AGI, Seriously + Foundation Model Polyamory

Summary

An episode of Eric Newcomer’s The Newcomer Podcast with Kyle as guest, built around Kyle’s viral “foundation model polyamory” chart — a screenshot mapping which venture firms have backed which competing frontier-model companies (OpenAI, Anthropic, xAI, SSI, Thinking Machines). The framing joke: infidelity among venture capitalists. What triggered it was an xAI fundraise announcement full of “we’re so excited to back Elon again” language from firms Kyle knew were already in OpenAI or Anthropic — and an LP account (“Endowment”) asking, “is backing competitive companies allowed now in venture?” Kyle went and checked the cross-section: yes, clearly.

The middle of the conversation is a firm-by-firm read of strategy. a16z is the most promiscuous — Martin Casado’s “why not all of them?” coverage play, buying exposure across every AI segment. Sequoia is active but more protective of its reputation at the earliest stages (Kyle recounts the old Stripe-vs-Finix episode, where Sequoia treated a competitive check as “just a donation”), with the observation that at the later stages you’re largely just garnering exposure to assets, not picking. Lightspeed Venture Partners is playing deliberate catch-up (Anthropic as one of its biggest positions, plus Mistral and xAI). The serial monogamists — Founders Fund, Thrive Capital, and Khosla Ventures — run concentrated books and back one winner up the truck (all-in on OpenAI), which fits Peter Thiel’s “you want to be the monopoly that chokes out the oxygen” logic; conspicuously absent are General Catalyst and NEA.

From there Kyle draws the real distinction: the model companies are differentiating. OpenAI is becoming a consumer business (ChatGPT is the cash cow), while Anthropic is planting itself in the API / B2B world — Kyle cites Anthropic’s own 2027 forecast of ~5x OpenAI’s projected API revenue. The bear case, which Kyle grants, is that API revenue isn’t long-term defensible against open source (the codegen agents — Cursor, Windsurf, Lovable, Bolt, Replit — can swap Claude for Gemini tomorrow), which is exactly why OpenAI paying ~$3B for Windsurf (formerly Codeium) is a logical move: you make money where people actually use a thing.

The connective idea is a dot-com analogy (a friend’s): the era had infrastructure, a networking layer (the Ciscos), and applications (Pets.com). Today the networking layer — the model companies — is also building the application. “It’s as if Cisco was launching Pets.com.” So the polyamory with models is really polyamory with everything, because the models collide with everything investable. On the bubble question, Kyle separates hype cycles (built on momentum) from value cycles (actual, replicable business value), and worries the volume of capital and hype is not commensurate with articulated business value — the biggest earners so far are Nvidia, OpenAI, and Accenture (helping enterprises figure out what to do), while enterprise case studies (Klarna, Shopify) feel “very aspirational.”

It closes on the founder takeaway — Kyle’s three-years-on-Substack thesis: the polyamory is a symptom of the rise of the capital conglomerators (the Blackstones of venture). The business model of most large firms has become asset management, even as the marketing still sounds like the scrappy John Doerr era. So founders should treat these firms as asset managers — manage information rights, pro rata, board control, and who you let into a round accordingly. “Just because they’re on your board does not mean they’re your buddy.”

Transcript

Eric Newcomer: Hey, it’s Eric Newcomer. I’m here with Kyle Harrison of Contrary, former investor at Index and Coatue. Thanks for joining me.

Kyle Harrison: Thanks for having me.

Eric Newcomer: We are here to talk about one specific thing: infidelity among venture capitalists. You had a viral image in the tech world that shows all the different investments VCs are making into foundation model companies — and a lot of the VCs are making investments into competitors. So, fun topic. Do you want to start off with the impetus for making the graphic, and then walk us through what you found?

Kyle Harrison: It was kind of a vibe that I had been noticing over the course of several months, where I’d see these fundraising announcements and it would be a big firm — I think it was xAI that really triggered me. I saw people who were like, “We’re so excited to back Elon again, generational company building,” all this stuff about how great the business was. And I was like, wait, weren’t these people invested in OpenAI or Anthropic or whatever? So I went and looked — and, oh, they did. Oh, I guess people are making a lot of cross, multi-layered investments.

And then a non-account that I really like, who’s one of the best LP accounts on here — this guy “Endowment” — he was like, “Okay, wait, is backing competitive companies allowed now in venture?” And I was like, let me go look at the cross-section. And it’s like, yeah, I guess so. Yes, clearly, is the answer.

Eric Newcomer: Do you want to walk us through it? Like, with Andreessen you have SSI — Ilya’s company — xAI, OpenAI, and Mira’s Thinking Machines. What are some of the other interesting ones?

Kyle Harrison: Yeah. I think the biggest thing — and even some of these I missed; this was not necessarily a purely academic exercise —

Eric Newcomer: But what’s crazy is, you made a screenshot, and now you’re the scholar on this topic. That’s the level of thought leadership on the internet. No — this is much more baked than most VC takes, with just a screenshot. Anyway, go ahead.

Kyle Harrison: Yeah. Harvard sent me a degree in the mail.

Eric Newcomer: Yeah, exactly.

Kyle Harrison: I feel like the biggest thing is, you can kind of clearly see — and somebody actually retweeted this and pointed it out — that it’s a very specific strategy. The one who’s most active is Andreessen, which is consistent; Andreessen is super broad, investing in a lot of different companies. Martin Casado, who I love, said on stage at Cerebral Valley — when someone asked, “How do you know which segment of AI to invest in?” — he just said, “Why not all of them?” They’re openly going for coverage at this point. And apparently even in the key categories, with competitors.

So Andreessen is everywhere. You’ve got Sequoia, who is also quite active — a little less than Andreessen, so a little more demure, or whatever. And then you have firms like Founders Fund—

Eric Newcomer: Sequoia just did it too. I mean, Sequoia is interesting, because — Affirm and Klarna — they have these famous ones where they’ve done both. They try, I think, to be loyal, but… I don’t know, it’s getting pretty broad; they’re in a lot of them. I was going to try to do some spin, like, oh, they were the ex in that relationship with the Klarna world, and then they invest in the leader, which is clearly OpenAI in my opinion. I don’t know.

Kyle Harrison: Yeah, be careful with any Sequoia analysis, because — I don’t know if you remember this — but they invested in Finix back in the day, on the payments side, and Stripe was pretty upset about it. So Sequoia was literally like, “Our bad, keep the money, we won’t take an equity stake, it’s just a donation.” So they’re definitely more thoughtful about where they’re investing.

To Sequoia’s credit, one of the things they’ve been very deliberate about is being more selective in the earlier stages. Somebody else pointed this out in a retweet: once it gets into the later stages, you’re largely just garnering exposure to assets. It’s not as much [about picking] once these guys are raising $10 billion-plus rounds — you’re really just getting exposure. I think Sequoia is a little more protective of its reputation at the earliest stages.

Lightspeed is another one. Anthropic is going to be one of their biggest positions overall — I dug into their returns, and that was an eye-opener. And then they’re in Mistral, xAI, and I think one other that somebody pointed out I’d missed. Lightspeed has also been super active, and I’d argue they’re really trying to deliberately demonstrate that they want to be in this bucket of the Andreessens and General Catalysts of the world — just be very, very active.

Eric Newcomer: What’s funny — where is General Catalyst? GC and NEA? If you’re going to do the Andreessen-style blast-them-all approach, shouldn’t you be in the defining category of our time?

Kyle Harrison: It’s a good question, and I’ll defer again to the lack of diligence. They might be in five of these, and I just didn’t want to look up another Crunchbase profile.

Eric Newcomer: Look up everyone. Okay.

Kyle Harrison: They’ll be like, “No, we’re in this one.” I don’t know.

Eric Newcomer: Who are the serial monogamists here? Of the big-brand firms, who do you think has stuck to one bet?

Kyle Harrison: Well, this was actually a sort of oversight on my side in making the chart. Originally it was supposed to show the cross-section, but I started at the bottom — okay, OpenAI, who’s in that, boom boom — and worked my way around, adding firms as I went. What I realized when I posted it and it popped off is that I’d left Founders Fund on there despite the fact that they are actually monogamous in this particular strategy. The other one I didn’t put on but that got called out is Thrive.

Going back to this being a pretty indicative sampling of firm strategies: Founders Fund and Thrive are famous for being really dedicated to running an incredibly concentrated portfolio and just doubling down, tripling down into their winners. And Founders Fund, despite being perceived as early and perceptive — Airbnb was a Series C for them — some of these, they’re like, “What is the momentum company that’s going to define the generation, and how can we back up the truck into it?” And clearly they see OpenAI here.

Eric Newcomer: Well, and I think that’s one of the reasons they’re not afraid to admit when they didn’t get there as early as possible.

Kyle Harrison: Right. And probably in that same vein of monogamy, I think Khosla is another one — they’ve been all OpenAI, all day, for as long as possible.

Eric Newcomer: It’d be insane to diversify away. It’s one thing — I think a consideration for a lot of these is they’re getting into OpenAI at such a high price that it fits into the growth story we’re talking about, where they were not traditional venture investors in OpenAI. They’re just like, “We can’t miss the company of our generation.” And then they’re like, “Well, let’s try to do real venture investments in other model companies if we get a shot at it.”

Kyle Harrison: Yeah. So to Founders Fund’s credit, even if they’re not there at the earliest days, they’re willing to acknowledge, “Hey, this has clearly broken out as a category definer.” I’ve heard Trae Stephens say this multiple times: their perspective is that there’s one company — if you’re going to invest in space and you didn’t invest in SpaceX, you probably lost money; if you want to invest in defense and you didn’t invest in Anduril, you’re probably going to lose money. That’s their thought process. So even if they’re getting there a little late, they’re going to back up the truck.

Eric Newcomer: And that’s Peter Thiel’s whole thing — you want to be a monopoly business. You don’t want to be a company that’s creating a category, because then you have all these competitors. You want to be the company in a category that chokes out all the oxygen for everybody else.

Kyle Harrison: That’s right. I also think there’s a narrative here — we can talk about this if you want — that there is increasing differentiation. It’s very subtle, and it’s going to start overlapping again, but there is some semblance of different companies trying to prioritize different places.

Eric Newcomer: No, let’s talk about it — especially in light of OpenAI potentially purchasing Windsurf, formerly known as Codeium. That speaks to them demolishing everything, but also to the idea that if you’re an investor it’s hard to know where these companies are going to go. They’re going to lean into where they’re successful. Foundation models in particular — you could become an application, you could be infrastructure. It’s hard to know. So no one is saying VCs shouldn’t invest in Windsurf, and if Windsurf becomes part of OpenAI’s turf, then the floodgates are open. What would you say about that?

Kyle Harrison: I think it’s very similar to what happened with the hyperscalers. If you look at the Amazons and Googles and Microsofts of the world, the difference is that their amalgamation of a bunch of different faces happened long after they’d been founded, long after they’d gone public. And even think about OpenAI launching a social app — a friend of mine tweeted, “So are network effects dead? Does nothing mean anything anymore? All the businessy stuff we’ve learned is nonsense?” And it’s like, if you think about it, it’s not that any of the fundamental principles have changed. It’s that the velocity — the speed of the thing happening — is just so much faster than it’s ever been.

At the end of the day, if OpenAI does launch a social app and it turns out successful, it’ll be for the same reasons that when Meta launched Threads they also got up there pretty quick: they have a big, broad, established install base they can benefit from. So it’s exactly the same dynamics, but they can play into different strategies.

When I think about each model company, I feel like they’re each trying to carve out their niche. Some are fuzzier than others, but the clearest distinction is when you look at the projected revenue of Anthropic versus OpenAI. There were 2027 revenue estimates somebody leaked or put out — and Anthropic believed that by 2027 it would have something like 5x the API revenue OpenAI was projecting. And it wasn’t Anthropic crapping on OpenAI; it was OpenAI’s own forecast for ‘27 compared to Anthropic’s. Anthropic assumes it’s going to have five times as much API revenue as OpenAI. So it becomes really obvious that OpenAI, whether they meant to or not, is going to become largely a consumer-driven business — ChatGPT is their cash cow right now — whereas Anthropic is trying to squarely plant itself in the API world, the B2B world. That’s a clear distinction that’s defining a lot of the decisions those companies are making.

Eric Newcomer: You know, I love Anthropic and they’re very competitive, but the infrastructure/API business is under threat from open source in a way that having a dominant consumer application isn’t — a consumer app is a type of moat we’re all familiar with. It feels like, besides OpenAI having a lot of users, everybody else is sort of like, “We need to run faster than our competition to have incrementally better stuff.” Obviously over time you build customer relationships and sales motions and product differentiation, but…

Kyle Harrison: It totally — I mean, that’s absolutely the bear case on Anthropic: that progress is just too quick for you to get in and establish any real, meaningful enterprise install base. Whereas on the consumer side, that’s different, because when you see these surveys that say something like 65 to 75% of adults in the US — not in tech, but period — have used an AI product, the vast majority of them are ChatGPT.

I think about this all the time: when we look back at the end of 2022 when ChatGPT came out, what’s so interesting is that if you talked to the experts at that point in time, that was not a technological breakthrough. It wasn’t something the companies operating in 2022 couldn’t have done — they could have put a chat interface on their model, but they didn’t. OpenAI put it out almost as an experimental tool: “Hey, isn’t this neat, you can play around with it.” And it exploded, because it was people’s first direct experience with generative text.

Eric Newcomer: Everybody was too afraid of their tails, and only OpenAI was willing to unleash AI onto us all in a cavalier way.

Kyle Harrison: Yeah. Well, I think they even think it was an accident, right? They didn’t expect it to do that either. Clearly some of the breakup with Anthropic was, you know, the rollout of ChatGPT and everything. I do think there’s a lot of room for consumer adoption, because I feel like I’m begging people to use ChatGPT in some cases. My family members — I’m like, it’s crazy. I’m super excited about o3, spending a lot of time on it. These companies clearly need to find ways to bring the product to people in a way that doesn’t take as much work, and that’s going to be a huge buildout.

Eric Newcomer: Going back to venture, because that’s really the focus of this conversation — do you think what we’re seeing with the models is playing out in other categories, or is this category unique, where VCs invest all over the place?

Kyle Harrison: I had one of my much-smarter-than-me friends articulate this analogy. When you think about the dot-com era, there were almost three layers that drove the distribution of the internet. You had the infrastructure side — building physical compute and things like that. Then the networking — the Ciscos of the world, broadening out access to the underlying tech. And then the applications — the Pets.coms and so on.

The way that’s similar to today: the chips companies are ripping again — the underlying infrastructure is still powerful regardless of what gets built on top of it. That segment has more problems with geopolitics than with distribution. But what’s really unique about this point in time, and what makes it really difficult for venture investors, is that in the dot-com era you had networking companies that were just trying to get the technology out into the world, and then applications came and said, “We’re going to package this and deliver it to people as a service.” Today, the networking layer — effectively the model companies — is also building the application. So it’s as if Cisco were launching Pets.com. It would be really difficult for Pets.com to compete with that. And so when you see folks like OpenAI buying Windsurf, that’s a perfectly logical extension.

Eric Newcomer: To reframe what you’re saying a little: the non-monogamy — the polyamory — with foundation models is polyamory with everything, because the models are competing with everything that’s investable and interesting right now. So if you’re doing applications, infrastructure, whatever, inevitably you’re on a collision course with your model investment. And that seems okay. It’s just the pure model-on-model that feels like, “Oh my god.” How many closed-source models do we need, I guess, is the backdrop question.

Kyle Harrison: Well, the thing people are struggling with is that the progress — not just of individual companies, but from model to model. You look at people’s reactions from Claude 3.5 to 3.7, and it’s like, that’s a misstep that could make or break the game for a while. It can happen that quickly. Somebody tweeted about this constant “we’re so back / it’s so over” cycle — “oh, actually 4.5 is not very good,” and then “oh, o3 is incredible.” It’s so quick with each model that it’s so much harder — so at the end of the day, if you’re a massive multi-billion-dollar firm, you’ve got to go get exposure to everything, because you don’t know what’s going to come out on top.

Eric Newcomer: This is what we want from capitalism, isn’t it? Everybody’s competing, lots of money flowing into an interesting space, and you’ve got to be the best of the best if you want to extract value. It’s a category where I’m happy for the world to burn some capital trying to crack it. The problem with dot-com was there was no distribution — there wasn’t the customer base when you set up the internet. In all the other bubbles, there’s still distribution to quickly get a bunch of customers. Are you worried about a big bubble?

Kyle Harrison: My biggest concern is this idea that hype cycles are built on momentum. Value cycles, if you will — actual value creation — happen when you’re able to push through all the hype and excitement and experimentation and get to the actual creation of business value. And if there’s underlying business value that’s consistent and replicable, that changes the way an organization works, then you can capture some of that value. My sense is that the volume of capital and hype is not commensurate with our ability to articulate and execute on actual business value. Those feel wildly out of step.

Part of what made that possible this time is that every enterprise in the world was willing to spend money initially to figure out AI — but that could be a false signal on how much they stick around when they don’t get the value. And that goes back to your question: the biggest thing we have to unlock is how to get people to use this consistently in a way that creates value. That’s not just true of trying to convince our mom to use ChatGPT — it’s also true on the business side.

I think all the time about — we’ve seen crazy scale-ups, but what are the three companies that have made the most money from this pop? Nvidia, which is demand — people need it. OpenAI, offering it to people to play around with and use. And then Accenture, making an ungodly amount of money helping people figure out, “What do we do with this? How does it actually impact our business?” Outside of that, there haven’t been a ton of case studies I’ve seen of, “Oh, John Deere is suddenly ramping up.” You hear some stories — I think it was the Klarna story, Klarna and Shopify.

Eric Newcomer: But some of it feels very aspirational, like they want to be seen as getting it.

Kyle Harrison: Yeah, totally. Listen, I would love for that to happen. The fears of job destruction and so on — I agree it’s the same as Charlie and the Chocolate Factory’s dad getting displaced by a robot and then getting a job fixing the robot that took his job. There are lots of opportunities for evolution; I’m not afraid of that, I want that to happen. The question is just, is it capable of doing that? And I don’t know that we’re there, which could set us up for a very painful pop. But you look back at dot-com, and Amazon and Google came through — there will be businesses built that are still generational, despite participating in short-term pain.

Eric Newcomer: Sam Altman / OpenAI reportedly did try to scare investors off investing — he had that list that was reported. Do you think that was a total failure, or any observations about it?

Kyle Harrison: What was interesting is, if I remember correctly, one of the biggest stinks made about that list was Glean — people not investing in Glean and OpenAI. And here’s the thing: to go back to the question of not being as worried about the infidelity between the model and application layer — I actually don’t think you can distinguish them, because increasingly it seems it’s going to be the same thing. The API revenue — it’s really difficult to believe that’s long-term defensible.

I thought it was fascinating when I dug into all the codegen agents — the Cursors and Windsurfs and Lovables and Bolts and Replits of the world. A couple months ago they were all using Claude 3.5. But tomorrow they can all switch over to whatever’s better — now Gemini is the better codegen model, or whatever. That progress means it’s going to be incredibly difficult for any of these companies to justify their existence at the model layer. Which means — well, where do you make money? You go pay $3 billion for Windsurf, because that’s where you’re going to make money, because people want to have a thing that they use.

The bigger question then, for Windsurf users, is, “Okay, wait, am I now going to be stuck on GPT models?” And it’s like — not if they want to keep that installed base. They’re going to have to let people use different models.

Eric Newcomer: And on that acquisition, people are like, “You can also train your models on the behavior on Windsurf.” So there’s an argument that OpenAI could leave it open — let customers use whatever model they want — and then learn a lot to make your own model better, so that eventually they want to use your model again.

Kyle Harrison: Yeah. But if I’m OpenAI and I want to rule the world — and I do think, again, some of these companies will differentiate. I think SSI is going to go off on this AI-safety, “let’s save the planet” stuff, and it’s like, let’s see how that goes; I don’t know what that’s going to evolve into. But it’s clear OpenAI is not that. Maybe the narrative and marketing is “making AGI accessible for everybody,” but they want to own the world and make a lot of money, and they’re trying to be a for-profit so they can raise more money to make more money. They’re clearly trying to own the world.

So if I’m OpenAI and I need to justify a hundreds-of-billions valuation, how am I going to do that? It’s not going to be “I promise I’ll keep making the next best model” on a constant hamster wheel. It’s going to be “I’m going to be the thing for your workplace, and the thing for your planning and your personal life and your AI agent.”

Eric Newcomer: I was going to say — initially OpenAI is a uniquely problematic juggernaut from an investor perspective, because it’s a nonprofit, and because it had the ousted-CEO drama. So I can see why you’d want to hedge your bets that it’s a shakier dominant player than in past generations. But then you reflect and you’re like — wait, Uber had the Travis stuff; Facebook, Mark Zuckerberg was firing his whole executive team and had to pivot to mobile. In some ways the juggernaut was never so clearly anointed as we write the story after the fact. But going through it, you’re like, “Man, OpenAI could totally fail to become a for-profit, and then I better have an investment somewhere else.”

I wanted to ask — and let this be the last question — without talking your own book, what should a founder do based on this information? Do I need to go to Coatue and Thrive? Do I just accept it? What advice would you give a founder faced with this polyamory?

Kyle Harrison: I don’t know that I’d go so far as to say polyamory is rampant in every category, at every stage, for every firm — everybody’s different. But what we’re seeing here is a symptom of something you and I have talked about a lot: this rise of the capital conglomerator, the Blackstones of our industry. This is a subsystem of that.

What founders need to understand — and what I’ve spent three years writing about on my Substack — is that increasingly the business model of the venture firm has become more focused on asset management. That’s not the marketing. They still market themselves like they’re the scrappy John Doerrs of the world from the early 2000s, or the Benchmarks. But that’s the marketing. The business model of most of these firms has become asset management. And if they’re going to build themselves as asset managers, you as a founder need to treat them as asset managers.

They are not your — even if they’re on your board, they are not your buddy just because they’re on your board. It does not mean they’re your therapist or your sounding board. There are firms and investors you can go get who will be your confidant and friend and support, and whose firms are set up more aligned with your outcomes. These larger firms are not that. So people need to take a crash course in how public companies manage the asset managers that buy their stock — because that’s the way you need to treat some of these firms. That goes to things like information rights, strategy, who you let into what round, what percentage of ownership, how much of your pro rata they control, majority voting rights, and so on. You need to manage them as the asset managers that they are, and be very, very careful. It’s not bad — it’s a good weapon you can use in a big fight — but you have to be really careful.

Eric Newcomer: Kyle Harrison, Contrary — thanks so much for joining us.

Kyle Harrison: Thanks for having me.

Connections

The chart

• The “foundation model polyamory” screenshot — Kyle’s viral map of which venture firms have backed which competing frontier-model companies. Triggered by an xAI fundraise announcement and an LP account’s question, “is backing competitive companies allowed now in venture?”

The firms

• a16z — the most active/promiscuous; Martin Casado’s “why not all of them?” coverage play across every AI segment.
• Sequoia — active but more protective of its reputation at the earliest stages; the Finix-vs-Stripe “just a donation” episode as evidence they’re thoughtful about competitive checks.
• Lightspeed Venture Partners — deliberate catch-up; Anthropic one of its biggest positions, plus Mistral and xAI.
• Founders Fund · Thrive Capital · Khosla Ventures — the serial monogamists; concentrated books, all-in on OpenAI, “back up the truck” into the category definer.
• General Catalyst · NEA — conspicuously absent from the chart.
• Coatue · Index Ventures — Kyle’s former firms (per the intro).

Model companies

• OpenAI — becoming a consumer business; ChatGPT the cash cow; the reported “do-not-invest” list (with Glean); the $3B Windsurf (Codeium) acquisition as a move down into the application layer.
• Anthropic — planting itself in the API / B2B world; its own 2027 forecast of ~5x OpenAI’s projected API revenue; the bear case that API revenue isn’t defensible vs. open source.
• xAI · SSI (Ilya Sutskever) · Thinking Machines (Mira Murati) — the other frontier-model bets on the chart; SSI as the likely AI-safety differentiator.
• Cursor · Windsurf · Bolt · Replit — codegen agents that can swap Claude for Gemini tomorrow, illustrating why the model layer is hard to defend.

People

• Eric Newcomer — host, The Newcomer Podcast.
• Martin Casado — a16z; “why not all of them?” at Cerebral Valley.
• Trae Stephens — Founders Fund; “there’s one company” (SpaceX, Anduril) philosophy.
• Peter Thiel — the monopoly / “choke out the oxygen” frame.
• Sam Altman — OpenAI; the reported investor “do-not-invest” list.
• Elon Musk — the xAI raise that triggered the chart.

Concepts & themes

• Venture capital as asset management — Kyle’s three-year Substack thesis; the firm business model has shifted from picking to gathering assets.
• Capital conglomerators / the Blackstones of venture — the polyamory as a symptom of the rise of scaled, multi-strategy firms.
• Later-stage exposure vs. early-stage selection — “you’re largely just garnering exposure to assets” once the rounds get to $10B+.
• Hype cycles vs. value cycles — capital and hype not commensurate with articulated business value; Nvidia, OpenAI, and Accenture the real earners; Klarna/Shopify case studies “aspirational.”
• Network effects / distribution moats — Threads-style install-base advantages; ChatGPT as “a distribution breakthrough, not a technological one.”
• The dot-com three-layer analogy — infrastructure / networking / applications, and why the model companies collapsing networking-and-application into one changes the game.