Latest / Newcomer Pod / AGI, Seriously + Foundation Model Polyamory (With Kyle Harrison)
Transcript
- 0:00Tom published his first post Tom Newcomer byline How does it
- 0:03feel? Feels like I'm born again.
- 0:08One of our mutual friends, I think Corey, the information was
- 0:11like, I'm going to call it Doe Tom whenever you're right here,
- 0:13it's like Tom a newcomer. I don't know it's.
- 0:16Funny, yeah. We haven't discussed branding on
- 0:18my stories, but if we could have some sort of like a hover
- 0:20graphic or something, we're like if the mouse goes over newcomer,
- 0:24like my name kind of fades in, flashes in.
- 0:27Yeah, let's talk about it. I'm willing to work it into comp
- 0:29discussion. This episode is presented by
- 0:32Breck's The Financial Stack that founders and VCs can truly bank
- 0:36on. Imagine what your founders could
- 0:38do with their runway if they had a banking solution that had no
- 0:41minimums, no transaction fees, and 20 times the standard FDIC
- 0:46protection. Plus they could earn an industry
- 0:49leading yield while maintaining access to funds whenever needed.
- 0:52Brex simplifies financial services for start-ups so they
- 0:56can focus on building. Connect your portfolio to the
- 0:59financial stack that one in three US venture backed
- 1:01start-ups already use. Check out brex.com/banking Dash
- 1:07solutions. Your piece was basically markets
- 1:12collapsing, but otherwise it would have been a great business
- 1:15news period for the foundation model companies, which is not
- 1:19your disposition. I feel like almost like I didn't
- 1:21order you to come in and say like tell everybody how AI is
- 1:23doing great. It's like you would be the guy
- 1:25to like, be like, I don't know, I'm worried about everything.
- 1:28Eric's hitting the techno optimism whip like you got to
- 1:31publish a positive story. Right.
- 1:33I wish you'd come in negative. So it didn't sound like, oh,
- 1:36you're writing for us, the AI conference people, You're going
- 1:39to give a booster take What? What convinced you that like
- 1:42things were going well in Foundation model?
- 1:45Work. Yeah, I really wanted to come in
- 1:46with a negative piece and then quit right after that and say,
- 1:49like Eric rejected my my contrarian take, my contrarian
- 1:53market take. Spiked.
- 1:54Yeah, exactly. Immediately, like this is who
- 1:57pays the bills here at Newcomer. No, On the contrary, Eric.
- 2:00He wants harder hitting stuff than what I'm delivering.
- 2:02I'm actually very worried. I was talking to some people and
- 2:06I just wanted to get, you know, a dipstick test of where AI was
- 2:10at the moment. And much to my chagrin, much to
- 2:14my disappointment, it was a very strong start of the year.
- 2:17You know, like I, I have this quote in the story, but I was
- 2:19talking to an investor who was telling me we all had in mind
- 2:23this idea that competition was the most important thing and
- 2:26there were only going to be a small number of winners and we
- 2:28needed to back those winners. And, and everyone else is going
- 2:30to be shit out of luck. And at least so far, it turns
- 2:34out that like the market size is large and there are enough
- 2:37people on the consumer and business side who want to pay
- 2:40for this stuff that the businesses, you know, and we
- 2:43were looking mostly at foundation model companies.
- 2:46So the ones that have raised the most money, but are building the
- 2:49tech that's powering almost all the applications.
- 2:53They had really good quarters and their business grew
- 2:57substantially. And, you know, in certain cases
- 2:59above initial projections, basically things are going very
- 3:03well in AI right now. And the only thing that could
- 3:05maybe derail it in the short term is the economy collapsing
- 3:08is the markets, you know, shitting the bed and, and
- 3:11everyone realizing that they're going to pull back on their
- 3:13spending because right now at least they're buying into the
- 3:15hype. And and so, yeah, that's the
- 3:18backdrop of it. It's like we've got all these
- 3:20numbers, we have these exclusive figures on, on how all the
- 3:24foundation model companies are doing, yeah.
- 3:25Give us a give us a tier ranking like who's who's doing the best
- 3:29of these foundation model companies right now for your for
- 3:31your numbers. Good, good week, bad week, good
- 3:34good quarter, bad quarter. It's kind of hard to dislodge
- 3:37open AI right now. They're really, they're really
- 3:40killing it as much as it kills me 'cause it's a much better
- 3:43story when they're not. I'm sorry, but you know,
- 3:46relentless product roll out pace, Yeah.
- 3:49We're gonna talk about O3, which I'm extremely excited about.
- 3:51We'll talk about the actual tech after we talk about sort of the
- 3:54business, you know, revenue. Right.
- 3:56So, so the the product stuff which we'll get to the revenue
- 3:59stuff which they are projecting that they could get as much as
- 4:0212, you know, almost 13 billion in revenue this year and that
- 4:06seems very plausible based on their user growth.
- 4:10You know, Sam had this. I hadn't watched it until I was
- 4:12doing the story. Very funny interview at the Ted
- 4:14conference where he kind of let slip that their users doubled in
- 4:20the last couple of weeks and, you know, back of the envelope
- 4:23you could. Thanks.
- 4:23Studio Ghibli, right? I mean, it's got to be mostly
- 4:27right. Like I that's got to be a lot of
- 4:29it. I feel like the yeah, the image
- 4:31generation has to just be overwhelming demand coming from.
- 4:35That nothing like a product. People actually like that.
- 4:37It's like, oh, let me show off using it to everybody.
- 4:39It's not like a social feature. It's like literally something so
- 4:42cool that it becomes like marketing for your product.
- 4:46But your Anthropic was doing well too.
- 4:48It is not just open AI. Right, Yep, Anthropic, which we
- 4:51really dove into their numbers, they finished off 24 with like
- 4:55950,000,000 in revenue. So very solid year.
- 4:58And you know, already at the beginning of the year they're
- 5:00tracking to do annualized 2 billion, well over 2 billion.
- 5:04And by the end of the year, if growth continues to pace, it
- 5:06could be as much as 4 billion. So that's still, you know, 1/3
- 5:09of open AI business, which is interesting, you know, and I
- 5:12think we're paying attention to the fact that one is like head
- 5:14and shoulders above the rest. But like Anthropic is really
- 5:17established itself as the enterprise AI model maker of
- 5:22choice. And if when it comes to coding,
- 5:25Claude is sort of where it's at. What's interesting with that is
- 5:28as I was doing this piece, like literally we were going to, you
- 5:30know, getting close to to putting the last touches on it.
- 5:33News breaks that open AIS and talks to acquire windsurf, which
- 5:37is this big AI coding. What big?
- 5:40But they're like, you know, a, a somewhat positive.
- 5:42Codium until like a couple days ago.
- 5:44Yeah, I didn't even know about windsurf, to be honest.
- 5:47Windsurf was the product you know they can be with cursor
- 5:50which is called any sphere. But for some reason, this
- 5:53category is full of companies with products that are more
- 5:55famous than the underlying companies Anyway.
- 5:58Yeah, Open AI is supposed to be buying it for 3 billion, which,
- 6:03you know, will scare the shit out of every application
- 6:06company. Or what do?
- 6:06Yeah. What do you guys make of that?
- 6:08I mean my first take on it 'cause I actually did not
- 6:11realize this until I was, you know, digging into Anthropics
- 6:13numbers. But those coding application
- 6:16start-ups, the any, What the fuck is it?
- 6:18Any sphere? Any sphere?
- 6:20They should. Just change that.
- 6:21Everyone calls it cursor and podium anyway guys, I I think
- 6:23generally across. AI so there is a company called
- 6:27Any Scale and you know, I'm like, I do all this conference
- 6:31organizing and like we we did this enterprise 30 list.
- 6:34So I accidentally emailed like any scale about any sphere and
- 6:39like they both ended up on the same thread.
- 6:41So I was on this like comical thread where like both of them
- 6:45were replying and I wanted one of them.
- 6:47I ended up talking to any scale just because it's like, oh, now
- 6:51we're on a thread. You're another interesting
- 6:52company. Like let's do a call.
- 6:54But yeah, it's extremely confusing.
- 6:57That's that. That's your growth pitch to
- 6:58these people. It's like since we're on the
- 7:00phone. Yeah, since I mistook.
- 7:02You for another company. Way it's like now that you've
- 7:04wasted our time, like you at least need to hear our spiel,
- 7:07you know, like. Anyway, we should have a whole
- 7:10episode on the shitty naming across all of AI.
- 7:12Like why O 4 is the less powerful to O3?
- 7:16Absolutely, yeah. I would say I feel like
- 7:19honestly, any sphere, any scale has become like a recurring
- 7:22character on this podcast to this point.
- 7:25Just like pick a name. Yeah, any, any port in the
- 7:28storm. So their acquisition of them is
- 7:30actually could be hugely problematic for Anthropic
- 7:33because one of the, you know, significant drivers of business
- 7:37for Anthropic are these coding startups that incorporate
- 7:40Anthropic's models into their application and that that powers
- 7:43the whole thing. And you got to assume if open
- 7:46AI, you know, continue like, like, like gets this acquisition
- 7:50across the line, they're going to cut out all of Anthropic's
- 7:53models. And I don't know exactly how
- 7:56much revenue I should look into that one, but it's certainly an
- 7:59offensive move on, on on open a is part.
- 8:02So that's something that to to watch that deal aside,
- 8:05Anthropic's in a great place and there's a huge amount of
- 8:07confidence there. And again, for the second place
- 8:11player to be in a good space generally means good things for
- 8:14the health of the sector. I I don't think it's just that
- 8:16Anthropic could lose the revenue.
- 8:19I think what matters is that Open AI is going further into
- 8:24the application space, right? It's not just chat.
- 8:27ChatGPT is their application. It would seem they want to
- 8:30provide other applications, which means anyone built on top
- 8:34of Open AI has their ears justified.
- 8:36I can reveal here for the first time that one sentence in Tom's
- 8:40story was written entirely by O3.
- 8:43You can pause here to go see if you can identify it.
- 8:46But it it was any AI app acquisition will jolt the
- 8:50remaining independence, confirming their fear that Chat
- 8:53GPT's parent won't settle for being just the brains behind
- 8:57everyone else's software. I had the one convoluted
- 9:02sentence and it rewrote it and you're like ChatGPT is better.
- 9:05I want to let listeners in. OK, You kind of got to behind
- 9:07the scenes here. This is a line that Eric wrote,
- 9:10not a line that I wrote. Because.
- 9:12Anytime anytime I put my language into ChatGPT and I say
- 9:16can you make this sound better? They say I can't.
- 9:19I truly can't. This is this is peak human
- 9:22intelligence, and it would be foolhardy for me to suggest
- 9:25anything. It just shuts off.
- 9:28When you try, yeah, I actually get a call from Brad Lightcap
- 9:30every time I put my wording in there and say we'd actually like
- 9:33to refund you $20, Sir. So it's the program. $200 now,
- 9:38yeah. Yeah, this stuff's getting
- 9:39better. I, I think the, the market right
- 9:42now is, is really is really strong and it pains me to say
- 9:45it. Well, what about, I mean, even
- 9:47in kind of apps you you looked into perplexity too, right?
- 9:49Like how are they doing? Briefly, we did, yeah.
- 9:53Good. You're doing good.
- 9:54Good. Yeah, I mean the CEO of
- 9:57Perplexity, you know, Arvind Srinivas, he's out there a lot.
- 10:00He likes promoting the company. And so he he tweeted out that,
- 10:03you know, they had reached 100 million in ARR in March.
- 10:07But digging a little bit deeper into them, I found that, you
- 10:10know, there's from the investor side of things, they're
- 10:13expecting that to almost double by the end of the year, so they
- 10:16feel really strong about stuff. You talked about opening eyes a
- 10:19big player. I mean, Google's Gemini 2.5 is
- 10:22hitting a bunch of benchmarks too.
- 10:23But meta seems like kind of fell off with Llama, like it's been
- 10:27doing nearly as well. I know obviously there's reports
- 10:30of engineers taking that they worked on the latest model off
- 10:33their LinkedIn or clarifying that they did not work on the
- 10:36latest model on their LinkedIn because it's just not.
- 10:39Doing the The accusation is that what Llama for was built to do
- 10:43well on the test, but not necessarily that big of an
- 10:46improvement. Yeah.
- 10:47I mean, first of all, I relate. There are certain jobs that I've
- 10:50had that I've taken off of my LinkedIn as well as I.
- 10:52Never know. So it's nice to know that, you
- 10:56know, you know, AI researchers are just like us.
- 10:59Llama's a kind of huge question mark right now.
- 11:02People were very impressed with the earlier iterations of it.
- 11:05I mean, this was like a high quality open, you know, or free
- 11:08to use, you know, thing that basically became the standard
- 11:11open source model. And then Deep sea came along
- 11:14and, you know, had deep reasoning capabilities, which I
- 11:17think really appeal to a lot of people.
- 11:18And then Llama 4 was supposed to come out and blow everything out
- 11:21of the water and they had, like you were mentioning Madeline.
- 11:26You know, these benchmarks look very impressive, but when people
- 11:28were using it, like when developers actually were trying
- 11:31to apply it, they were like, man, this sucks.
- 11:34You know, it's like very verbose and it's answers are much worse
- 11:38than, you know, what the competitive ones are, are
- 11:40offering. And it got to a point where AVP
- 11:43of AI at Meta had to get out there and say like, well, we're
- 11:47working through some bugs right now.
- 11:49We're trying to make sure that everything works for our for our
- 11:51users. And these accusations that we
- 11:53were training Llama on the test set so that it could perform
- 11:56well on the benchmarks is totally unfounded.
- 11:59And look, I don't know yet and I'll look into it as to, you
- 12:03know, what true or isn't true. But once you're doing that kind
- 12:05of damage control, you kind of already lost like it's a it's
- 12:08APR game to a degree. And so much of open source is
- 12:11like, hey, we're cool in the space.
- 12:13We're going to attract engineers.
- 12:15I mean, to some degree, Mehta's just trying to kneecap everybody
- 12:18else and be like, we give it away for free.
- 12:20You know, there are fewer giant tech companies.
- 12:22And if DeepSeek is playing that role, you know, Mehta still
- 12:25wins. And sort of making it hard for
- 12:28for everybody else to compete. But yeah, llamas need to, you
- 12:31know, need to rebound, I think on the next one.
- 12:34And I'm kind of, I'm, I'm sympathetic to them because
- 12:37these models like you do these training runs and you, you're
- 12:40kind of get what you get. And if it's not good enough, you
- 12:43do another one And, and either you release it or you don't.
- 12:46And they're clearly, I mean, if it truly is as bad as some
- 12:49people are saying, you know, there was a lot of pressure on
- 12:51them to just release something and, you know, they put out
- 12:54something that maybe wasn't very good and like.
- 12:56That's the beauty of everybody's competing.
- 12:58We should be happy, you know. I do feel like separate from all
- 13:01of these numbers on how these companies are doing, looking at
- 13:04the technical specs, like people are really excited about 03.
- 13:09It's like some people are saying we're hitting the AGI moment
- 13:11right now. Eric, you were AGI momenting all
- 13:14over Twitter this week. You're like, oh, yeah, you.
- 13:17I was up. I was declaring on Twitter like,
- 13:20oh, history will reflect. This is a huge moment, but the
- 13:23headlines are about everything. But you were.
- 13:25In our slack last night at like 9:00 PM just being like wow, how
- 13:28did it do this? You were.
- 13:30Like I told you guys, you're not obligated to engage with me when
- 13:33I'm just like slacking away. But Tyler Cowen said AGI,
- 13:37seriously, he really thinks his AGI.
- 13:39I was getting some pushback from AI diehards that are like, OK,
- 13:44O3 isn't so much better than Gemini and Anthropic, and I
- 13:47guess I'm not a student of which one is better.
- 13:51I think what's amazing to me about O3 is how great it is like
- 13:55searching and sort of pulling a bunch of stuff together along
- 13:58with reasoning, I think. And we're going to make a
- 14:01concerted effort to prove this true.
- 14:03I think we can get a genuine scoop thanks to ChatGPT, which I
- 14:09think would be a landmark moment from reporters.
- 14:11I don't know how much that's sustainable versus like there's
- 14:15a lot of stuff that's sort of like on the Internet, the search
- 14:18isn't as good at finding the now ChatGPT, with its ability to
- 14:22analyze stuff, can sort of put pieces together.
- 14:25I don't know how. Exactly.
- 14:26I was going to say scoops were like the, I mean, scoops are
- 14:28like the the one thing that AI doesn't have on writers, right,
- 14:32you know? Right, exactly.
- 14:32If they can do that, then it's AGI.
- 14:34That's what I'm saying from my perspective.
- 14:36I actually don't think that they will be able to turn out scoops
- 14:40because you pay $200 a month for ChatGPT and that's much less
- 14:46than you're paying me for scoops.
- 14:48And I'm very concerned that you think $200 a month equals the
- 14:52scoops that I could be able to give you.
- 14:54So by that metric, I do not believe.
- 14:57That they're able to get scoops says otherwise.
- 14:59You're yeah, right. Just that's true.
- 15:02But investment? Wise speaking, it's just not
- 15:04possible because otherwise you would just pay for that and not
- 15:06me. You know, it's, it's common for
- 15:09you, you know is what. Yeah, I believe that's not true
- 15:13and I believe it can't do that. And there's a huge amount of
- 15:15value for paying people to write scoops, Significantly more,
- 15:19significantly more than the $200 a month that you were paying
- 15:23for. For cheek it is.
- 15:24Partially tongue, tongue in cheek, but like, I mean, I do
- 15:26think there's a reality that like lovable can code, but like,
- 15:30I'm not a coder and so I'm not seeing it through.
- 15:32And there's going to be the same way that Chachi BT can do some
- 15:36of these things. But if you're not like a
- 15:38craftsperson, you don't really have the taste for it.
- 15:41You don't know what to look for. It's sort of like it's hard to
- 15:43really extract it. And so I think some of the
- 15:46threats of job replacement in these sort of areas where
- 15:50there's a lot of taste and you have to know all the sort of
- 15:52steps, even if it does some of the steps well, it's, it's only
- 15:56going to be a tool for the craftspeople.
- 16:00Yeah. I mean, what was interesting
- 16:02with the stuff that you were, you know, going on about on
- 16:05Slack, basically? You know, it can find filings
- 16:07for like venture Performance. So we hadn't found it right, It
- 16:10can find. It can find unlisted links that
- 16:12are listed by, you know, like these venture the LP's and the
- 16:16endowments and, you know, pension funds where they have
- 16:20these filings already on the web.
- 16:21But you have to know the URL to find the specific filing
- 16:25document. So it's not, you know, fully get
- 16:27guarded, but it's, you know. And it and it analyzed some
- 16:30filings from Amazon. I don't think we would ever do
- 16:33this story. It analyzed some findings from
- 16:34Amazon that broke down the nature of Amazon's investment in
- 16:39Anthropic that, you know, Reuters or wherever who had
- 16:43written the story hadn't really gotten the full breakdown.
- 16:46And so it argued that there were like nuances of how the
- 16:50investment was structured, that it caught on to that, that
- 16:53reporters hadn't yet found which.
- 16:55So it's already doing like incremental scoops.
- 16:57It's already doing, you know, like actually this is new
- 16:59information. It's arguing with its editor.
- 17:02It's the needle, you know? That's honestly the strongest
- 17:05case for AGI I've seen yet. If it's already like a very
- 17:08self-conscious in, you know, sensitive reporter that's trying
- 17:12to say, no, this is new Intel, this is good stuff.
- 17:14Why? Why do you think this is just
- 17:15the same thing Reuters did? I've never more deeply related
- 17:18to an AI. No, I, I seriously though, I, I
- 17:21think there's an interesting case to make that if this
- 17:23somehow does get more wrapped up into the reporting process and
- 17:27the pulling of public documents or semi public documents is
- 17:30something that like O3 or O3 Pro can do pretty well.
- 17:34I think it puts more pressure on the aspects of reporting, which
- 17:37is gathering information from human beings.
- 17:39And it's not just document diving which is value.
- 17:42Like I feel like one of those woke people I need to be
- 17:44snapping to you. I couldn't agree agree more with
- 17:46what you're saying. Yeah, that's why we're video
- 17:48too. Exactly.
- 17:50I mean this is about the. People.
- 17:52AI is about the people. AI makes us more.
- 17:53Human there's 2 levels about it. One like people, the the sort of
- 17:58like what people are like isn't really like on the Internet.
- 18:01Like maybe you see YouTube videos, but it's like hard to
- 18:04it's not a big data problem. It's like a small data problem
- 18:07where you need to, you know, talk to people, interacting with
- 18:09people. So getting the story of what
- 18:11people like and how organizations are working there
- 18:14isn't like the data for the AI to consume.
- 18:17And then obviously humans are just better at understanding
- 18:19like what's interesting about a person and like telling a
- 18:22personal story. So I totally agree with you.
- 18:24Like AI will be better at solving data stories and stuff
- 18:28where it's like pulling together what's what's happening with
- 18:31like money and numbers. And reporters are going to have
- 18:34to focus on people like more than ever.
- 18:36It's interesting to see how this tech will be incorporated into
- 18:40newsrooms because there's a desperate move by all top
- 18:43people. And I know this from my time at
- 18:44the Wall Street Journal to force down the newsroom's throat AI
- 18:48and into the process because it makes you seem futuristic in
- 18:51advance. And you can pitch to, you know,
- 18:53your non journalistic business execs that were embracing AI.
- 18:57And so far, that's mostly meant using like AI generated talking
- 19:00points or like bullet points at the tops of stories, which at
- 19:04best maybe removes one person whose job was to write those
- 19:07things. But it's not AI.
- 19:10It's not actually disruptive in the in the flow of news or
- 19:12reporting. So maybe what you were messing
- 19:14around with Eric will be closer to it and that can actually
- 19:17change things. But aside from that, it's been
- 19:19pretty, I think. It's a good sounding board, good
- 19:21proofreader. I mean, we still pay a
- 19:23proofreader, but you know, it's not.
- 19:25It's going to be more and more helpful.
- 19:26But I agree that the heavy-handed attempts to to use
- 19:30it can be heavy-handed. Do you think O3 and similar
- 19:34models represent AGI as Tyler Cowen seems to say, or as crank
- 19:41Gary Marcus? Please, this is nowhere close.
- 19:44Can't even do math with decimal points.
- 19:46It's idiotic and there's a propaganda machine and.
- 19:49I, I don't know, man, I don't know that the bar is always
- 19:53moving. I don't really know what it
- 19:54means. I mean, there's like financial
- 19:56definitions for it. There's philosophical
- 19:58definitions. I have a feeling that like this
- 20:01back reverse engineering of AGI and say like we've actually had
- 20:04it for the last couple of years of the last year or so because
- 20:07of reasoning models that's weak. I'm sorry, Like I, I want a lot
- 20:10more than that. Like AGI to me needs to be like
- 20:13in SIM City 2000 when like a headline in the newspaper said,
- 20:17like nuclear energy exists and you're allowed to like build
- 20:20nuclear plants and and and. I'm most sympathetic to that,
- 20:24that it hasn't produced any real innovation or new ideas, and
- 20:30that is some barrier. But yeah, is does that mean that
- 20:34the threshold for AGI has to be novelty?
- 20:36Because I feel like while I'm as a creative person would love to
- 20:41say yes, that's the threshold and therefore you can't do what
- 20:44I do, There are plenty of things that it can do.
- 20:47And if your definition of AGI is to replace someone in the
- 20:50workforce at a competent level, it's getting closer and closer
- 20:54to being able to do that to the point where like this could be
- 20:56the AGI moment. Like I think what you were
- 20:58working with Eric just for it pulling up filings and things
- 21:01like that. Like it can do stuff that
- 21:02analysts can do just hands down already.
- 21:05I, I think it's safe to say I know a lot about venture capital
- 21:08and I think open AI knows more than me.
- 21:11You know, it's just like it can, it can pull all the world's
- 21:14information and sort of synthesize it.
- 21:16It makes an occasional error, but so do I.
- 21:19Like I think even now in areas where I'm an expert, it has more
- 21:25context and information than me. I mean, I, I, you know, I, I'll
- 21:29be honest, I've known Gary for a while.
- 21:31I've probably muted him at this point.
- 21:32And that's more just out of like respect for him as a person
- 21:35because I like him, but he's so fucking annoying on X and I'm
- 21:39willing to keep that in the pod like he just is, even though I
- 21:42think he's a really smart guy. But I've seen some interesting
- 21:45stuff on there. Just pointing out like it's
- 21:47inability to do what would be basic logic that a human, you
- 21:51know, average and even maybe slightly below average
- 21:53intelligence could do. There was this one picture, I
- 21:55don't know if you guys saw it, of like these curvy lines
- 21:59pointing from a name to like a stick figure.
- 22:01And it was saying like it, you know, it put the image into O3
- 22:05and asked it to identify the person with the color of the
- 22:08stick figure that it was pointing to.
- 22:10And it got it wrong, completely wrong.
- 22:12And that's something that probably 1/3 grader could do.
- 22:16And I think those sorts of mistakes are meaningful.
- 22:20Like sure, it has research tool capabilities.
- 22:23And I think, you know, this gets into like the copilot versus
- 22:27agent debate, like is it something that works as an
- 22:30assistant or is it like autonomous?
- 22:32And I think those mistakes, until they're fixed on a regular
- 22:36enough basis, stopped me from getting too excited about like
- 22:39human level intelligence. It just doesn't meet.
- 22:42It just doesn't meet that. All right, we've looked at how
- 22:45foundation models have been doing really well.
- 22:48Madeline, your story for Friday looks at RVC is reacting to
- 22:53market force. Well, they love to say that they
- 22:55do not react to the markets, right, Eric.
- 22:57They love to say that they're, you know, thinking of the future
- 23:00and that the markets now don't really interplay with what
- 23:03they're doing. True, true or false?
- 23:06Or what do you? What have you found?
- 23:08Charles Hudson, a precursor, told me best and said the
- 23:11tailwind of AI is overcoming the headwinds of the market at the
- 23:15late stage. They're delayed IP OS.
- 23:16We cannot disregard that. Obviously that's happening.
- 23:19I would not say that late stage is 100% affected.
- 23:22I'd say that there's more of a pause there.
- 23:24But like at early stage deals, deals are still flowing.
- 23:27Companies are getting backed. I most of the investors I talked
- 23:30to had not, you know, increased the pace of deals that they plan
- 23:34to do over the next few months and by the end of the year, but
- 23:37most of them also had not slowed down.
- 23:39The money is still flowing. It's just, of course, you know,
- 23:42the classic AI versus everybody else story.
- 23:44Do you think they're in denial about how bad the economy could
- 23:48get? I mean, Ryan Peterson, who
- 23:50everybody in Silicon Valley trust we had on the podcast, is
- 23:53like, if these tariffs don't get reversed, like tons of
- 23:56businesses are going to die. Like we're already seeing, you
- 23:59know, companies sell to their factory in China.
- 24:02Like, obviously those aren't tech companies.
- 24:04Those are like, you know, T-shirt manufacturers probably,
- 24:07but that could have a huge impact.
- 24:09Or like, how much do you think they're in denial about how bad
- 24:12the economy could get? I mean, I think that the early
- 24:15stage investors, it's not as much as they're in denial, they
- 24:18just like are looking at a 10 year horizon.
- 24:20So like in their mind, maybe the economy's fixed in 10 years, who
- 24:23cares? This is going to be the next
- 24:24generational wave of technology and we got to get in it right
- 24:27now. I'd say more at the Series B
- 24:29plus level, there's a little more doom and gloom.
- 24:32It's not entirely rosy. I've heard a couple rumors of
- 24:35term sheets getting pulled for companies in logistics, chip
- 24:39making tools, anything like kind of that touches things that'll
- 24:42be heavily tariffed that are raising right now.
- 24:45I don't think they're it's not that they're not going to be
- 24:47able to raise. It's just looking a little
- 24:48tougher. Anything in retail not looking
- 24:51so great. So yes, there are carve outs
- 24:53where there are like things that are not doing well, but there's
- 24:55also like the Consumer Packaged Goods companies like the D2C
- 24:58marketplaces that were like really big in the boom times.
- 25:01They haven't been doing super well honestly for a while.
- 25:04And this is just kind of like nail in the coffin for them,
- 25:07unfortunately. You know, if you're sticking to
- 25:09like VC, like software company, AI engineering, AI application,
- 25:14those deals are fine. That has not slowed down at all.
- 25:17No one, especially in early stage VC is like worried about
- 25:20the markets right now other than just like when you get to that
- 25:23sort of mid stage is when people start to like slow down a bit.
- 25:26If you had to do business, tech is where you'd want to do it.
- 25:30You know, the least physical goods possible.
- 25:32I mean, obviously what Tom you, you've touched on the fact that
- 25:36software is not just ones and zeros.
- 25:39There is like a hardware component that influences the
- 25:42cost structure. What's your view on how much
- 25:45worse things are going to get for the tech industry?
- 25:48I mean, I feel pretty good right now.
- 25:51Patriots are in control. You know, we've got the All In
- 25:55podcast, people mostly running policy.
- 25:58It's kind of hard for me to feel more confident than that.
- 26:00It seems like everything's gone according to plan.
- 26:02So I don't know where any of this doom and gloom is coming
- 26:04from. I mean, look, if you want to get
- 26:06things done, you've got to do things.
- 26:08It doesn't matter what they are. You just got to do things.
- 26:11And I think we're in the era of doing things.
- 26:13And I couldn't be more impressed with the administration's do
- 26:16things, whether the thing that they do undoes the previous
- 26:19thing doesn't matter as long as it was a thing that was done.
- 26:22And I also want to say I'm very impressed with the number of
- 26:25Twitter ads that I'm seeing that then have community notes saying
- 26:28this is a scam propagated by a drop shipping company.
- 26:32Well, they're. All I think drop shipping is
- 26:34getting killed. You're literally seeing tweets
- 26:36of like Donald Trump. I I voted for you, I love you, I
- 26:39worship you, you can do no wrong, but you're killing my
- 26:43business. Please.
- 26:44Yeah, yeah, Trump's completely turning his back on the whole
- 26:47drop shipping community I think was a huge mistake.
- 26:51Granted, these people don't seem to have money or it's not
- 26:53denominated in dollars, it seems.
- 26:55But I'm, I really think like Trump versus the drop shippers.
- 26:59Like I don't want to make any predictions, but like I I, I, I
- 27:03make sure I keep a close eye on the drop shippers if I would
- 27:05farmers. Are going to get a bailout.
- 27:07Maybe they'll do a targeted bailout on the drop shippers.
- 27:09If there's a if there's a drop shipping bailout, that would be
- 27:12incredible. Drop shipping Bailout.
- 27:15I we're we're cooked. It's over.
- 27:17I'm sorry. I want Liberation Day 2 is going
- 27:20to be a huge press conference where like everybody that's
- 27:23selling, you know, stuff that was a drop ship from Vietnam is
- 27:28that that is the end state of America that I've been very
- 27:31excited about. So.
- 27:32I. That's as much as I can offer on
- 27:34this one. Well, with that last little dash
- 27:37of Tom nihilism, I think we pretty much covered the whole
- 27:40gambit of news this week. We got, you know, the foundation
- 27:44models crushing it in terms of business.
- 27:46We've got VCs excited, markets be damned, except maybe a little
- 27:51bit worried, but we won't talk about it now because we're still
- 27:53doing deals. Eric, you had a nice chat with
- 27:55Kyle Harrison about VC Polyamory.
- 27:58How did that go? Yeah, he he did great thought
- 28:01leadership. He made it a viral image on
- 28:04Twitter. The ultimate influencer.
- 28:06That's like a research report in venture anyway.
- 28:08So Kyle Harrison at Contrary friend of the pod, came on to
- 28:12talk about venture capitalists investing in many, many
- 28:16foundation model companies in his chart of all the
- 28:18incestuousness in venture and foundation model companies.
- 28:21So it was a fun coda to Tom's reporting on the state of
- 28:25models. Kyle and I talked about how
- 28:27venture capitalists are playing all the money flowing into
- 28:30foundation model companies. Give it a listen.
- 28:33Hey, it's Eric Newcomer. I'm here with Kyle Harrison of
- 28:35Contrary, former investor at Index and CO2 Thanks for joining
- 28:42me. Thanks for having me.
- 28:43We we are here to talk about one specific thing, infidelity among
- 28:48venture capitalists. Where you, you had a viral image
- 28:52in tech world that shows all the different investments that VCs
- 28:56are making into foundation model companies.
- 28:59And you know, a lot of the VCs are making investments into
- 29:02competitors. So fun topic.
- 29:04Do you want to just start off like the impetus for making the
- 29:07graphic and and then sort of walk us through what you found?
- 29:11It was kind of a vibe that I had been noticing over the course of
- 29:15several months where I would see these fundraising announcements
- 29:18and it would be a big firm talking about, you know, I think
- 29:20it was XAI. I was the one that really
- 29:23triggered me where I saw people that were like, you know, we're
- 29:25so excited to back you on again, you generational company
- 29:28building all that. And it was all this like stuff
- 29:29about how great the business was.
- 29:31And I was like, weren't wait, didn't these people invest in
- 29:34opening eye or in traffic or whatever?
- 29:35And I went and looked and was like, oh, they did like, oh, I
- 29:37guess people are making a lot of cross multi layered investments
- 29:41and stuff. And then and then in a non
- 29:43account that I really like, who's one of the best LP
- 29:45accounts on here? Is this guy endowment Eddie?
- 29:47He was like, OK, wait, is backing competitive companies
- 29:50allowed now in venture? And I was like, let me go look
- 29:53at the cross section and it's like, yeah, I guess so.
- 29:55Yes, clearly is the answer. Do you just want to walk us
- 29:58through it like Andreessen? You have SSI, Ilya's company,
- 30:03XAI, Open AI and Mira's Thinking Machines.
- 30:07What are some of the other interesting ones?
- 30:10Yeah. I mean, I think the biggest
- 30:11thing was that you've got, and I think even some of these I
- 30:13missed, like this was not necessarily a purely academic
- 30:16exercise, but what's crazy. Is you made a screenshot, now
- 30:19you are the scholar on. This.
- 30:21That's right. That's the level.
- 30:22So it works. On the Internet, this is much
- 30:25more baked than most VC takes with just a screenshot anyway.
- 30:28Go ahead. Yeah.
- 30:29No, Harvard sent me a degree in the mail.
- 30:32Yeah, exactly. I, I feel like the biggest thing
- 30:35is like, you can kind of clearly see, and I think somebody
- 30:38actually retweeted this and pointed this out, that it's like
- 30:41a very specific strategy where like the one who's most active
- 30:44is Andreessen, which is consistent.
- 30:46Like Andreessen is super broad and investing in a lot of
- 30:48different. Martine Casado, who I love, set
- 30:50on stage at Cerebral Valley when I was like, how do you know
- 30:53which segment of AI to invest in?
- 30:56He was just like, why not all of them?
- 30:59They're they're like openly going for coverage at this
- 31:01point. And apparently even in the key
- 31:03categories with competitors. Yeah, that's Andreessen.
- 31:07So Andreessen I think is everywhere.
- 31:08You've got Sequoia who is, is also quite active, a little less
- 31:13than Andreessen. So a little bit more, you know,
- 31:16demur or whatever. And then I think you have firms
- 31:19like. Sequoia just did it to I mean,
- 31:21Sequoia's interesting because they obviously what Affirm and
- 31:24Chlorina, they have these famous ones where they've done both.
- 31:27They try, I think, to be loyal, but they would have more of a
- 31:31story of like, you know, I don't know, it's getting pretty broad.
- 31:35They're in a lot of them. I was going to try and do some
- 31:37spin that it's like, oh, they they're an X in that sort of
- 31:41relationship with the rule off world, and then they invest in
- 31:44the leader, which is clearly open AI in my opinion.
- 31:47I don't know. But yeah, what it was Any
- 31:49Sequoia analysis? I think cuz it's also, you think
- 31:52about the story of like, I don't know if you remember this, but
- 31:54they invested in Phoenix back in the day on the payment side and
- 31:57Stripe was pretty upset about it.
- 31:59And so Sequoia was literally like our bad keep the money.
- 32:02We won't take an equity stake. Like it's just a donation, you
- 32:05know, right. So they're definitely more
- 32:08thoughtful about where they're investing.
- 32:10I think to Sequoia's credit, one of the things that they've been
- 32:13very deliberate about is that they are more selective in the
- 32:16earlier stages. And somebody else I think
- 32:19pointed this out in a retweet where it's like once it gets
- 32:22into the later stages, you're largely just, you're garnering
- 32:25exposure to assets. Like it's not as much once these
- 32:28guys are raising, you know, $10 billion plus rounds or whatever,
- 32:32it's like you're really just getting exposure.
- 32:34I think Sequoia is a little bit more protective of their
- 32:37reputation at the earliest stages.
- 32:39Lightspeed is another one. I mean, anthropic is going to be
- 32:43one of their biggest positions overall.
- 32:46Like because you know, I dug into their returns and that was
- 32:48like an totally didn't realize that.
- 32:51And then they're in Mistral NXAI.
- 32:54I think they're also in one other, somebody else, I can't
- 32:57remember which one, but somebody pointed out that maybe there is
- 33:00one of on here that I missed. And because Lightspeed has also
- 33:04been super active. And I, I would argue that I
- 33:06think Lightspeed has been playing not necessarily like
- 33:09catch up, but they're, they're really trying to like
- 33:11deliberately demonstrate that they want to be in this bucket
- 33:14of like the Andreessen's GCS of the world, like just be very,
- 33:18very active. What's funny, where is GCGC and
- 33:21NEA like? If you're going to do the
- 33:24Andreessen style blast them all approach, shouldn't you be in
- 33:27the defining category of our time?
- 33:30Yeah, it's a good question. And I I will.
- 33:31I will defer back again to the lack.
- 33:34Of my they. Might be they might be in five
- 33:38of these and I just didn't want to look up another crutch based
- 33:41profile. Look up everyone.
- 33:42OK, they'll they'll they'll be like, no, we're we're sluts in
- 33:46this ad. Who?
- 33:50Who are the serial monogamists here or who who do you think of
- 33:53the sort of big brand firms has sort of stuck to one bet?
- 33:58Well, I think that and now it was actually a sort of oversight
- 34:01on my side like in making this chart because originally it was
- 34:03supposed to show the cross section, but I kind of just
- 34:06started at the bottom. I said, OK, open AI, who's in
- 34:08that boom, boom. And I worked my way around with
- 34:10the companies adding firms as I went.
- 34:12And what I realized when I get when I posted it and it kind of
- 34:15popped off is that I had left founders found on there despite
- 34:18the fact that they are actually monogamous in this in this
- 34:21particular strategy. And the other one that I didn't
- 34:24put on here, but that that called it out is Thrive.
- 34:26And I think that again, going back to this like this actually
- 34:28pretty indicative sampling of different firm strategies.
- 34:33I think that Founders Fund and Thrive are famous for being
- 34:37really dedicated to running an incredibly concentrated
- 34:41portfolio and just like doubling down, tripling down into their
- 34:45winners. And founders fund, despite being
- 34:46like perceived as like, I don't know, perceptive and early, you
- 34:51know, Airbnb was like series C. Some of these they're like, what
- 34:54is the momentum company that's going to define the generation?
- 34:56How can we like back up the truck into it?
- 35:00And you know, clearly they, they see open AI here.
- 35:03Well, and I think that's one of the reasons why they're not
- 35:06afraid to admit when they did, even if they didn't get there as
- 35:10early as possible, right? I think that and probably in
- 35:12that in that vein of monogamous, I think Khosla is another one
- 35:15where they've been all open AI all day for AS.
- 35:18Long as possible, like you roll that company in just right.
- 35:22Jesus, right? Yeah, that would be insane.
- 35:24I think to diversify away. It's one thing, you know, I
- 35:28think a consideration for a lot of these is they're getting an
- 35:30open AI. It's such a high price that it
- 35:31fits into the growth story where you're talking about where they
- 35:35were like, not traditional venture investors in open AI.
- 35:37They're just like we can't miss the company of our generation
- 35:41and then they're like, well, let's try and do real venture
- 35:43investments in other model companies if we get a shot at.
- 35:46Yeah. So I think the and so, but I
- 35:48think to founders funds credit, even if they're not there at the
- 35:51earliest days, they're willing to acknowledge that like, hey,
- 35:54listen, this is clearly broken out as a category definer and
- 35:56they'll say that too. Like I've heard Trey say this
- 35:59multiple times where they talk about, listen, like our
- 36:02perspective is that there's one company and like, if you're
- 36:04going to invest in space and you didn't invest in SpaceX, you
- 36:06probably lost money. If you want to invest in defense
- 36:09and you didn't invest in Anderol, you probably are going
- 36:11to lose money. Like they, that's their thought
- 36:13process. And so even if they're getting
- 36:14there a little late, they're going to back up the truck.
- 36:17And that's Peter Thiel's like, you want to be a monopoly
- 36:19business. It's sort of like you don't.
- 36:20You don't want to be in a company that's creating a
- 36:21category necessarily, because then you have all these
- 36:24competitors. You want to be the company in a
- 36:26category that chokes out all the oxygen for everybody else.
- 36:29That's right. I, I, I also think that there is
- 36:31a narrative here, we can talk about this if you want to, that
- 36:34there is increasing differentiation.
- 36:37It's very subtle and it's, and it's going to start overlapping
- 36:40again, but there is some semblance of different companies
- 36:44trying to prioritize different places.
- 36:46We can talk about that if you want to, but.
- 36:47No, I think, yeah, let's let's talk about it, especially in
- 36:50light of Opening Eye potentially purchasing windsurf, formerly
- 36:54known as Codium, that that speaks to them, you know,
- 36:57demolishing everything, but also speaks to the idea that if
- 37:01you're an investor, it's hard to know where these companies are
- 37:04going to go, right? They're sort of going to lean
- 37:05into where they're successful foundation models in particular.
- 37:08It's like you could become an application, you could be
- 37:11infrastructure. You know, it's hard, hard to
- 37:12know. So are you really?
- 37:14No one is saying VC shouldn't invest in windsurf.
- 37:18And if windsurf is part of Opening Eyes turf, then it's
- 37:21sort of like, you know, the floodgates are open.
- 37:22Or what would you say about that?
- 37:24So I think that the, it's actually very, it's very, very
- 37:27similar to what happened with the hyper scalers where you look
- 37:30at like the Amazons and Googles and Microsofts of the world.
- 37:34The difference was that they're sort of amalgamation of a bunch
- 37:37of different faces happened long after they had been founded,
- 37:41long after they'd gone public. The difference was it's not a,
- 37:44it's not necessary because even like think about like open AI
- 37:46launching a social app and, and I had a friend of mine tweet and
- 37:51say, so are just like our network effects dead, Like
- 37:54there's nothing mean anything anymore.
- 37:55All the businessy stuff that we've learned is, is nonsense.
- 37:58And it's like, if you think about it, it's not that any of
- 38:01the fundamental principles have changed.
- 38:03It's that the velocity or like speed of the thing happening is
- 38:06just so much faster than it's ever been before.
- 38:09But at the end of the day, like if open AI does launch a social
- 38:11app and it does turn out to be successful, it's going to be for
- 38:15the same reasons why when Meta launched threads, they also got
- 38:19up there pretty quick is because they have a big broad
- 38:21established installed base they can they can benefit from.
- 38:24So it's exactly the same dynamics, but they can play into
- 38:28different strategies. And, and so when I think about
- 38:31like each company, each each for a model company, I feel like
- 38:35they are trying to carve out their niche in the universe.
- 38:38Some are fuzzier than others, but I do think the clearest
- 38:41distinction is when you look at the projected revenue of a
- 38:45specifically Anthropic versus Open Eye.
- 38:50I think it was like 2027 revenue estimates that that somebody had
- 38:53leaked or put out or whatever. But it was basically like, hey,
- 38:55by 2027, Anthropic believed that they would have I think like 5X
- 39:00the API revenue that Open AI was projecting for.
- 39:03And it wasn't Anthropic crapping on Open AI.
- 39:05It was Opening Eyes forecast for 27 compared to Anthropic's
- 39:07forecast. And Anthropic assumes they're
- 39:10going to have five times as much API revenue than Open AI.
- 39:13And it becomes really obvious that Open AI, whether they meant
- 39:17to do this or not, is going to become largely a consumer driven
- 39:23business. Like ChatGPT is their cash cow
- 39:25right now. And whereas Anthropic is trying
- 39:28to squarely plant itself in the API where it's sort of the B to
- 39:32B world. That's a clear distinction
- 39:35that's defining a lot of the decisions those companies are
- 39:37making. You know, I love, love Anthropic
- 39:39and they're very competitive, but there is the, the sort of
- 39:42infrastructure API business is like under threat from open
- 39:47source in a way that having a dominant consumer application
- 39:51is, is a type of Moat that we're all all familiar with.
- 39:54It feels like besides open AI, having a lot of users, everybody
- 39:58else is sort of like we need to run faster than our competition
- 40:02to have sort of incrementally better stuff.
- 40:05Obviously over time you build customer relationships, sales,
- 40:09sort of product differentiation. But it is totally, if I mean
- 40:12that's absolutely the the bear case on Anthropic is that
- 40:17progress is just too quick for you to get in and establish any
- 40:20real meaningful like enterprise install base.
- 40:22Whereas I think on the consumer side that's different because
- 40:26when you see these surveys that say something like 6575% of
- 40:30adults in the US like not in tech, but like period, everyone
- 40:34has used an AI product. The vast majority of them are
- 40:37ChatGPT. And I think about this all the
- 40:39time that like when we look back at the end of 2022 when Chad TPT
- 40:43came out, What's so interesting is when you talk to the people
- 40:45who are the sort of experts in this world at that point in
- 40:49time, that was not a technological breakthrough.
- 40:52Like it wasn't, that wasn't something that a lot of the
- 40:54companies who were operating in 2022 couldn't have done.
- 40:57Like they could have put a chat interface on their model, but
- 41:00they didn't open AI put out it as like almost an experimental
- 41:03tool to say, hey, in this neat, you can kind of play around with
- 41:06it. And it exploded because it was
- 41:07people's first direct experience with generative text.
- 41:10Everybody was too afraid of their tails, and only open AI
- 41:13was willing to unleash AI onto us all in a cavalier way.
- 41:17Yeah. Yeah, well, I think they even
- 41:19think it was an accident, right? They like they didn't expect it
- 41:21to do that either. Right, But I mean, clearly, I
- 41:23think some of the break of anthropic was, you know, the
- 41:26roll out of Chachi BT and everything.
- 41:29I, I do think there's a lot of room for consumer adoption just
- 41:35because I, I feel like I'm begging people to use open AI or
- 41:38to use Chachi BT in some case. I'm like, like my family members
- 41:41are like, it's crazy. I'm like, you know, super
- 41:43excited about O3 spending a lot of time on it and it it's, you
- 41:47know, people, these companies clearly need to find ways to
- 41:49bring the product to people in a way that doesn't take as much
- 41:53work. And that's going to be a huge
- 41:54build up going back to venture because that's really the focus
- 41:56of this conversation. Do you think what we're seeing
- 41:59with the models is playing out in other categories or do you
- 42:04think this category is unique where VCs invest all over the
- 42:09place? So I had one of my much smarter
- 42:14than me friends kind of articulate this analogy of like
- 42:18when you think about the.com, there was sort of, you know, the
- 42:22almost the three layers of what was happening.
- 42:25And that really drove like the distribution of the Internet.
- 42:28And you basically have there were folks on the infrastructure
- 42:31side you're building physical, you know, almost like compute
- 42:34and stuff like that. Then you have the networking
- 42:37where it's the Cisco's of the world getting to broaden out and
- 42:40get access to the underlying tech.
- 42:42And then you had the applications, right, the
- 42:45pets.com and stuff like that. The difference there when you
- 42:48think about like how that is similar to today is that you
- 42:51should you look at the chips company and chips company once
- 42:53again are ripping, right. The sort of underlying
- 42:54infrastructure is still powerful regardless of what's getting
- 42:56built on top of it. That's very powerful.
- 42:58And so I think that segment has more problems with, you know,
- 43:02geopolitical stuff than it does like distribution.
- 43:05But what's what's really unique about this point in time that
- 43:08makes it really difficult for venture investors is that in
- 43:10the.com you had networking companies that were like, we're
- 43:13just trying to get this out into the world.
- 43:14And then you had applications coming and saying we're going to
- 43:16get this package and deliver it to people as a service.
- 43:20Today you have the networking layer, which is effectively the
- 43:23model companies also building the application.
- 43:26So it's effectively as if Cisco was launching pets.com, it would
- 43:29be really difficult for pets.com to compete with that.
- 43:32You know, and you know, maybe it falls apart because it wasn't
- 43:34the underlying thing. But today it is like AI is the
- 43:37underlying aspect of that product.
- 43:39And so when you see folks like open AI buying Winsurf like that
- 43:43is a it's a perfectly logical extension, right?
- 43:46You see this? To reframe what you're saying a
- 43:49little bit, you're saying the non monogamy or the polyamory
- 43:52with foundation models, it is polyamory with everything
- 43:55because the models sort of are competing with everything that's
- 43:58investable and interesting right now.
- 44:01And so it's like, yeah, if you're doing applications,
- 44:02infrastructure, whatever, inevitably you're going to be a
- 44:05collision course with your model investment.
- 44:08That that seems OK. It's just I think it's just the
- 44:10pure model on model that feels like, Oh my God, like whatever,
- 44:14you know, how many, how many closed source models do we need?
- 44:16I guess is a backdrop question. Well, and I guess the question,
- 44:20I think the, the thing that people are sort of struggling
- 44:23with is that like the, the not just the progress of individual
- 44:29companies, but the progress from model to model.
- 44:31I mean, you look at like people's reactions from Claude
- 44:353.5 to 3.7 and it's like that's a miss.
- 44:38That's a misstep that that like could make or break the game for
- 44:41a while, right? Like that it can happen that
- 44:43quickly. Or like, you know, I think
- 44:45somebody else tweeted this of this like constant like we're so
- 44:48back. It's so over cycle of like, oh,
- 44:50like I actually think, you know, 4.5 is not very good.
- 44:53And then it's like 03 is incredible and like it's so
- 44:56quick with each model that it's so much harder that like at the
- 44:59end of the day, yeah, if you're a massive multi billion dollar
- 45:02firm, you've you've got to go get exposure to everything
- 45:05because you don't know what's going to come out on top.
- 45:06This is what we want from capitalism, isn't it?
- 45:09Everybody's competing, lots of money flowing into an
- 45:11interesting space. And like, yeah, you got to be
- 45:14the best of the best if you want to extract value.
- 45:17I don't know. I'm not.
- 45:18And it's a category where I'm like happy for the world to burn
- 45:22some capital, you know, trying to crackitsoidontknow.com
- 45:27theproblemwith.com was like there was no distribution.
- 45:30It's like they didn't they you couldn't, there wasn't the
- 45:32customer base when you set up the Internet now in all the
- 45:36other bubbles, it's like, well, there's still a distribution to
- 45:38quickly get a bunch of customers.
- 45:40I mean you, you invoke.com, are you, are you worried about like
- 45:44a big bubble? So I my biggest concern is this
- 45:48idea of like, like hype cycles are built on momentum value
- 45:55cycles, if you will. Like actual value creation
- 45:58happens when you're able to push through all of the hype and all
- 46:01the excitement, all the experimentation and get to the
- 46:04actual creation of business value.
- 46:06And if there is underlying business value that's consistent
- 46:09and replicable and changing the way an organization works, then
- 46:11you can capture some of that value.
- 46:14My sense is that the volume of capital and the volume of hype
- 46:20is not commensurate with the our ability to articulate and
- 46:24execute on actual business value like that.
- 46:27Does not those feel wildly? And part of what made that
- 46:30possible this time is every enterprise in the world was
- 46:33willing to spend money initially to say, oh, how do we figure out
- 46:37AI? But that could be a sort of
- 46:39false signal on how much they stick around when they don't get
- 46:42the value. And that goes back to your
- 46:44question about like, you know, we, we the, the biggest thing
- 46:47that we have to unlock is we have to figure out how do we get
- 46:51people to use this more consistently in a way that
- 46:53creates value. That's not just true of like you
- 46:56and I trying to convince our mom to use ChatGPT or whatever.
- 46:59That's also true. Yeah, we look too similar.
- 47:04We can't be saying something like that, but like that that
- 47:08kind of stuff like it's from the consumer side is also true on
- 47:12the business side. And I think all the time about
- 47:14like, yeah, we've seen crazy like Rev scale UPS, right?
- 47:18But it's like, OK, what are probably the three companies
- 47:19that have made the most money from this this pop?
- 47:23And it's like NVIDIA, which is like, OK, demand, demand, people
- 47:25need it, whatever open AI, which is like innate offering it to
- 47:28people to play around with and use.
- 47:31And then like Accenture, who's like making an ungodly amount of
- 47:34money helping people figure out like, what do what do we do with
- 47:37this? Like, how does this actually
- 47:39impact our business? But like outside of that, like
- 47:41there haven't been a ton of case studies, at least that I've seen
- 47:44of like, oh, like John Deere is suddenly ramping up.
- 47:49You hear some stories like I think it was the what the Klarna
- 47:51story of like. Yeah, Klarna and Shopify, but
- 47:53some of it feels very aspirational, like they want to
- 47:55be seen as getting it. Yeah, but it's like.
- 47:58I want, yeah, totally. Listen, I would love for that to
- 48:01happen. Like I think that like, yeah,
- 48:02the fears of like job destruction and stuff like that.
- 48:04Like I agree that it's the same as like Willy Wonka or what's
- 48:08his name Charlie and the talking factor.
- 48:09He's dad getting displaced by a robot and then getting a job
- 48:12fixing the robot that took his job.
- 48:13Like there's there's lots of opportunities for evolution.
- 48:16I'm not afraid of that. I want that to happen.
- 48:19The question is just is it capable of doing that?
- 48:22And I don't, I don't know that we're there, which could set us
- 48:25up for a very painful pop. But like you look back at
- 48:28the.com and like Amazon and Google camethroughthe.com, like
- 48:32there will be businesses that get built that are still
- 48:34generational businesses despite the fact that they will
- 48:37participate in short term pain. Sam Allman, Opening eye, did
- 48:42reportedly try to scare investors off investing.
- 48:45You know, he had that list that was reported.
- 48:48You think that was just like a total failure or any
- 48:50observations about that? Well, So what was interesting is
- 48:53if I remember correctly, one of the biggest sort of stinks that
- 48:56was made about that list was Glean was was not people not
- 49:00investing in Glean and an open AI.
- 49:02And here's the thing, like, so to go back to the question of
- 49:05like you're not as worried about the infidelity for between like
- 49:10the model and application layer. Like I actually don't think that
- 49:13you can distinguish because it, it at the end of the day, it
- 49:17increasingly it seems that it is going to be the same thing
- 49:21because you're not like the again, like your point, like the
- 49:24API revenue, it's really difficult to believe that that's
- 49:27long term defensible. Like I thought it was
- 49:29fascinating when I dug into all of the Cogen agents right where
- 49:33you look at like the cursors and windsurfs and Lovables and bolts
- 49:36and replics of the world and stuff like that.
- 49:38They're all using Clod. Like they at the time they were
- 49:40using clod 3.5. This is a couple months ago, but
- 49:43they were all using that model. But it's like tomorrow we can
- 49:45all switch over to this one because now Gemini is the better
- 49:48code Gen. model or whatever like that.
- 49:50Progress means that it's going to be incredibly difficult for
- 49:53any of these companies to justify their existence at the
- 49:56model layer, which means that, well, where do you make money?
- 50:00You go pay $3 billion for windsurf because that's where
- 50:03you're going to make money because people want to have a
- 50:05thing that they use. The bigger question there then
- 50:07for, for windsurf users is like, OK, wait, but am I going to be
- 50:10stuck now then on GPT models? And it's like not if they want
- 50:14to keep their that installed base like they're going to have
- 50:16to let people use different models.
- 50:19And on that acquisition, there is people are like, oh, you can
- 50:22also train your models on the behavior on windsurf.
- 50:24So there's an argument that, you know, you could be open AI,
- 50:28leave it, leave customers allowed to use whatever model
- 50:30they want, and then learn a lot to then make your model better
- 50:33so that eventually they want to use your model again.
- 50:37Yeah. But if I think about like, OK,
- 50:38if I'm open AI and I want to rule the world and I'm not
- 50:42because I do think also like again going back to this, like
- 50:44some of these companies will differentiate.
- 50:46I think that SSI is going to be sort of go off on this like AI
- 50:53safety, like let's save the planet stuff.
- 50:55And it's like, let's see how that goes.
- 50:56I don't know what that's going to evolve to look like, but it's
- 50:59clear that open AI is not, you know, again, maybe the narrative
- 51:02and the marketing is making AGI accessible for everybody or
- 51:05whatever. It's like, right, They want to
- 51:07own the world and they want to make a lot of money and they're
- 51:08trying to be a for profit so that they can raise more money
- 51:10to make more money. Like they're clearly trying to
- 51:12own the world. It's like, OK, if I'm open AI
- 51:15and I need to justify a hundreds of billions of dollar valuation,
- 51:20how am I going to do that? It's not going to be, I promise
- 51:23I'll keep making the next best model and like a constant, you
- 51:26know, a hamster wheel, it's going to be, I'm going to be the
- 51:29thing for your workplace and I'm going to be the thing for your
- 51:32planning and your personal life and your AI agent and whatever.
- 51:35I, I was going to say initially, open AI is a uniquely
- 51:39problematic juggernaut from an investor perspective because it
- 51:42is a non profit and you know, because it had the ousted CEO
- 51:47and everything. It's like, I can see why you'd
- 51:49want to sort of hedge your bets that maybe it's sort of like a
- 51:52shakier dominant player than in past generations, but then you
- 51:56reflect and you're like, wait, Uber add the Travis stuff like
- 52:00Facebook. Mark Zuckerberg was like firing
- 52:02his whole executive team and had to like pivot to mobile in some
- 52:05ways, like the juggernaut says company was never like so
- 52:09clearly anointed as we we write the story, but you know, going
- 52:13through it, you're like, man, opening eye could totally failed
- 52:16to become a for profit and then I better have an investment
- 52:20somewhere else. I wanted to ask and I guess let
- 52:23this be sort of the last question.
- 52:27What without talking your own book, what should a founder do
- 52:31based on this information or like, what do you think?
- 52:34Like if everybody's investing in everybody, like to I need to go
- 52:36to a CO2 and thrive or it's like I need to just accept it or
- 52:41what? What do you What advice would
- 52:42you give to a founder faced with this polyamory?
- 52:47So increasingly, and I don't know that I'd go so far as to
- 52:52say that polyamory is rampant in every category forever, at every
- 52:55stage for every firm. Like everybody's, everybody's
- 52:58different. But what we're seeing here is a
- 53:00symptom of something that you and I have talked about a lot
- 53:02before of this like sort of capital agglomerator, like rise
- 53:06of the, you know, the black stones of our industry, right?
- 53:08Like that's what's happening. This is a Subs.
- 53:10This is a. For lack of a better term.
- 53:14Yeah, like this is a sub system of that.
- 53:17And I think that in in thinking about that because it is true
- 53:21that they're going to do more, there's going to be more
- 53:22applications that they're investing in that may be
- 53:24competitive and whatever because they're just trying to play
- 53:26capital. What founders need to understand
- 53:28and what I've spent three years writing about on my on my sub
- 53:31sack, is that increasingly venture firms, at least the
- 53:35business model venture firm has increasingly become more focused
- 53:38on asset management. That's not the marketing.
- 53:42They still market themselves like they are the scrappy, you
- 53:45know, John Doors of the world from the early 2000s or whatever
- 53:48and the benchmarks and stuff like that.
- 53:49But that's, that's the marketing.
- 53:51The business model of most of these firms has become asset
- 53:54management. And if they're going to build
- 53:56themselves as asset managers, you as a founder need to treat
- 54:01them as asset managers. They are not your, even if
- 54:04they're on your board, they are not.
- 54:06Just because they're on your board does not mean that they're
- 54:07your buddy. It does not mean that they're
- 54:09your therapist or your sounding board.
- 54:11There are firms and there are investors that you can go get
- 54:14who will be your confidante and your friend and support and
- 54:17whatever. And their firms are set up more
- 54:19aligned with your outcomes. These larger firms are not that.
- 54:23And so people need to take a, like a crash course in how
- 54:28public companies manage the asset managers that buy their
- 54:31stock, because that's the way you need to treat some of these
- 54:34firms is like that. And that goes to like things
- 54:36like information rights and strategy and, and how and who
- 54:39you let into what round and what percentage of ownership and how
- 54:42much they control of your parata and your, you know, majority
- 54:46voting rights and stuff like that.
- 54:47Like you need to manage them as the asset managers that they are
- 54:51and be very, very careful. It's not bad.
- 54:53It's a good weapon that you can use in a in a in a big fight,
- 54:57but you have to be really careful.
- 55:00Kyle Harrison Contrary, thanks so much for joining us.
- 55:03Thanks for having me.