.
- 27:10Here's what I had. Oh, you.
- 27:14Found something chachi BT is saying.
- 27:16Yes, EU AI Act. Exactly, Yeah.
- 27:19UKSMHRFUS to regs online chachi BT things we're like in good
- 27:24shape all. Right.
- 27:25Well you took 70% so never forget that I get 2 to one.
- 27:28Yeah. I don't know how you're going to
- 27:29do the overall math at the end, but I'm sure whatever.
- 27:31Onward and upward. All right, Tesla full staff
- 27:34driving approved for unsupervised driving in one or
- 27:37more US state, 40% probability. Max had the over, Eric had the
- 27:43over, and I had the over. I think we're in good shape
- 27:46there, right? Isn't Texas gonna happen like
- 27:49yeah in June, supposedly. Like, yeah, in two weeks,
- 27:52correct? But it does seem like right?
- 27:54Chachibi says not yet Pilot cars on private roads in Texas.
- 27:57No public permit. It's coming to Austin in June.
- 28:00Now. I think this gets into the
- 28:02letter of the prediction, though.
- 28:04Is it a state law or city law, whatever.
- 28:08OK, we'll come back to that in a few months.
- 28:11AI will write the copy for greater than 50% of a major news
- 28:16outlets articles. 30% chance Max had the over, Eric had the over
- 28:22and I took the under. I just feel like my, my under is
- 28:27great here. I mean, nobody's.
- 28:28Nobody's. Claiming this nobody's first
- 28:30newsroom. Everybody is like, what's the?
- 28:32Eric's gonna do this by the end of the year.
- 28:35Just to go build. A new media company.
- 28:37I won't be major though. I don't think I'll I'll qualify.
- 28:40You could be major by the end of the year.
- 28:41Why not? Apparently, the Chicago Sun
- 28:43Times inadvertently ran an AI generated book list filled with
- 28:47errors, sparking backlash. Clearly not paying for 03 there
- 28:50I think. I just think we're gonna see so
- 28:52many of these things, like hallucinations, like, aren't we
- 28:54already seeing this from the Trump administration?
- 28:56Like just random things that don't make any sense.
- 29:00I certainly don't want, you know, AI to replace the
- 29:03newsrooms. I'm just expecting a lot of like
- 29:05stories over the next year about like academic papers and news
- 29:09articles just having obvious hallucinations, right?
- 29:11Because people are going to be using.
- 29:13But shouldn't we be more disturbed that the Trump
- 29:15administration is like, I feel like if Democrats were in charge
- 29:18and they were releasing government reports that were
- 29:21clearly written by AI, yeah, it would be like the biggest
- 29:23cultural story of the moment. It's like, well, the governments
- 29:26already phoning it in with Trump, it's like, oh, at least
- 29:29they're using it is. Literally what I was going to
- 29:31say. I was like, I would honestly
- 29:33prefer the AI to be doing this than the miscellaneous Trump
- 29:37administration employee. Like knows nothing about
- 29:40anything. I just want Sam Altman to give
- 29:42every administration staffer free O3 so that they have O3
- 29:46right. Reasonably intelligent fake
- 29:47reports instead of real ones. I do think this core take that
- 29:51the Internet split everybody's view and this is part of why
- 29:55originally Marc Andreessen was so pro crypto and anti AI.
- 29:59I do think AI is going to bring us potentially closer together
- 30:02where people are asking rock like is this bullshit true or
- 30:05false? I mean, it's possible people
- 30:07build crazier models, but for now, while the models are sort
- 30:11of generally in agreement with each other about how the world
- 30:14works, it could be a major force for cultural consensus over the
- 30:19next couple decades. It's basically the network TV of
- 30:22the Internet essentially, right? We're all watching the three
- 30:25major channels and they all broadcast kind of like exactly,
- 30:29but not family friendly content. And your take is that Marc
- 30:32Andreessen? Marc Andreessen doesn't want to
- 30:35bring us together. He was weird, crazy shit.
- 30:37Yeah, that's what crypto is all about.
- 30:39Like, do whatever you want, like, and AI is a source of
- 30:42conformity. But then Andreessen Horowitz
- 30:44basically saw what was winning, and it was like, we not need to
- 30:47go where the money is. And, you know, that's how I see
- 30:49the story playing out. I mean, they were resistant at
- 30:52first. Khosla led the open AI venture
- 30:55round. Anyway, keep going.
- 30:56Didn't Peter Thiel say like AI was communist or something?
- 31:00Yeah. Yeah, communist and crypto was
- 31:02libertarian or whatever. Right.
- 31:04OK, Fully AI scripted and AI rendered feature film gets a
- 31:08theatrical release 25% chance. Max with the under, Eric with
- 31:13the over and James with the under.
- 31:15This was this was Eric using his insider knowledge of pay to play
- 31:19tactics within the movie industry to to try to grab us on
- 31:23this that someones going to pay a theater chain to take their
- 31:25movie even if it sucks. Basically, make sense?
- 31:28I mean, we still have time. I had to double check this that
- 31:30it wasn't hallucination, but according to chat GVT there is a
- 31:35fully AI generated movie releasing soon, Pirate Queen
- 31:40Zheng Yi Sao, billed as the world's first fully AI generated
- 31:44feature film. Exactly.
- 31:45That's going to get a festival run.
- 31:46I gotta watch it. All right, moving on, more than
- 31:49three major smartphone OEMs chip phones with AI Co processors
- 31:55running 7 billion parameter plus models on device. 7 billion is
- 32:01pretty high. The rumor mill on Apple is
- 32:03saying they're going to be able to run 3 to 4 billion.
- 32:06So even if you believe in James's claim that the chip
- 32:09they've had in there for 12 years is an AI chip, it still
- 32:12might not be able to do a 7 billion parameter model.
- 32:14So we were specifying an Apple, Samsung, Google, Chami,
- 32:18dedicated AI Co processors that run 7 billion perimeter LLMS
- 32:23fully locally. We had the probability at 75% to
- 32:29unders Max and Eric and I took the over and I yeah, I was
- 32:33counting the the current technology as capable of running
- 32:38those types of models. But yeah, to your point, Max, I
- 32:40think 7 billion might be a bit high, right?
- 32:443 to 4 is what the rumor mills saying for this year, but we'll
- 32:47know in a few more months I guess.
- 32:48Next. Anthropic releases a model
- 32:51scoring over 90% on U Bar, the unified benchmark for AI
- 32:57reasoning, which does not exist according to our own research.
- 33:02During the podcast recording last year, Complete Hallucinated
- 33:07prediction from Claude. How are you feeling about that
- 33:10one guys? Great.
- 33:13I should, yeah. Certainly raises red flags,
- 33:16yeah. Moving on to #8 / 4 Fortune 500
- 33:22firms 5 or more cut greater than 25% of their middle management
- 33:28roles by the end of this year, crediting AI explicitly with a
- 33:3425% probability. Actually interesting.
- 33:38Here we have an over and over from Max and over from Eric and
- 33:44on under from myself. I'm feeling pretty good about
- 33:47the under. Yeah, yeah, look, I'm.
- 33:49Gonna do the research on this. We're only asking for five firms
- 33:52to cut 25% of only middle management.
- 33:54So that's a pretty that's a low bar.
- 33:56According to my research with ChatGPT, this has not
- 33:59materialized. Many large companies are
- 34:01experimenting with AI and none have reported cutting 1/4 of
- 34:05their management. Yeah, maybe we overestimated,
- 34:08first of all how much we thought it would give them air cover for
- 34:11all sorts of things. But yes, exactly.
- 34:13Maybe we didn't want to get right into the AI narrative.
- 34:17If we got a tariff induced recession, this actually might
- 34:19happen so. OK #9 Deepmine and Google
- 34:22discover a new drug that clears phase one trials within 2025.
- 34:27We gave that a 20% probability and Max took the under, Eric
- 34:33took the under, and I took the under.
- 34:35I think all looking pretty strong here.
- 34:38And D mine spun out isomorphic, which would be the start up.
- 34:42I think that would potentially do this.
- 34:43So there's a there's a chance that even if it happens, we can
- 34:47all claim technicality that it doesn't.
- 34:49But I I think it's, it's not looking likely, right?
- 34:52They have to be in phase one trials already.
- 34:55Yeah, I mean I. Think we?
- 34:56Yeah, I think that they are planning to enter trials by the
- 34:59end of this year, so unlikely to have completed phase one trial.
- 35:04And lastly, we have the international AI treaty with
- 35:09greater than or equal to 15 signatories, including three of
- 35:13the US, China, EU and UK. 50% probability we all took the
- 35:19under. Seems like a good bet so far.
- 35:22Yeah, the AI believes too much in human institutions.
- 35:25Right, because US, China both seem unlikely, right?
- 35:29Is Europe getting EUUK viable getting the US which under Trump
- 35:35is now like no AI regulation and China which is we do what we
- 35:40want. The fact that the EU and UK get
- 35:42separate credit here doing a lot of work, but three of the four
- 35:46seems high do. You know, do you know what the
- 35:48score is? So just pulling together the
- 35:50scores, I asked Chachi PT to create a scoring system.
- 35:54It came out with a formula inspired by Breyer style scoring
- 35:59which I had never heard of but. Yeah, that's how close you are
- 36:01to the probability. Yeah exactly.
- 36:03Yeah, seems like a good scoring system.
- 36:06It gives 3rd place to Max Dinged for betting against AI film
- 36:13better calibrated on his conservative bets like hardware
- 36:16and policy. 2nd place Eric solid instinct, slight overconfidence
- 36:21on a few misses and in first myself.
- 36:25Great balance of bold but accurate calls.
- 36:28Thank you, Chachi I. Love it, love it.
- 36:30I think I think this AI film take is complete bullshit.
- 36:33So I need to I need to score adjusted for that.
- 36:38This is just a check in, no medals awarded yet, but I will
- 36:44take the pole position and see you guys in a few months where
- 36:49we can do the final tally. All right, let's do our fantasy
- 36:53draft. Max, you want to talk us through
- 36:56the game and then we'll get into our picks.
- 36:58Yes, we invented an ingenious game based on fantasy football
- 37:04that allowed us to draft teams of startups into imaginary
- 37:09rosters. We've done two different drafts.
- 37:12We did one about a year and a half ago and one about six
- 37:14months ago. We restricted the draft list,
- 37:18the draft board as it were, to companies that had raised over
- 37:21$100 million at the time. So if there's obvious omissions
- 37:25that come up in your mind, it's probably because they hadn't
- 37:27raised 100 million at the. Cursor Cursor.
- 37:29Being the most. I think like there was some like
- 37:32where we've that we didn't include because.
- 37:34Yeah, We also excluded specific like chip based companies,
- 37:39Chinese companies, I don't know for robotics, I can't remember
- 37:42healthcare, we anything we thought we were even Dumber than
- 37:46normal about we left off the list.
- 37:47So we all drafted teams. We did a snake draft.
- 37:51Most notably, the first draft involved a discount that we had
- 37:57one person had to take for getting the first pick because
- 38:00the first pick was very obvious. It was Open AI.
- 38:02Eric paid, I believe, $75 billion in handicap to draft.
- 38:07Opening auction 1st and I and I really good about it.
- 38:10Aged pretty well since they're valued at 300 million right now
- 38:123. 100 billion, would you say? 300 billion.
- 38:15I'm sorry. Yeah, like coughed.
- 38:16Yeah, it doesn't matter. And I will say before we say our
- 38:21teams, I have not yet had the first pick in any draft.
- 38:25So I just want everyone to remember that when making our
- 38:27teams, OK, I will go through my team first, which I will admit
- 38:32upfront is in last place right now.
- 38:34All right, my team, Databricks, my star worth $62 billion,
- 38:38Cohere AI, model company, modular AI, language company
- 38:42Scale AI, Sierra AI, Sakana AI and Hebia.
- 38:47And if you know all of those names, you are far too online.
- 38:51I'm bullish on Sierra. I think that's, I think scale.
- 38:54Has room to run as well. Oh yeah, yeah, and obviously
- 38:57Databricks isn't going anywhere. But Anthropic surpassing data
- 39:01bricks has been a bit of a sob story for my team because James
- 39:04got Anthropic at Crazy crazy money if I recall.
- 39:07The 4th pick that wasn't even your third pick because it's a
- 39:10snake draft. So it was me then Max the data
- 39:13bricks, then James with you'll say in a second, which doesn't
- 39:17make any sense, and then fourth with Anthropic.
- 39:20It's so embarrassing in retrospect, just.
- 39:23To just to clarify, like we drafted these teams originally
- 39:26in 2023 and then we did, I don't know, we drafted it.
- 39:30We, we had an ad drop waiver period last November and this
- 39:35again is a mid year check in. No, no ads, no drops, but
- 39:39checking in on the teams. So Max, I have your score so far
- 39:43right now at 93 billion, mostly because some of your teams have
- 39:48not raised or exited since you drafted them.
- 39:51I have no valuation on Sierra and scale is at 25.
- 39:55I'm somewhat optimistic on both of those.
- 39:57I managed to somehow pick the only foundation model company in
- 40:00the world that isn't wildly overvalued.
- 40:02Cohere SSI thinking machines, Anthropic.
- 40:08Like just throw a dart boarded foundation models.
- 40:11You've got a $40 billion company.
- 40:13But I'm. So Silicon Valley, the show said
- 40:16this from the beginning. No.
- 40:17Revenue is so much better than revenue.
- 40:19Cohere is a real business, so people can value.
- 40:22And. SSI and thinking machines are a
- 40:25dream. I have made a huge real.
- 40:29Revenue I I think SSI has real revenue, but anyway.
- 40:32Yeah, Max, what's your learning from this so far?
- 40:35I would say we already knew this, but you know the winners
- 40:38keep winning, right? Obviously Open AI swamps
- 40:41everything else that's happened in the entire draft.
- 40:43So we have a true power law which is nothing else matters
- 40:45comparison to Open AI even with the handicap which.
- 40:49Which we knew, which we knew was the risk when we came into it.
- 40:51Was happening, but regardless it still happened.
- 40:55Secondly, I would say Databricks is, you know, a merely a $60
- 41:00billion company, but that looks lame compared to, you know, like
- 41:05XAI being valued 80 like. You know, all right, all right,
- 41:08let's not spoil. James, you want.
- 41:10To go next. My team with the first pick on
- 41:13my draft that you guys were making fun of just the moments
- 41:16ago hugging face, No value because they have not raised
- 41:21since 2023. I've drafted them because they
- 41:24had one of the highest valuations at the time of the
- 41:27draft. They were valued, I think, over
- 41:28a billion dollars. I thought they were valued at 4
- 41:30or 4 billion. Dollars at the time, yeah,
- 41:32something like that, yeah. Yeah.
- 41:33OK So Anthropic giving me 61 1/2 billion value replit hasn't
- 41:42raised since the original draft. I exited Adept at 1 billion.
- 41:47I snagged XAI with the first pick last November.
- 41:52I'm. So jealous of that.
- 41:53Locked in 80 billion of value right there because they raised
- 41:57earlier this year and runway also raised $4 billion
- 42:02valuation. 11 Labs also raised at a $3.3 billion valuation and
- 42:08poolside no raise recently. Also one of those foundation
- 42:14model companies that we have yet to really see appear on the
- 42:18draft board, but I am happy with my overall team and my score of
- 42:23close to $150 billion currently. You know, something I just
- 42:27thought about, You're extremely lucky that XAI purchased X and
- 42:33not the opposite way, because if it had been X, it purchased XAI.
- 42:37We'd be able to like, force you to disown whatever growth, but
- 42:40now you get to benefit from this combined monstrosity, which
- 42:45kudos to you. I would just say had I been able
- 42:48to draft first, I would be in James's position of being in
- 42:51second place with XAI, so I don't personally think that a
- 42:55coin flip should be dictating my performance right now.
- 42:59What is you? I know.
- 43:01What was me I I will say to give James credit here, I I truly
- 43:05believe that anthropic is is the pick of the draft or, you know,
- 43:08so far I think that just getting anthropic at the fourth position
- 43:12in retrospect looks insane. And so I think that that is the
- 43:16the greatest. I don't even know if I call it a
- 43:18sleeper is sort of a semi sleeper pick, but that that
- 43:21clearly to me has had the most appreciation.
- 43:23Eric, why don't you all? Right.
- 43:25So yeah, I picked Open AI with a $75 billion handicap.
- 43:29Now it's worth 300 billion. So I'm getting basically 225
- 43:33billion for that. Inflection sold for 1.43
- 43:37billion. Character sold to Google for 2.5
- 43:41billion. Glean we're scoring at 4.6
- 43:45billion, but rumored to be raising at 7 billion.
- 43:49Miss Straw AI worried about that one. 6 billion right now.
- 43:54Perplexity. Oh man, I'm getting no credit
- 43:56for that. That's going to be a good 10
- 43:58right now, but it will be 14 is apparent according to the rumor,
- 44:02so we'll see. Safe super intelligence.
- 44:05I knew this was buzzy, but I don't even know if I could have
- 44:08seen this one raised at $32 billion already.
- 44:13It's more perplexity like That's insane.
- 44:18Kodium sold for 3 billion do. You want to explain that it's
- 44:23they renamed to Windsor for their name.
- 44:25Kodium is windsurf. Yeah, they sold to Open AI and
- 44:28then Harvey. No credit right now but rumored
- 44:31to be raising at 5 billion. Total value $274.5 billion.
- 44:39Yeah. I feel really good about this.
- 44:41I mean, hysterically, as I think I mentioned on the last episode,
- 44:44I I wrote a bear case about open AI after this at 157 billion, I
- 44:49think, but whatever. So I'm getting it narratively
- 44:52both ways. But yeah, I mean, I'm proud of
- 44:55all my picks. I think even my sort of singles
- 44:58are selling and I'm bullish on basically everything except I
- 45:04would like to hear what's going on with Mistral.
- 45:07But yeah, it's I mean, it's a power law business.
- 45:10It's crazy that like I'm like, oh, glean that's that's a good
- 45:13company. I was totally right that that
- 45:14would be a good company, but it it doesn't really matter for my
- 45:17my performance. Yeah, I mean I think you are
- 45:19consistently hitting singles and double s, but you could have
- 45:22nothing on your team except opening eye and be beating us by
- 45:25100 plus billion dollars at this point.
- 45:26So it. Doesn't, no.
- 45:28Which is why I made a that we could not randomly assign the
- 45:31first one and I I said we know. You were right.
- 45:34So I'm complaining, I I'm not complaining.
- 45:36You made the right decision 100%.
- 45:38It's just it, it is remarkable. It's like the whole game is just
- 45:42like, open. AI who drafted Open AI who
- 45:44drafted one Open AI go back. To that episode.
- 45:47Yeah, exactly. Yeah, No, we talked about it.
- 45:49I mean, I think it was it would they were raising at 90 at the
- 45:52time. And so you ended up with a $75
- 45:54billion handicap on a $90 billion company, which seemed
- 45:57like a reasonable deal to us. But you know, we were.
- 46:00We were all wrong, obviously. Or at least we're wrong with us.
- 46:02Check in. These do have to last, right?
- 46:04It's like 5 years or something. We're we're like, yeah.
- 46:07Five years, yeah. Yeah, we had a good shot with
- 46:10Sam Hoffman getting fired of your team.
- 46:13Going up sync. But but then he came back in
- 46:16force. All right.
- 46:17What's come on the market that you think we'll be looking at at
- 46:19the end of this year? Thinking machines.
- 46:21Cursor for sure. Manus.
- 46:24Oh yeah, yeah, yeah. Thinking Machines, cursor and
- 46:27manus are the ones that come to mind for me.
- 46:28Yeah, I mean, it's, it's going to be a tight band because we're
- 46:30we're picking them up and we're only interested in ones What
- 46:33that. They have to have raised $100
- 46:34million. All right, well, that's that's
- 46:36basically our episode. We're gonna have two more before
- 46:40the Cerebral Valley AI Summit in London on June 25th.
- 46:43And then at some point, once we've gathered ourselves, gone
- 46:47to that wedding in France we mentioned and relaxed a little,
- 46:50we'll come back to you and give you our thoughts from the event.
- 46:53I'm super excited about next week.
- 46:56No pressure James, our our game master over here, but we're
- 47:01trying to come up with some good concepts, but we'll be talking
- 47:05about voice and video and certainly in light of what VO3
- 47:10Googles new video creation model, it's an exciting time in
- 47:14video. So see you next week.
- 47:16Thanks guys. See ya.
- 47:18Thank you.