More or Less Podcast

Chinese AI Model GLM 5.2 Beating Frontier Models | Meta Glasses, Polymarket Scandal, AI Talent War

2204 summary words 10 min summary Watch video

Start with the signal

10 min read

Summary

At-a-Glance

  • Verdict: Skim
  • Core thesis: The highest-signal argument is that fast open-weight models such as GLM 5.2 may already match frontier models for coding while running much faster on rented infrastructure, weakening frontier-lab pricing power and making model routing and self-hosting increasingly practical.
  • Why it matters: The firsthand GLM test, discussion of runaway agent-loop costs, and examples of agents grading meetings are directly relevant to OpenClaw's model economics and operating design, but they occupy only a small portion of a long, repetitive roundtable.
  • Best use: Use the AI-model and agent-operations discussion as a prompt for an internal GLM benchmark and cost-control review; rely on this brief for the consumer-tech, prediction-market, and capital-markets segments.

Executive Summary

The most consequential segment is Dave Morin's account of running the open-weight GLM 5.2 model across eight rented Nvidia B200 GPUs. He reports roughly 150 tokens per second—about 10 times his experience with frontier services—and says it was as good as or better than frontier models on the same coding tasks. For an engineering team with heavy aggregate token consumption, he argues that rented or owned infrastructure could provide much faster, effectively uncapped generation without frontier API pricing.

The panel links this to uncertainty around frontier AI valuations and the talent war. Their argument is that investors cannot model these companies conventionally, so they reduce the outcome to an 'infinity or zero' scenario and overreact to weak proxies such as researcher movement. If open models continue closing the capability gap, the enormous capital and valuation positions in frontier labs become harder to justify through durable model scarcity alone.

The practical agent lesson is that costs are driven less by ordinary one-off prompts than by recurring jobs and long-running goal loops. One participant discovered that some unattended recurring agents were spending far more than others and had to reduce their usage. The panel also describes agents grading founder and LP meetings, reviewing content, and recommending more concrete agendas and action-oriented conclusions.

Elsewhere, the group sees Meta's more fashionable smart glasses as meaningful product evolution but not yet an everyday habit: video, audio, calls, sports, travel capture, and visual Q&A work, while image quality and routine usage still lag phones. They regard real-money prediction markets as potentially valuable information systems but addictive and reputationally vulnerable, particularly after an allegation that Polymarket encouraged influencers to portray fake trades and earnings. The remaining IPO, media-app, and Cannes discussion is thoughtful but mostly tangential to Ken's core work.

Key Takeaways

  • Claim: An open-weight model may now be good enough to replace frontier APIs for at least some production coding workloads. | Evidence: Dave Morin says he tested GLM 5.2 on eight rented Nvidia B200 GPUs and obtained about 150 tokens per second, which he characterized as roughly 10 times frontier-model throughput. On the same coding tasks he normally sends to frontier models, he judged GLM 5.2 to be as good or better. | Implication: OpenClaw should treat model choice as an empirical routing and unit-economics decision rather than assume frontier APIs are always required; coding workloads are an immediate candidate for an open-model evaluation. | Caveat: This is one participant's recent, anecdotal test rather than a controlled benchmark; no task suite, accuracy measurements, total infrastructure cost, utilization level, or operational overhead was provided. Eight B200s are also expensive high-end hardware, so faster inference does not automatically mean lower total cost.
  • Claim: The economic case for self-hosting emerges at the team or fleet level, not necessarily for an individual power user. | Evidence: Morin explains that the B200 experiment was motivated by an engineering team's aggregate monthly token expenditure, framing the choice as paying frontier prices repeatedly versus renting or purchasing a rig that can generate far more tokens at higher speed. Another participant said his own heavy individual use still costs only a few thousand dollars per month and that a large market-analysis and back-testing task cost about $100. | Implication: Ken should evaluate dedicated inference only against measured organization-wide demand, concurrency, latency, and utilization; isolated anecdotes about large token bills are insufficient. | Caveat: The conversation supplies no break-even calculation and does not account for deployment engineering, uptime, batching, idle capacity, security, model updates, or fallback requirements.
  • Claim: Autonomous loops and recurring jobs are the main hidden source of agent spending risk. | Evidence: Participants distinguish ordinary prompting from long-running goal loops and cron-like recurring work. One says she had multiple jobs running continuously, did not realize some were spending much more than others, and had to 'bash' the costs down. | Implication: OpenClaw's control plane should expose per-agent and per-loop cost attribution, hard budgets, time and iteration limits, anomaly alerts, and kill controls before unattended workflows scale. | Caveat: No specific orchestration platform, token budget, or remediation mechanism is described.
  • Claim: AI talent moves are being overinterpreted because frontier-lab valuations lack conventional operating anchors. | Evidence: The panel notes employees moving among Meta, OpenAI, Thinking Machines, Anthropic, and Google, sometimes reversing previous moves. It characterizes investor models as 'infinity or zero' rather than forecasts based on revenue growth or a five-year CAGR, causing every personnel move to become a tradable proxy for the probability of an extreme outcome. | Implication: For investment work, researcher movement should be weighted behind product performance, distribution, compute economics, customer retention, and evidence of durable differentiation—especially as open models improve. | Caveat: Talent concentration can still matter when a specific researcher controls unique capabilities, teams, or execution knowledge; the panel discusses the signal generically and does not analyze individual departures.
  • Claim: Meta smart glasses have credible situational utility but have not crossed into a twice-daily consumer habit. | Evidence: Users praised hands-free travel video, calls, audio, sports capture, and the ability to ask the glasses what they are looking at; one participant used that visual-Q&A feature around Rome. However, an owner of roughly 10 pairs acknowledged low hours of use, and another said photos remained below smartphone quality even though video felt more like a breakthrough. | Implication: The near-term opportunity is likely contextual, hands-free capture and multimodal assistance rather than wholesale phone replacement; any agent integration should be designed around specific high-frequency situations. | Caveat: The panel is unusually enthusiastic and owns multiple generations, so its experience may not represent mainstream retention. It also flags surveillance and training-data concerns without detailing Meta's policies.
  • Claim: Prediction markets need real stakes to produce engagement, but addictive design and irresponsible promotion can destroy their information-market credibility. | Evidence: The panel argues that a Meta market using only virtual currency would not matter, while real-money markets can aggregate dispersed knowledge. It also cites parents reporting serious sports-betting and prediction-market problems among people in their 20s, and alleges that Polymarket directed influencers to make fake videos about fake trades and report fake earnings. | Implication: Any investment or partnership thesis in the category must include promotion disclosure, market integrity, responsible-gaming controls, and regulatory exposure as core product requirements rather than peripheral compliance work. | Caveat: The Polymarket conduct is an allegation made in the discussion and is not independently documented in the transcript. The speakers also blur prediction markets, sports betting, and gambling in places, despite differences among those products.
  • Claim: Earlier public listings could democratize access to venture-scale gains, but only by accepting weaker protections and more fraud risk. | Evidence: The panel says Dell raised $30 million at an $85 million valuation 38 years earlier, equivalent to an approximately $239 million market capitalization in current dollars. Morin contrasts that with a stated Anthropic valuation of $965 billion and argues that late listings reserve most appreciation for private investors. The counterargument is that lighter listing standards undermine the same consumer protections used to distinguish investing from gambling. | Implication: For investment-thesis work, market-access reform should be evaluated together with disclosure quality, liquidity, passive-index effects, and fraud tolerance rather than framed as a simple public-versus-private choice. | Caveat: The Anthropic calculation is a hypothetical based on the valuation quoted in the conversation, not a verified return scenario. The panel also notes that few retail holders would actually retain a stock through decades of appreciation and drawdowns.

Detailed Brief

Agents as continuous management and product-review systems

  • Claims: One participant has agents grade external meetings with founders and limited partners, and also uses them to review new content and recommend edits.; Another uses Claude against product screenshots to compare design alternatives and asks an agent to diagnose meetings that felt unsuccessful.; The recurring feedback from the meeting-review agent was not abstract stylistic advice: it called for a more concrete opening agenda, clearer discussion, and explicit actions at the end.
  • Evidence: The workflow is described as part of a recurring loop rather than an occasional manual prompt.; The panel proposes extending the same grading pattern to podcast episodes and app experiences.
  • Caveats: No grading rubric, source-of-truth outcome data, privacy process, or evidence that the agent's scores correlate with actual meeting success is supplied.
  • Implications: Agentic evaluation becomes more useful when it converts qualitative work into a repeatable feedback loop, but the evaluation criteria need to be tied to real outcomes to avoid generic advice.

Distribution design and the public positioning of AI

  • Claims: The Information made short video clips the default in its mobile video experience because users preferred concise clips to full episodes.; The panel favors a personalized ranking layer and debates replacing article feeds with a one-story-per-screen, card-like interface that gives users a finite completion count.; At Cannes Lions, Google and Meta were interpreted as repositioning AI defensively: Google emphasized that AI can preserve creativity, while Meta's 'the future is for everybody' message was read as an attempt to broaden acceptance and preempt regulation.
  • Evidence: The redesigned app places a technology-focused deep-research chatbot prominently so users can ask what they missed or retrieve the latest reporting before meetings.; Suggested product changes included ranking shorts by interest, using a full-screen stories format, and inserting advertising or owned promotions at a regular cadence.
  • Caveats: The app discussion is an informal live critique without retention, conversion, or engagement data.; The interpretation of Cannes messaging is the panel's inference, not a statement of the companies' explicit regulatory strategy.
  • Implications: Specialized data and workflow context can differentiate an AI interface even when its underlying interaction pattern resembles general chatbots.; AI companies increasingly need a social-license narrative alongside technical capability, particularly in creative and consumer markets.

Notable Concepts & Terms

  • GLM 5.2: The Chinese open-weight model that Morin says matched or beat frontier models on his coding tasks while delivering about 150 tokens per second on eight B200 GPUs.
  • Nvidia B200: A high-end Blackwell GPU used in the self-hosted GLM test; its performance is central to the claim, while its cost and operational requirements are central to the missing break-even analysis.
  • Long-running goal loops: Autonomous agent executions that continue iterating toward a goal and can consume substantially more tokens than ordinary requests.
  • Infinity-or-zero model: The panel's shorthand for valuing frontier AI labs primarily by the probability of a dominant or failed outcome rather than conventional revenue and growth forecasts.
  • Toothbrush test: A consumer-product heuristic asking whether a product is useful at least twice per day; the glasses have compelling use cases but, for these owners, have not consistently met that threshold.
  • Markets as information machines: The Hayek-inspired argument that prices aggregate dispersed knowledge, providing the intellectual case for prediction markets beyond entertainment or gambling.
  • Forced illiquidity: The observation that private-company employees can become extraordinarily wealthy because they cannot sell during rapid appreciation, whereas public-market holders would often diversify after an early gain.
  • AI social license: The implicit need for technology companies to persuade creators, consumers, and regulators that AI is inclusive and creativity-enhancing rather than destructive.

Operator Notes / Why Ken Should Care

  • Run a blinded GLM 5.2 evaluation against the frontier models currently used by OpenClaw, using representative coding, tool-use, long-context, and agent-loop tasks rather than generic benchmarks.
  • Build a full inference break-even model covering GPU rental, utilization, batching, engineering labor, observability, redundancy, model refreshes, and frontier fallback—not merely cost per token.
  • Add per-workflow budgets, maximum loop iterations, wall-clock limits, recurring-job inventories, anomaly detection, and emergency termination to the agent control plane.
  • Review whether any recurring OpenClaw jobs continue running after their output has ceased changing or after their downstream consumer has stopped using the result.
  • Create an outcome-linked rubric before deploying agents to grade meetings or content; compare agent scores with follow-up rates, decisions reached, and completed actions.
  • Monitor open-weight coding models as a strategic pressure on frontier-provider margins and use benchmark results as leverage in API pricing and capacity negotiations.
  • Treat claims based on newly released model tests and quoted private-company valuations as diligence leads, not investment facts, until independently reproduced or verified.

Source/Metadata

  • Title: Chinese AI Model GLM 5.2 Beating Frontier Models | Meta Glasses, Polymarket Scandal, AI Talent War
  • Transcript words: 16788
  • Duration seconds: 2974
  • Timestamp note: No timestamps or chapter markers were present. The supplied transcript also contains substantial duplicated passages and repeated closing material.
Search this video Resume transcript at 0%
Full transcript 9755 words · 74 min read
0:00

This happened at Facebook. There's so many people on Facebook, right, in the early days, who were very, very junior engineers and got hilariously uber-wealthy in the first three years of their career purely because they literally couldn't sell and it appreciated so quickly. That would have never happened in the public market. Zuck is a guy that can carry a grudge, and I think Zuck is still savage about when he wanted to buy Snap and it didn't work out, so they copied everything, and he's just putting the nail in the coffin. To me, trading cards are infinitely dumber than prediction markets. I would rather our son be addicted to prediction markets

0:33

than to stupid unboxing trading cards. Here's where Dave used them. We went to the Vatican. We're showing our kids the Vatican, obviously very historical place. They're like, no photos, Sistine Chapel. Who puts on the meta glasses right away? Dave and Britt plus Sam and Jess put it all right to the test. More or Less. Why, hello, friends. Welcome to More or Less.

1:12

And Lauren's, we're so happy to have you back. Welcome to the quad. Glad to be back. I did listen last week. Good to see you guys. I thought you guys had a good conversation. Yeah. We got a lot of people, a lot of comments, a lot of listens to just the lesses. Guys, guess what happened? We were in the Zurich airport in an underground train. Lumi, our three-year-old, needed a place to sit down because her legs were tired. Some woman moved over. Then that woman's husband looked at us and goes, wait, I know you guys. I watch your podcast. Guys, look how famous we are. Isn't that exciting? The other day, I got recognized on an airplane, and it is really funny

1:48

because it is like, talk about being niche famous. It's like, it turns out no one knows who we are except for the six routes we travel. Hey, Zurich was weird. I would say Zurich was an outlier. My grandfather is proud of me for the Zurich one. The only time that I have a minor fan posse is at a Silicon Valley developer conference. I was very popular at NVIDIA GTC. I couldn't even buy myself a coffee, but that was it. So that's okay. We'll take it where we can get it. This is the future. It's all about having your community, which is very small, but with your relevance. It's pretty amazing. This guy who we met on the Zurich airport,

2:27

we ended up getting some really cool stuff done for Open Claw within 48 hours. And so I'm very grateful for it. He gave Dave a hot tip, and then they vibe-coded that up real fast. Blue horseshoe loves Anacott Steele type stuff. What are we doing here? Can't tell you the secrets. I'll tell you offline, Sam. Whatever is smaller than a Pico influencer, that's what we are. What's smaller than a Pico influencer? Nano? Nano. I thought it goes Nano Pico. Micro. Quantum? See, I'm not even smart enough to figure that. Guys, it's been another exciting week in technology. Here are some of the things that have happened. Has it? Sam says no. I try to bring the hype,

3:02

but you guys can tear it down. It's good. We had Cannes Lions. That is not a thing. Not to be confused with Festival de Cannes, right? But actually, I have a profound tech takeaway from that, that I'm willing to share with people. We have 20 million new styles of Meta glasses backed by Kendall Jenner. They're cool. I like them. I actually like the look too. Guys, you guys can't talk about the news. We're not talking about the news. We've always been Meta glasses fanboys. Well, this is the agenda section. We've got Meta glasses. We have Cannes. We have hot AI recruiting poaching, Summer Part Two. You may remember last summer when Meta raided OpenAI.

3:36

Well, the sequel is back and Anthropic is raiding Google. We should talk about that or talk about why it doesn't matter, but the markets seem to think it matters. Dave has some thoughts about IPOs. I love when one of you tweets. It gives me so much fodder. SpaceX is also tanking. And Britt, there's something happening with the app store that you wanted to talk about in Europe. Yeah, that sounds good. And Meta is launching like a Polymarket competitor. We should talk about that too. But for virtual currency, no real money? I know. Sam, I wanted to ask you your opinion because I want to do this for the information, but I get stuck on the virtual currency part

4:11

because I don't think anyone would care. No. Well, we've talked about this six times and it has to be real money. It's not a thing if it's not real money. I know. Well, also guys, I love the headline for that article broken by the New York Times was "CEO directs company to look at." I was like, I just thought that was very funny. Isn't that what CEOs do? In an era of superheroing shares, everything is "CEO directs company to do thing." It's more like "CEO directs company to do thing, company doesn't do thing." Well, it's like, the headline should be "In shocking twisted events, a non-CEO directed a company to do something." That doesn't happen at that company

4:37

or any other company. Let's start with the glasses. So we had Snap come out with its latest Spectacle thing. Meta basically went independent, dropped the Ray-Ban relationship. That's not true. What did it do to Ray-Ban stock? That's not true. It's all Luxottica, right? Luxottica owns Ray-Ban. They just made their own label, right? So it's all the same. It's just that they've got their own label now, but they're manufacturing it in the exact same factories. Fine. They've got their own label. Every glass is manufactured in the same factory. For some reason, there's only two things you can't manufacture in one factory. One is the most high-end ships in the world

5:16

and the other is sunglasses. Interesting. Well, I like them. I'm going to buy them. So we like these new styles. Do we think it will matter? Britt, is this going to become a new category? Is this the breakout moment for wearable glasses? Isn't it already? The thing I think is hilarious is that last week, Evan from Snap launched... What are those glasses called? Spectacles. The $2,000 ridiculous looking... Miranda Kerr is attractive. She's attractive in anything, though. She could wear a paper bag on her head and she'd be attractive. Yeah, that's true. Guys, Kylie Jenner is... I mean, Kendall Jenner is also attractive. Is it Kendall? It's Kendall. It's true.

5:54

But the Snap glasses... No, this is not going to happen. Hilariously, I would bet you Zuck pushed the timeline up to get this announcement live within a week. Zuck doesn't care. No, he's been making this push for a while. Guys, Snap is a $7 billion company. It's completely irrelevant at what Snap does. It's like a pre-Series A AI startup. Zuck is a guy that can carry a grudge and I think he is still savage about when he wanted to buy Snap and it didn't work out so they copied everything and he's just putting the nail in the coffin. No, this is some conspiracy. I love conspiracies, Dave. Trace the launch of these glasses to Mark going to the Prada show.

6:28

This has been a long, simmering fashion thing. I'm not saying they weren't in the pipeline. I just think he might have buffered the marketing announcement a little bit. No. They did that with Threads maybe, but with this... Keep in mind the fact that Snap is now about two Box.nets. That's how important and big a company it is, right? It's about two Box.nets. We love Aaron. We don't understand that company. We love Aaron. I hung out with Aaron the whole Apple event. To be clear, I adore Aaron. I don't understand the company. We all love Aaron. Aaron, we need you back on the pod. There's a difference between loving the person and the company. So it's a two Box.nets company.

7:15

You guys, stay focused on the glasses. Two... Well, I wait for Box.nets glasses. But the second... You can talk to your files in your glasses. It's good news. You can see files in your glasses. All of the files. And second, in terms of these... The marketing announcement a little bit. No. No, they did that with Threads maybe, but with this... Keep in mind the fact that Snap is now about two box.nets. That's how important and big a company it is, right? It's about two box.net. We love Aaron. We don't understand that company. We love Aaron. I hung out with Aaron the whole Apple event.

7:36

To be clear, I adore Aaron. I don't understand the company. We all love Aaron. Aaron, we need you back on the pod. There's a difference between loving the person and the company. So it's a two box.net company. You guys, stay focused on the glasses. Come on. Two... Well, I wait for box.net glasses. But the second... You can talk to your files in your glasses. It's good news. You can see files in your glasses. All of the files. And second, in terms of these... I do have to give credit. I don't know Evan at all. I think I've met him once in passing. I actually like Evan.

7:41

The only thing I know about Evan, what I do appreciate and I think is hilarious, is that on LinkedIn, his profile says something to the effect of head of products for Meta, which is awesome. See? I told you. There's still a grudge, you guys. So this... Did we decide if this was the watershed breakout moment for this category of wearable or is that the jury's out? No, Jess. It's just the evolution. Okay, Dave. It's not the evolution. You know why? Because from day one... Britt, you're... When... Stop interrupting. From day one... Dave, stop the head shaking and let the woman speak. This is like Austin Powers.

7:48

Yeah. Our editors are... Or else the editors are going to cut this out. Okay. I said from day one when Meta launched the first Rave. Women are not going to wear these things. They... Some women wore them fine. They are not attractive for women. They never have been. The Oakley ones sucked too. Finally, this is the watershed moment of women buying smart glasses. Mark my words. Saying it now. That's all. So yes, I do think it was a big deal. Okay. Britt says it's a big deal. I think it's a slow evolution. I think it's a slow evolution.

7:55

Look, I love the smart glasses. I have probably ten pairs of them. I think they're great. I've had them for every generation. I used to have the original ones. How many times a year do you wear them, Sam? How many times a year? That's the thing. I love them and they're really cool for action shots. I do use them occasionally, but my hours of use per pair is quite low, right? Still, right? Is my problem.

7:58

I love the idea of them. They have great audio. I love popping them in Miata. They're great on phone calls. They're a very specific use case. They're fun for photos, but they have not yet made it, despite my ownership of so many pairs of them, into my daily rotation in any form. Yeah, that's a good point. I use them on... Dave wore them all last week on our travel. No, I didn't use them all last week. I brought them along and I had them in my... You know, I love photography.

8:02

Here's where Dave used them. We went to the Vatican. We're showing our kids the Vatican. Obviously, very historical place. They're like, no photos, Sistine Chapel. Who puts on the MetaGlasses right away? I tried to do that at the Toronto Space Needle where they're like, you can't have anything except for your glasses. So I put on my glasses. I'm like, great. I'm going to get my own footage. I don't want to pay you forty dollars for your pre-recorded roll of me on the Space Needle. They're like, no smart glasses. I'm like, what if these are my prescription glasses? You can't tell me not to wear smart glasses. So I had to fight with Canadians about the same thing.

8:06

Well, it's interesting. I've used them throughout the trip here and there. And even if you're doing that and if you have your smartphone and then I have I carry another camera with me usually. The quality isn't there yet to exceed the smartphone quality. Pretty good. It is pretty good. This sounds great. The quality is still right below smartphone quality. And so when I look at my photo library in context, I'm less psyched about the photos from the MetaGlasses than I am about. But Dave, what you have to do is include at least one photo from the glasses in your Instagram carousel so you get boosted distribution. Oh, yes, indeed.

8:12

Look, the videos are definitely breakthrough. In terms of video, it's certainly a breakthrough. So my big news of the week, which is relevant to this, is I unfortunately dropped my phone off a mountain and I lost the greatest video I've ever taken in my entire life. Oh, man. And it's funny because you're like, I should have been wearing my glasses, which case I wouldn't have dropped my camera, I mean, my phone off the side of the mountain. And actually, it would have been amazing for climbing the middle of Teton. But for whatever reason, despite owning several pairs and having them in my house, I didn't bring them.

8:17

So there's some gap still to cross there, if that makes sense. You know, for sure. I was in the middle of the Roman Forum. As one is. I had them on. You guys are just pissing in each other's where I was last week. Sorry. It's just I was at the Forum. I was on the middle Teton. I was, you know, okay. Let us do our man thing. Just keep going. Tell us when you're done. Yeah. Listen, we're early stage VCs for a reason, Jess, which is we get to do these things. We don't have real jobs. You know, to the same point, Sam, I realized, oh yeah, I can ask it what I'm looking at and have it tell me the story. And it's going to be like, the Sistine Chapel, you idiot.

8:28

I will say. It's a pretty good use case, though. You can say like, hey, Meta, tell me what I'm looking at. He was doing that a lot in Italy. I learned a bunch of things about Rome that I didn't know. And that use case was pretty great. Some of which may even be true. Wait, does Meta's click tracking apply to the glasses? Are they like. Of course they are. What are you talking about? Doing getting training data off of everything. And by the way, is this what's going to start happening? Because I've started getting startup pitches about this where everyone's wearing voice recorders now. So the thesis is everyone's hustlers. Everyone meaning you?

8:32

Well, no, some people in Silicon Valley. So the thesis is they need more training data. Real world people are construction workers and Disney ride staff going to start recording everything. Sam knows all about this. I know of a deal he's working on in this zone. It's done. Oh, now we're just ruining each other's deals. I love this. Okay. So we've decided the hardware is making progress. The software and use cases. Don't you have to be at the toothbrush test to be a consumer product? You have to be used twice a day? You use your toothbrush twice a day? That's a very old analogy. I forgot about that. Yeah. It's like coffee and toothbrushes.

8:42

I just teed that up for you. I don't know. I'm optimistic. I also really like the sporting use cases of these things. I think they're fun to play tennis with and all that kind of stuff. So we're pro. We're pro. It's great. It's great. It looks great ski racing. We're pro. It's just there's something still to cross. How about this? Credit where credit's due. It is a great product and I like it when Meta gets out of its own way and doubles down on products that it has that are already great. And so I kind of give them props for doing that on this. It's a great product. Just do it. You know? Okay.

8:46

Dave, how would you feel about a Meta prediction market? Because that will seg us into— Doesn't matter. Don't care. Yeah. Doesn't matter if it's not real money. I think you have all these companies unleashing vibe coding within their companies, which means you're going to have a lot of side projects and interesting things. This predates vibe coding. Okay. Well, I don't know. I mean, it's fine. I'm actually not that interested in it either, but I think you're going to— There'll be a lot of random apps from a lot of random companies. Credit where credit's due. It is a great product and I like it when Meta gets out of its own way and doubles down on products that it has

9:01

that are already great. And so I give them props for doing that on this. It's a great product. Just do it. Okay. Dave, how would you feel about a Meta prediction market? Because that will segue us into— Doesn't matter. Don't care. Doesn't matter if it's not real money. I think you have all these companies unleashing vibe coding within their companies, which means you're going to have a lot of side projects and interesting things. This predates vibe coding. Okay. Well, I'm not that interested in it either, but I think you're going to— There'll be a lot of random apps from a lot of random companies. I'm just not interested in the category in general. It's doing

9:44

quite a bit of damage to kids in their 20s. They're very addicted. They're very addicted to sports gambling. Okay. Everything's bad because of kids. But Dave, are you talking about just prediction markets or are you talking about— Prediction markets. Prediction and sports gambling because— They're losing a lot of money. We're not calling it gambling. Sorry. They're bigger sports betting platforms than the prediction markets from my understanding. Yeah. Not like Kalshi. Listen, guys, I'll take the other side of this for once. Look, I actually absolutely love this category and think it does make sense in the very abstract for Meta to be doing it. Why?

10:29

Because my favorite person to read in college was Hayek. Markets are information machines, right? And especially if the world gets harder to predict. Yeah, yeah, yeah. We can talk Austrian economics all day. I love this stuff and I think it makes total sense abstractly. If there were companies I would want to go run, this is very high on the list of things I think matter and are interesting. That's a big one. No, it's super important to get right. This is one of the key machines that we need for the internet to be the internet. We need pricing mechanisms and machines that tell us what's real. So I'm fully on board with this being awesome and more distribution is good.

11:17

The problem is they don't always tell us what's real. Sometimes they tell us what people think you want to be real. Well, no, that's the same thing as the stock market. The stock market reflects not reality. The stock market reflects what number goes up and number goes down and there's a second order game to the whole thing. But there are a lot of problems and these fundamentals— If that's already the truth of reality right now, Sam, then why wouldn't it just apply to this? You've launched another one of these markets, you end up with number goes up, number goes down, crypto behavior all over again. No, no, no. I get it. The reason it is intellectually interesting

11:57

for Meta— Of course. Is because you have a bunch of normies who actually do have information about the world from their personal lived experiences and if you can get them engaged in a market on that, you can learn about the real world and that's actually using the humans to pump all the knowledge, which is one of the most interesting things possible about a social network. Now, how it actually plays out on the less intellectual side is all— It's just never happened that way. We can go back through the history of social networking launches, not just Meta but across all of them, and it just never plays out that way. I agree that it's extremely unlikely

12:47

that an enormous company platform will successfully launch something as avant-garde as a prediction market. It's not impossible but very unlikely. I don't know. It might be good but I actually— I worry that it will be too good. Say more, Dave. When you talk to parents, I've got a lot of friends around town that have kids in their 20s and they are having serious problems with sports prediction markets and they can't stop. They are spending way too much money. These things are really good at addicting you to getting money out of you while watching a sports game and I actually worry that Meta will be insanely good at this and that's the problem. Would you rather your kids

13:25

be addicted to this or Pokémon? Pokémon, probably. Our oldest son has discovered trading. What do you mean by Pokémon? The trading cards. Our oldest son has discovered trading cards and it's the most annoying thing in the world to me because it's so stupid. You have to go to a physical store. They grow out of it. This is the year that happens. But you don't actually, Jess. There are entire apps now that are about trading cards. I know. To me, trading cards are infinitely dumber than prediction markets. I would rather our son be addicted to prediction markets than to stupid unboxing trading cards because at least there's some risk, reward, learning markets.

14:01

It's interesting how hard it is to process that. What if he's addicted to playing with his friends outside and doing his homework and playing tennis? An old dream. I'm just saying in the spectrum of stuff for teens to do, I would rather them be learning about bet sizing, risk reward, stochastic markets than watching idiots unbox Pokémon cards for the dopamine hit of getting some stupid rare cards. The unboxing videos, I hate. I hate that. Not just the unboxing, but I think they call it ripping. You rip open the trading card pack and there's an entire market around this. I'm not into it, but all the guys I know that coach baseball teams around the Bay Area,

14:41

they're super into this and it's become a major category. Okay, Britt, do you have prediction market thoughts and then I have one. The one thing I wanted to say back to the voice of the female user and consumer: women are only less than a quarter of Kalshi and Polymarket right now. It really hasn't struck with most of the women I know. I actually think it's going to hit with Meta if they do this and they build it into things like Instagram because I think— These things are already at large scale. Why would the behavior change? It has to be betting on Love Island. Exactly. Like The Bachelor, and people are going to be, "Oh, I bet it's going to be Jake, not Chris."

15:12

And the women are going to get really vicious with each other about which guy's going to be the winner or what Taylor Swift's going to wear to her wedding. And I think it's going to become tribal and really community oriented and women are going to get really into it. I was looking at the numbers on X and X is actually 85% men. I could have told you that. Instagram's the opposite. It's all women. And so really what Meta has to do to make this successful is make it successful on Instagram and then it will crush. That's what's going to happen. I like that angle. So I like Britt's angle. I have another angle. Intellectually, Polymarket and Kalshi

15:45

are really interesting businesses for all the reasons Sam talked about and markets and truth. And as a journalist, there's a complicated relationship with them, but there is a wisdom of crowds, undeniable, truth, democratizing information element that is very compelling. The problem is these companies are not acting particularly responsibly, especially Polymarket, which I think is actually a shame because it's allowing, especially the media narrative around them, to drop like low hanging fruit for the media to rightfully criticize and attack them, which is actually obscuring some bigger issues. So the latest one is Polymarket's influencer program. You can agree

16:17

or disagree that influencer programs are good or bad. But they should be labeled. Polymarket has been directing influencers to make fake videos about fake trades and report fake earnings for their influencer program. I think that's immature behavior. Honestly, I don't know. I'm not saying it's illegal. I'm just saying if you're in a business like this that is so potentially interesting yet disruptive and you're going to go—

16:35

Truth, democratizing information element that is very compelling. The problem is these companies are not acting particularly responsibly, especially Polymarket, which I think is actually a shame because it's allowing, especially the media narrative around them, they're dropping low hanging fruit for the media to rightfully criticize and attack them, which is actually obscuring some of these bigger issues.

16:35

So the latest one is Polymarket's influencer program, which you can agree or disagree that influencer programs are good or bad. You should say they should be labeled. But Polymarket has been directing influencers to make fake videos about fake trades and reporting fake earnings for their influencer program. I think that's just immature behavior. Honestly, I don't know. I'm not saying it's illegal. I'm just saying, but if you're in a business like this that is so potentially interesting yet disruptive and you're going to rattle some cages, okay, but rattle the right cages. Just don't do stuff like that that honestly reporters, some of whom aren't going to take the time to understand the intricacies of the platform, which is actually quite interesting. There are many things about the platform that are transparent and you can see all the transactions, which is unlike sports betting. All of those things just are going to get drowned out by paying influencers to make fake videos, reporting fake earnings, and then denying it and being nasty about it.

16:36

So that is my current take on the prediction markets. They're not going away. And I think it's going to be hard for anyone to compete with them if they're based off a fake currency. For sure. No one cares about a fake currency. Yeah, no question. Fake currency is dumb. Awesome. Okay, we got that topic. Let's talk. Will you guys indulge me on the AI talent wars? I feel no one's interested in this, but the markets are. It's boring. It's ping pong. You know why I think it's not boring? People want money. AGI is not happening at any of these places. It's all a big thing. But that's what's interesting about it. People are willing to do anything to make money.

16:40

I think if you look at the layer of who goes where, it's profoundly uninteresting at this moment. Some of the people who went to Meta from Open AI have gone back. People who were at Thinking Machines who went back to Open AI are going back to Thinking Machines. What's interesting to me is what investors and journalists are putting into it. And it makes me realize everyone is deeply insecure about the future of all of these companies, that they're hanging so much on these signals that are objectively determined by so many other factors. So to me, it's actually a comment on how no one knows how to model out the future of all of this stuff.

16:42

So you're saying that's why they're switching? No. This is why everyone's hanging so much on this. Why the entire Google stock is sliding X percent on the news. Investors don't know. They're looking for any signal or any alpha, but they're picking the wrong one. Well, that's the problem when your model is infinity or zero. Interesting. I think you're right, but let me try to build on that. I think here's the simple thing. Everyone's model looks like this: infinity or zero probability of infinity, right?

16:46

Yeah, exactly. There's no more modeling than that. And everything is this risk adjusted. Is this going to be infinity or not? And so because there's no actual year over year growth, there's no model, right? Then you take any incremental signal and especially because everyone else is going to trade on it. So you have to trade on it faster and it just cycles into being a much bigger thing than it should be, right? Because the models are so, we've all made these models when we were 22, right? When working for other people is like, you know what it's supposed to look like and then you know what it looks like now and that's why it's happening because there is no, oh, this was growing at 32% and now it's going to grow at 38% compounded five-year CAGR. It's just infinity, zero probability of infinity.

16:47

I would add one more thing. The level of commitment people have made to investing into these things is extraordinary. The amount of risk. The stakes are high. The amount of risk that is actually in all of these positions is really big and so I think you've also got that anxiety out there in the market because people don't know whether or not, the example I'll give is yesterday I've been testing eight Nvidia B200s, basically stocked eight of these things running GLM 5.2 and it's really amazing. This is an open source model. Dave, what is the context of the B? Give us the... The B? Where did you get eight B200s? It's a long story. It's a friend of mine.

16:51

That's the thing I'm interested in is not what you're doing with them but where did you source eight of them? I can tell you later. Dark web, Sam. Those are expensive. Are you just personally paying for those? We're renting them right now but they're really expensive. That's a lot. That's expensive, Dave. And what is special about that B200? It's the top-end GPU that Nvidia makes, right? It's the new hotness. Oh, I see. It's the Blackwell. Yeah, they're the Blackwell 200s. It's not the Vera Rubin though. No, the Vera Rubin's... You can't get Vera Rubin's right now unless you're way out on the frontier. If it's a Porsche, then an H100 is like a Miata.

17:00

Okay. And so, Dave, what's your takeaway from this?

17:01

GLM 5.2 is a new open source open weights model that was released last week. And I was playing with it in the cloud and was really shocked by its performance. And so I started talking with a couple of hacker friends of mine actually from the 2000s. And we were like, let's see if we can get a rig together and test this. And we got this thing up and running and out of the box, it's doing 150 tokens per second, which is 10 times what you get out of frontier models. And it's really extraordinary to play with. And this is an open source model. I was testing it doing the exact same coding tasks that I'm doing with frontier models. It's as good or better.

17:02

And so that's a really interesting thing that's happening right now. I'm sitting here going, I don't need the frontier models anymore. I can run these open weight models on standard Nvidia hardware, get a bunch of people together, rent one, and it works really, really well. And this is kind of Sam's point for years that we've just been waiting for this moment to come. And I felt I experienced it yesterday. So what does that mean for all of these enormous positions that are out there in these frontier labs? I think it's kind of shaky.

17:05

Yeah, I agree with all that. But I'm also, Dave, confused as to what heavy lifting you're doing here because I do a lot of coding with these things and a lot of tasks and I'm not even close to pushing the limits of what they can do. What are you doing in your spare time that's requiring this level of compute? Because I'm a pretty heavy user. Building OpenClaw. Yeah, but what specifically? What are you finding you need that level of sophistication for?

17:10

Well, this is not about I need it individually. It's that we've got an engineering team that's doing an enormous amount of token spend to do all of the engineering across the month. And so you start to ask this question, do you want to pay frontier prices for these things? Or do you want to buy the rig yourself and then be able to generate infinity token at 10 times the speed?

17:11

I totally understand. I'm actually just asking a more specific question. I gotta say, I'm a pretty heavy user of this stuff. I build apps all the time, blah, blah, blah. I am shocked at some people's level of spend on these things, at least what they say they spend. Because I'm literally like, what are you possibly doing? What are you spending, Sam? What's your monthly bill? I don't even know, but it's probably a few thousand bucks a month, which is a lot by human standards. But it's not like when people are like, oh, I'm crushing tokens, I'm spending...

17:13

across the month. And so you start to ask this question: do you want to pay frontier prices for these things? Or do you want to buy the rig yourself and then be able to generate infinity token at 10 times the speed? I totally understand. I'm actually just asking a more specific report. I gotta say, I'm a pretty heavy user of this stuff. I build apps all the time. I am shocked at some people's level of spend on these things, at least what they say they spend. Because I'm literally like, what are you possibly doing? What are you spending, Sam? What's your monthly bill?

17:18

I don't even know, but it's probably a few thousand bucks a month, which is a lot by human standards. But it's not like, when people are like, oh, I'm crushing tokens, I'm spending, I'm like, doing what? I've had this thing do entire market analysis, back testing on crazy stock theories, and it's like, a hundred bucks. What are we doing? I think the cron jobs is where they get you. No, no, no, it's loops. It's these long running goal loops.

17:21

Okay, what's a loop? Now I get to play the dumb person part. By the way, I think a lot of people are going to not know how much money they're spending and freak out because once you start getting really into this, I have all these recurring jobs that run all the time, and I had no idea that some of them were spending way more than others, and I really bashed it down yesterday. I'm usually able to figure out how to spend the maximum amount of money on something, and I've tried, and I'm just like, I can't do it. I want to be the type of venture capital that's like, I spent $20,000 yesterday. How? How? I'm trying. I go to the thing, like, spend up to $1,000 complete this task and comes back and it's like, I spent $14.95. I'm like, that's not really...

17:23

Dear listeners, please help Sam spend more money. Just tell him how to do it. We'll all be grateful and then we don't have to talk about this. No, just kidding. Token maxing. Yeah, all I know is that you go through these eras where there's the information version of clickbait, like, the thing that, if you just put it in the headline, you're going to get a huge number of subs and it's changed over the years. It's always something different. Believe it or not, it was Kubernetes at one point. That's how wonky our audience is. Then it was like SoftBank and the Vision Fund. And now, if you have AI bills in a headline, the information community is here for it.

17:26

Now, interesting. Being responsible journalists, we do not, therefore, maybe only one article a week will be on such topics. So maybe we're dumb journalists, but it is amazing how tuned people are. Is this? The other thing that does really well is anything to do with data center loans and new structures there. But anyway, that's current window into the psyche of our subscribers. Speaking of which, guys, our new app will be out probably by the time this hits, I hope. So anyway, if... Oh, what are the new features? I'm glad you asked. Did you code it? No, we probably should have. How much did you pay in tokens?

17:33

No, you did vibe code it. You just vibe code it via someone else who paid target it. I paid someone else to vibe code it. You double got double charged: tokens and a person.

17:36

Brett, let me tell you the two things I love about this app. First of all, video. So we brought TITV to mobile where it has always belonged. You can see the awesome work of our team. There's a video tab that is a great way to just get, it's basically summarizing all the biggest news. So I love it. And then our deep research chat bot is also slightly redesigned and front and center. So I use it all the time to say what I miss, what is the biggest thing that was reported this week. I go into a meeting and say, what's the latest we've reported on this or that. And it's unlike ChatGPT or Claude or Gemini. It's really smart about tech. So those are the two things and I'd love feedback on it.

17:37

What else is happening, guys? Do you actually want feedback on it or do you just want to say that is what you say? I am so, no one believes this about me. I went talking to someone. I was like, how do we get more feedback? I love feedback because I can't use all of our products every second of the day. So I would love feedback. And Sam, you have objectively been testing it for a month and giving me no feedback. So I'm sure it's perfect. Jess. I mean, all feedback is helpful. Jess, do you have your agents giving you feedback every day? No, but I've had some great agent use this week. It's been really helpful to me. But Britt, how does that look? What does that look like?

17:47

Oh, well, so I, this is part of my recurring loop. I have my agents now. When I have an external meeting with an LP or a founder or something, I get graded on how the meeting went. But for them, the whole meeting, and I think that you should also, and then every time I have new content or something, you can also run it through and get a grade or get edits. So I was just curious if you've run your app through any sort of agent and had them pick apart the pros and cons.

17:49

It's various points in this. I got some screenshots and I would ask Claude if it had a perspective on it. So if I wasn't sure of X violent style or Y violent style, but I haven't done it in a super sophisticated way. I have had meetings that I just felt haven't gone well. And I have proactively asked my agent how to improve the meeting or something I could have done better. It always just tells me that there should have been more action. It just should have been more concrete every aspect of the meeting. The beginning of the meeting should have been more concrete with the agenda. The end should have been more concrete takeaway. So we should grade all of our episodes too.

17:51

I'm using your app now and I'm actually, you changed my number one piece of feedback I didn't give you, which is that the TITV shorts, I don't want to watch the whole episodes. The episodes are long and boring. I want the clips. I made shorts default for you. I made it. No, the shorts is great. That's the only thing. That was my only feedback. The shorts are good. You will never use Reels or TikTok again. Not quite. But it's really good. And Sam, the next piece of feedback you gave, which I've put to the team is to sort them based on a recommendation algorithm based on your interests. Yeah, I don't want random stuff. But it's not that random.

18:00

You know what you should totally do? Why don't you put some ads in this, in this shorts thing? Make a lot of money. Because we don't have to because we already make so much money off TITV. You can always make more. Why don't you just like every fifth should be an ad. It can be an ad for more or less podcasts. Okay, you hear that, Lindsay? We got to put some ads. We got to put some ads in. Lindsay, fire it up. Every fifth scroll can be like, you know what everyone wants? Fifth scroll. Gemini. Woo! Throw in a more or less ad every fifth scroll. My favorite part of Instagram is the ads. So I like this, but I need ads. I need to buy something. You need to buy something.

18:06

This is really good. We're, you know, we have designed a fully designed merch store that I just haven't gotten around to launching. But maybe we should be putting our merch in this. Is this like a full app review now? Okay. With Jess. I also, for what it's like, I like the full flip vertical script thing. I think you should change the homepage so the articles are also full screen flippers as opposed to scrolling. Because I don't want to scroll anymore. Scrolling's lame. You can go sideways through the sections. You can scroll horizontally.

18:10

No, that's not what I want. I want to like one story, one view, best quote from it. Click on it to see more. Just kind of like you have the shorts. He wants an Instagram stories view.

18:12

But I need ads. I need to buy something. You need to buy something. This is really good. We've designed a fully designed merch store that I just haven't gotten around to launching. But maybe we should be putting our merch in this. Is this like a full app review now? Okay. With Jess. I also, for what it's like, I like the full flip vertical script thing. I think you should change the homepage so the articles are also like full screen flippers as opposed to scrolling. Because I don't want to scroll anymore. Scrolling's lame. You can go sideways through the sections. You can scroll horizontally. No, that's not what I want. I want one story, one view, best quote from it. Click on it to see more. Just like you have the shorts. He wants an Instagram stories view, not a feed. He wants to take the writing out of the journalism. I get it. I get it. But I would guess the internet's split on that. Feeds versus stories and people want discretion. It's an interesting thing. Do you guys remember like Vox launched on cards? You explain the world through cards. That's sort of what Sam wants. You want the playing cards. A bunch of companies did that. Yeah. I could vibe code that experience on top of this app and you could pick which one you wanted, right? No, never give users choice. You have to tell them what to do. But I do think that way I could just flip, flip, flip and give me a count of how many stories I haven't flipped through and I can be done because I just don't scroll stories anymore. Okay. Cover your ears, all journalists. Now, I do want to make a quick point about Can, but Dave, do you want to make a point about IPOs because I've teed it up and I thought you had a good point?

18:12

[SPEAKER_00] I put out this tweet a couple days ago because I saw Michael Dell posted that 38 years ago today, Dell Computer went public. We raised $30 million at an $85 million valuation when they went public. So like a series seed. [SPEAKER_01] Yeah. Yeah. [SPEAKER_00] And so I said, 84.5 at a $239 million market cap in today's dollars. If Anthropic had gone public at the same price and grown to its current valuation of $965 billion, a $100 investment would be worth $1.35 million today. [SPEAKER_01] Wow. [SPEAKER_02] It sounds like you're a degenerate teen on a gambling app there, Dave, with wanting those types of returns.

18:18

[SPEAKER_00] It's true. But the question is, and then I just said, bring back earlier public listings so that every person can bet on tech too. Right? So you do like prediction markets. I've made this point a lot of times, odd, but I think that it's interesting the responses that I got. I got a lot of people being like, there's definitely the same amount of growth from here on out for the next 30 years. And I'm like, people are idiots. You're going to see that would imply that Anthropic's going to be worth $11 quadrillion. [SPEAKER_01] Yeah. Those are the people who put $100 in SpaceX when it goes public. Those are not real people.

18:20

[SPEAKER_02] It's shocking to me the sheer amount of them though, Sam. Speaking of which, that $100, well, let's see. I don't think that $100 is worth. It's 96. It's shocking. But Dave, here's the thing. Here's the thing. I actually, look, it's funny because I'm the one who's pro-prediction market and I'm fairly libertarian. I'd be fine if there was no Sarbanes-Oxley and you had a much more open public market. I think that Sarbanes-Oxley is so stupid and bad for everyone. But here's the flip side is I don't know how you can believe that but then also be anti-prediction market because the whole point of all these regulations and accounting standards and why things aren't public and the cost, all the reasons it's expensive to be public is effectively, if in theory, if not practice, consumer protection so you're investing in real companies at real prices, right? And so it's a really hard one, right? Because I think I'm more in your camp of like there should be much lighter regulation. Everything is effectively gambling, right? Like don't worry about accounting standards. Don't worry about what's real and what's not real. Have at it. But I don't think you can simultaneously be against the prediction markets and pro earlier companies going public with lighter standards.

18:21

[SPEAKER_00] That's interesting. I'm willing to consider that. Because it's the same thing. It's the same thing. It's like here's some made up shit I made up that's worth 15. I mean, I don't know. It's not exactly the same thing, Sam, right? Like there is a something to going public and committing to report something to every man, right? Like, I don't know. We just had a company. I actually am proud of this. So TeamShares, which I seeded originally at an even lower price than Dell. It was like, love that company. $400K on a 4 million post was my first check into that when they were getting started. [SPEAKER_02] How are they doing in the public market?

18:24

[SPEAKER_00] They're public and they're up, but like that. But there's SPACs. [SPEAKER_02] Yeah, there's SPACs. SPACs are everywhere, I feel like. A robotics company did a SPAC today. There's like a SPAC every day.

18:26

[SPEAKER_00] Because it's so hard and expensive to deal with getting public, right? And the standards to it as well as just where you need to be scale-wise, the attention is so high. That's why you're seeing the SPACs. That's the closest thing you have to that type of asymmetry. Like TeamShares being public is obviously much bigger than Dell going public, but it's not many orders of magnitude. It's like one order of magnitude more, not two, not ten, right? And so like that is, you say let things be public earlier. TeamShares just went out and I'm really proud of is a great example of something that goes out at a really low price where there should be 100x ahead of it, in theory. Like that's not advice, but it's got the dynamics where it's not a quadrillion dollars, right, if it works. But it's really, really hard for companies to get out that way right now and expensive. And you just have to either accept much lower standards and totally change the dynamics, or, but then you have to accept the fact there's going to be a lot more fraudulent companies that are public and people are going to lose more money. And like it even relates to the indices, right? Which is like, the extent that everyone's gotten into passive investing and like auto clicking on the whole market, well then it really matters what's in the market, right? Like versus it being individual stock picking. So I don't know, it's a complicated set of issues. It's easy to agree with it intellectually, but when you get into the details, it's just a shame, right?

18:27

[SPEAKER_02] It's, well, yeah, it's also just like the rich got richer. [SPEAKER_00] Yeah, it's just a shame. Like the rich got richer. Like you don't need a public market when 1% or even 0.1% of the people have plenty of money to pay for everything.

18:31

[SPEAKER_02] Yeah, it's just, you know, I look at Dell and like a lot of people, we all grew up in the era that Dell and Apple and all of these great companies that are still public today, there was about a 3,000 X between when Dell went public and today. But, wait, but Dell also got taken private and then went public again and there's a lot of good things in the middle. Dell was very nuanced. There were drawdowns. I mean, I think I don't know if this is at all knowable. My bet is it's not. I'm very curious how many people actually put $100 into Dell at zero and they couldn't have rode it all the way through or Apple and still holds it today. Like no one does that, right? So it's a very intellectual argument based on like, oh, in theory, if you had done this, whereas in practice, I know one of our neighbors actually who's not at all in the tech world is one of these guys who put money in Apple in the 80s and now kept it and good for him. But I think it's very few people.

18:32

between when Dell went public and today. But Dell also got taken private and then went public again and there's a lot of good things in the middle. Dell was very nuanced. There were drawdowns. I don't know if this is at all knowable. My bet is it's not. I'm very curious how many people actually put $100 into Dell at zero and they could have rode it all the way through or Apple and still hold it today. Like no one does that, right? So it's a very intellectual argument based on, in theory, if you had done this, whereas in practice, I know one of our neighbors actually who's not at all in the tech world is one of these guys who put money in Apple in the 80s and now kept it and good for him. But I think it's very few people that actually do that, right? But I think that's why the Anthropic example is so interesting because it's happened in under five years. The Anthropic thing is funny because that's not a natural design pattern. It's also not based on anything other than speculation, really, right? And I think part of the story there is here's a funny thing. There's a bunch of people who are going to make a ton of money on Anthropic purely because they didn't have the time to sell, right? This happened a little bit with Facebook, but it's definitely happening with these companies. If you had gone to a rational early employee and it had been public, hypothetically, and they went from having $2 million to $20 million, they'd be like, "I'm out, right?" They'd be thrilled. The fact that it grew so quickly is you have this whole class of people that rationally should never have been concentrated in that asset and never would have been had they had the option and had been public, but because they had no option to sell and it happened so meteorically fast, there's all these random people that are now going to be worth hundreds of millions of dollars by accident, right?

18:33

Interesting. It's an interesting counter-argument to this whole thing, which is there's random walk. This happened at Facebook. There's so many people at Facebook, right, in the early days who were very junior engineers and got hilariously uber-wealthy in the first three years of their career purely because they literally couldn't sell and it appreciated so quickly, right? That would have never happened in the public markets because people would have been like, "I made 10 times my money. I can pay off my mortgage and pay for my mom. Fabulous, right?" Yeah. That's an interesting point, but somewhat niche compared to the bigger point, but interesting.

18:34

Well, it's just the point is that whether these things are public or private, you're locked up, you're not locked up, there's a whole bunch of random elements to this, and there are trade-offs to the whole thing. Okay, I'm going to end us by sending us over to the Riviera where none of us were. Britt, did you ever do Cannes in your media days? Oh, I've done many Cannes before, lots of rosé on yachts. Is it Cannes or Can? Cannes and Can if you're American and not Cannes, like Dave said at the beginning of this podcast. It's like padel, padel. Isn't it all relevant anymore? It was never relevant. No, what happened was it used to be legit and then—

18:44

No, Britt, it's always been an ad conference. Well, Dave, ads are legitimate. Ads are legitimate. It's a big business. The business of the internet, actually. Yes, but people get very confused. They think it's the Cannes film festival, which is actually cool and interesting. There are two festivals. There's the film festival and the media festival. The media festival is always about advertisers, creative industry, and creative people and media. And then big brands came in. So of course, Google and every big tech brand comes in. And then the whole thing does the same as South by Southwest did, which is jump the shark. And you have like Pepsi booths and all kinds of things.

18:48

Get real. It's an excuse for a boondoggle. And so people still go. You drink a lot of rosé on yachts. You go to a lot of parties. There's speaker panels and it's fine. But it's really hot in France right now. It's like 112 degrees. And I'm really glad I'm not there. Here's the thing. I heard someone was like, the cool event is the UTA beach. And I'm like, I can't think of a company that is less irrelevant than UTA. So what planet are we on?

18:51

I think my experience is a bunch of people, like the people who go there, like there's a guy on my team who went. And I was like, I'm not paying for that. He's like, it's fine. I'll just go for fun and keep working. He's like, if I find a deal, can I expense part of it? I'm like, fine. If you find a deal, what you're not going to find. So it just seems to me like a bunch of 20-year-olds who are underemployed or can work remote, pretending like they're doing important things when they're not. We're complaining about a lot of things. Britt's totally right. There's this underlying creativity component that is totally trumped by the business. It's a sales conference.

18:52

No, it's like the Super Bowl. It's like ads, which has been trumped by the boondoggle conference. I am now on record as a fan of the boondoggle known as Davos. You are? What? Because I did a ton of business there this year. So I am like, I was so productive in those three days that I said, I helped so many parts of my team by being there for three days. So I should probably be less cynical. A boondoggle done well is effective. Better than Zoom.

18:58

I will say though, my takeaway from—I always watch and Britt points out the media—the tech companies have all the money, put up these fancy things and I'm always interested in their messaging and their messaging this year basically boiled down to AI is not evil. Google was out there showing the pro-creativity side of AI, which I believe in, by the way, but their whole thing was you can still be creative with AI. It doesn't kill creativity. And then Meta, which riffing off the glasses launch, which I don't think was there, their slogan is the future is for everybody, which I think is really weird, but is also like, including your AI friends, right? But also like trying to say you too, person who doesn't feel like AI is for you, AI is for you, AKA don't regulate us out of oblivion. So my take was that the interesting threads to look at—we've talked about how tech companies could work to change the positioning around AI. And this is a small example in the creative. Guys, AI is evil.com is for sale for $100,000.

18:58

Oh God. What about AI is good.com? Well, AI is not evil might be available. We might need that. We can sell it at the Cannes festival for millions of dollars to tech. Join the AI is not evil consortium. I can't believe you're not a member of the AI is not evil consortium. Oh God. Sam is going to hold people to the— See, I'm going to have Claude make that for me in two seconds and go pitch it to all the heads of tech companies. They're going to cost like $12, not $10,000. Good luck with that.

19:05

And dear listeners, I hope you enjoyed having the quad back. I hope that you heard us because these editors are going to have to do some major interruption cleanup from this episode. But I hope you take that as a sign of how happy we are to be back together talking tech, talking shop. And I maintain it's going to be an exciting summer. And you should enjoy the Information's new vertical scrolling news feed because I kind of like it. I'm glad you made that change. And actually, I made the biggest decision I made was to make shorts the default in the video. That's a great decision. Good job. That is really the only consequential thing I did.

19:10

Compared to Instagram, there are far fewer bikinis in this though. And that is exactly where I sign off. Because anything else I say can and will be used against me in all aspects of my life. And so with that, I say thank you, dear listeners. And we'll see you back here next week for another episode of More or Less. Bye. Bye. See you later.

19:17

Some major interrupting cleanup from this episode. But I hope you take that as a sign of how happy we are to be back together talking tech, talking shop. And I maintain it's going to be an exciting summer. And you should enjoy the information's new vertical scrolling news feed because I like it. I'm glad you made that change. And actually, I made... The biggest decision I made was to make shorts the default in the video. That's a great decision. Good job. That is really the only consequential thing I did. Compared to Instagram, there are far fewer bikinis in this though. And that is exactly where I sign off. Because anything else I say can and will be used against me in all aspects of my life. And so with that, I say thank you, dear listeners. And we'll see you back here next week for another episode of More or Less. Bye. Bye. See you later. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less at Dave Morin at Lesson at Jay Lesson and as for me, I'm at Britt. See you guys next time.

19:18

Sure, but- The unboxing videos, I hate. I just hate. It's a huge thing. Not just the unboxing, but I think they call it ripping. It's like you rip open the trading card pack and there's like an entire market around this. I'm not into it, but all the guys I know that coach baseball teams around the Bay Area, they're like super into this and it's become like a major category. Okay, Britt, do you have prediction market thoughts and then I have one. The one thing I wanted to say, back to the voice of the female user and consumer, women are only like less than a quarter of Call She and Polly market right now. It really hasn't like struck with most of the women I know.

19:54

I actually think it's going to hit on with Meta if they do this and they build it into things like Instagram because I think- These things are already at large scale. Why would the behavior change? It has to be like betting on Love Island. Exactly, like The Bachelor and people are going to be like, oh, I bet it's going to be Jake, not Chris. No, it is. And the women are going to like get really vicious with each other about which guy's going to be the winner or what Taylor Swift's going to wear to her wedding. And I think it's going to become tribal and like really community oriented and women are going to get really into it. I was looking at the numbers on X

20:28

and it's like, you know, X is actually like 85% men. I could have told you that. Instagram's the opposite. You know, it's like all women. And so like really what Meta has to do to make this successful is make it successful on Instagram and then it will crush. That's what's going to happen. I like that angle. So I like Britt's angle. I have another angle. Intellectually, Polymarket and Kalshi are like really interesting businesses for all the reasons Sam talked about and markets and truth. And as a journalist, like there's, you sort of have a complicated relationship with them, but like there is a wisdom of crowds, undeniable, like truth, democratizing information element

21:04

that is very compelling. The problem is these companies are not acting particularly responsibly, especially Polymarket, which I think, you know, is actually kind of a shame because it's allowing, especially the media narrative around them, like they're dropping like low hanging fruit for the media to rightfully criticize and attack them, which is actually obscuring some of these more bigger issues. So the latest one is like Polymarket's influencer program, which you can agree or disagree that influencer programs are good or bad. You should say they should be labeled. But Polymarket has been directing influencers to make fake videos about fake trades

21:43

and reporting fake earnings for their influencer program. I think that's just like immature behavior. Honestly, I don't know. I'm not saying it's illegal. I'm just saying, but like if you're in a business like this that is so potentially interesting yet disruptive and you're going to go and like rattle some cages, okay, but rattle the right cages. Like just don't do stuff like that that honestly reporters who some of whom like aren't going to take the time to understand like the intricacies of the platform, which is actually quite interesting. Like actually, there are many things about the platform that are transparent and that you can see all the transactions,

22:20

which is unlike sports betting. Like all of those things just are going to get drowned out by like paying influencers to make fake videos, reporting fake earnings, and then like denying it and being nasty about it. So that is my current take on the prediction markets. They're not going away. And then I think things like, I think it's going to be hard for anyone to compete with them if they're based off like a fake currency. For sure. No one cares about a fake currency. Yeah, no question. Fake currency is dumb. Awesome. Okay, we got that topic. Let's talk. Will you guys indulge me on the AI talent wars? I feel like no one's interested in this, but the markets are.

22:54

It's just boring. It's like ping pong. You know why I think it's not boring? This is why I think it's not boring. It's just like people want money. AGI is not happening at any of these places. It's all like a big. But that's what's interesting about it. People are willing to do anything to make money. I think if you look at the layer of like who goes where, it's sort of profoundly uninteresting at this moment. And if you actually look like some of the people who went to meta from open AI have gone back. Like, you know, people who were at thinking machines who went back to open AI are going back thinking machines. What's interesting to me is like what investors

23:31

and journalists are like putting into it. And it makes me realize everyone is like deeply insecure about the future of all of these companies, that they're hanging so much on these signals that are objectively determined by like so many other factors. So to me, it's actually a comment of like the kind of no one knows how to model out the future of all of this stuff. So you're saying that's why they're switching? No, no. This is why everyone's hanging so much on this. Like why the entire, you know, Google stock is sliding X percent on the news. Like investors don't know. Like they're looking for like any signal or like any alpha, but they're picking the wrong one.

24:13

Well, that's the problem when you're modeling is infinity or zero. Interesting. I think you're right, but let me try to like build on that. I think here's the simple thing. Everyone's model looks like this. Infinity or zero and probability of infinity, right? Yeah, exactly. There's no, there's no more modeling than that. And like everything is this risk adjusted. Is this going to be infinity or not? And so as a result, because there's no actual year over year growth, there's no model, right? Then you take any incremental signal and especially because everyone else is going to trade on it. So you have to trade on it faster and it just like cycles into being a much bigger

24:45

thing than it should be, right? Because the models are so, I mean, we've all made these models when we were 22, right? When working for other people is like, you know what it's supposed to look like and then you know what it looks like now and like that's why it's happening because there is no, oh, this was growing at 32% and now it's going to grow at 38% compounded five-year CAGR. It's just infinity, zero probability of infinity. I would add to that maybe one more thing to add is that the level of commitment people have made to investing into these things is extraordinary. Like the amount of risk. Yeah, the stakes are high. The amount of risk that is actually

25:20

into all of these positions is really big and so I think you've also got that anxiety out there in the market because people don't know whether or not, like the example I'll give is yesterday I've been testing this eight Nvidia B200s, like basically stock eight of these things running GLM 5.2 and it's like really amazing. This is an open source model. Dave, what is the context of the B? Give us the... The B? Where did you get eight B200s? It's a long story. It's a friend of mine. That's like the thing I'm interested in is not what you're doing with them but where did you source eight of them? Ah, I can tell you later. Dark web, Sam. Those are also quite expensive.

26:03

Are you just like personally paying for those? We're renting them right now but they're... But they're really... That's a lot. That's expensive, Dave. And what is special about that B200? It's the top-end GPU that Nvidia makes, right? It's the new hotness. Oh, I see. It's the Blackwell. Yeah, they're the Blackwell. Blackwell 200s. It's not the Vera Rubin though. No, the Vera Rubin's... You can't get Vera Rubin's right now unless you're way out on the frontier. If it's a Porsche, then an H100 is like a Miata. Okay. And so, Dave, what's your takeaway from this? GLM 5.2 is a, you know, a new open source open weights model that was released like last week. And I was playing

26:42

with it in the cloud and was really shocked by its performance. And so, I started talking with a couple of... These are like hacker friends of mine actually from the 2000s. And we were like, let's see if we can get a rig together and test this. And we got this thing up and running and out of the box, it's doing 150 tokens per second, which is like 10 times what you get out of frontier models. And it's really extraordinary to play with. And this is an open source model. I was testing it doing the exact same coding tasks that I'm doing with frontier models. It's like as good or better. And so, that's like a really interesting thing that's happening right now.

27:21

Like, I'm sitting here going, I don't need the frontier models anymore. I can run these open weight models on standard NVIDIA hardware, get a bunch of people together, rent one, and it works really, really, really well. And this is kind of Sam's point for years that we've just been waiting for this moment to come. And I felt like I experienced it yesterday. So like, what does that mean for all of these enormous positions that are out there in these frontier labs? I think it's kind of shaky. Yeah, I agree with all that. But I also, Dave, I'm confused as to what heavy lifting you're doing here because I do a lot of freaking coding with these things and like a lot of tasks

27:58

and like, I'm not even close to pushing the limits of what they can do. Like, what are you doing in your spare time that's requiring this level of compute? Because I'm a pretty heavy user. Building OpenClaw. Yeah, but like, what specifically? Like, what are you finding you need that level of sophistication for? Well, this is not about I need it individually. It's that we've got an engineering team that's doing an enormous amount of token spend to do all of the engineering across the month. And so you start to ask this question, like, do you want to pay frontier prices for these things? Or do you want to buy the rig yourself and then be able to generate infinity token

28:37

at 10 times the speed? I totally understand. I'm actually just asking a more specific report. I gotta say, I'm like a pretty heavy user of this stuff. I build apps all the time, blah, blah, blah. I am shocked at some people's level of spend on these things, at least what they say they spend. Because I'm literally like, what are you possibly doing? What are you spending, Sam? What's your monthly bill? I don't even know, but it's probably a few thousand bucks a month, you know, like, which is a lot by, like, human standards. But it's not like, when people are like, oh, I'm crushing tokens, I'm spending, I'm like, doing what? Like, I've had this thing

29:08

do entire market analysis, back testing on crazy stock theories, and it's like, a hundred bucks. Like, what are we doing? I think the cron jobs is where they get you. No, no, no, it's loops. It's these, like, long running goal loops. Okay, what's a loop? Now, now I get to play the dumb person part. By the way, I think a lot of people are going to not know how much money they're spending and freak out because once you start getting really into this, like, I have, like, all these recurring jobs that run all the time, and I had no idea that some of them were spending way more than others, and I, I really bashed it down yesterday. I'm not usually someone who's,

29:47

I'm usually able to figure out how to spend the maximum amount of money on something, and I've tried, and I'm just like, I can't do it. I want to be the type of venture capital that's like, I spent $20,000 yesterday. Like, how? How? I'm trying. I'm like, I go to the thing, like, spend up to $1,000 complete this task and comes back and it's like, I spent $14.95. I'm like, that's not really... Dear listeners, please help Sam spend more money. Just tell him how to do it. We're all be grateful and then we don't have to talk about this. No, just kidding. Token maxing. Yeah, all I know is that, you know, you go through these eras where there's, like, the information version

30:19

of clickbait, like, the thing that, if you just put it in the headline, like, you're going to get a huge number of subs and it's changed over the years. Like, it's always something different. Believe it or not, it was Kubernetes at one point. That's how wonky our audience is. Then it was like SoftBank and the Vision Fund. And now, if you have AI bills in a headline, like, the information community is here for it. Now, interesting. being responsible journalists, we do not, you know, therefore, you know, maybe only one article a week will be on such topics. So maybe we're dumb journalists, but it is amazing how tuned people are. Is this? The other thing

30:54

that does really well is anything to do with data center loans and new structures there. But anyway, that's current window into the psyche of our subscribers. Speaking of which, guys, our new app will be out probably by the time this hits, I hope. So anyway, if... Oh, what are the new features? I'm glad you asked. Did you code it? No, we probably should have. How much did you pay in tokens? No, you did vibe code it. You just vibe code it via someone else who paid target it. I paid someone else to vibe code it. You double, got double charged, tokens and a person. Brett, let me tell you the two things I love about this app. First of all, video. So we brought TITV to mobile

31:30

where it has always belonged. You can see the awesome work of our team. There's a video tab that is a great way to just get, it's basically summarizing all the biggest news. So I love it. And then our deep research chat bot is also slightly redesigned and front and center. So I use it all the time to say what I miss, what is the biggest thing that was reported this week. I go into a meeting and say, what's the latest we've reported on this or that. And it's unlike ChatGPT or Claude or Gemini. It's like really smart about tech. So those are the two things and I'd love feedback on it. What else is happening, guys? Do you actually want feedback on it

32:04

or do you just want to say that is what you say? I am so, no one believes this about me. I like kind of went talking to someone. I was like, how do we get more feedback? I love feedback because you know what? I can't use all of our products every second of the day. So I would love feedback. And Sam, you have objectively been testing it for a month and giving me no feedback. So I'm sure it's perfect. Jess. I mean, all feedback is helpful. Jess, do you have your agents giving you feedback every day? Uh, no, but I've had some great agent use this week. It's been really helpful to me. But Britt, how does that, what does that look like? Oh, well, so I,

32:38

this is part of my recurring loop. I have my agents now. When I have an external meeting with like an LP or a founder or something, I get graded on how the meeting went. But for me, for them, the whole meeting, and I think that you should also, and then every time I have new content or something, um, you can also run it through and get a grade or get like, you know, edits. So I was just curious if you've like run your app through any sort of agent and had them pick apart the pros and cons. it's various points in this. I got some screenshots and I like would ask Claude if it had a perspective on it. So like if, you know, I wasn't sure of like X violent style

33:16

or Y violent style, but I haven't done it in a super sophisticated way. I have had like meetings that I just felt haven't gone well. And I have, I have proactively asked my agent, like how to improve the meeting or something I could have done better. It always just tells me that like, you know, there should have been more action. Like it just should have been more concrete, every aspect of the meeting. Like the beginning of the meeting should have been more concrete with the agenda. The end should have been more concrete takeaway. So, we should grade all of our episodes too. I'm using your app now and I'm actually, you changed my number one piece of feedback

33:47

I didn't give you, which is that the TITV shorts, I don't want to watch the whole episodes. The episodes are long and boring. I want the clips. I made shorts default for you. I made it. No, the shorts is great. That's the only thing. That was my only feedback. The shorts are good. You will never use Reels or TikTok again. Not quite. But it's really good. And Sam, the next piece of feedback you gave, which I've put to the team is to sort them based on a recommendation algorithm based on your interests. Yeah, I don't want random shit. But it's not that random. You know what you should totally do? Why don't you put some ads in this, in this shorts thing? Make a lot of money.

34:20

Because we don't have to because we already make so much money of TITV. You can always make more. Why don't you just like every fifth should be an ad. It can be an ad for more or less podcasts. Okay, you hear that, Lindsay? We got to put some ads. We got to put some ads in. Lindsay, fire it up. Every fifth scroll can be like, you know what everyone wants? Fifth scroll. Gemini. Woo! Throw in a more or less ad every fifth scroll. My favorite part of Instagram is the ads. So like, I like this, but I need ads. I need to buy something. You need to buy something. This is really good. We're, you know, we have designed a fully designed merch store that I just haven't

34:55

gotten around to launching. But maybe we should be putting our merch in this. Is this like a full app review now? Okay. With Jess.

35:05

I also, for what it's like, I like the full flip vertical script thing. I think you should change the homepage so the articles are also like full screen flippers as opposed to like scrolling. Because I don't like, I don't want to scroll anymore. Scrolling's lame. You can go sideways through the sections. You can, you can scroll horizontally. No, that's not what I want. I want to like one story, one view, best quote from it. Click on it to see more. Just kind of like you have the shorts. He wants an Instagram stories view, not a feed. He wants to take the writing out of the journalism. I get it. I get it. But you know what? I would, I would guess the internet's split

35:36

on that. Feeds versus stories and people want discretion. You know, it's an interesting, do you guys remember like Vox launched on like, it's now a big company that's now a smaller company that's now been split into four companies. But at the time it was cards. You explain the world through like cards. That's sort of what Sam wants. You want the playing cards. A bunch of companies did that. Yeah. I could vibe code that experience on top of this app and you could pick which one you wanted, right? No, never give users choice. You have to tell them what to do. But I do think that way I could just like flip, flip, flip and give me a count of how many stories

36:09

I haven't flipped through and I can be done because I don't, I just don't, I'm not going to scroll stories anymore. Okay. Cover your ears, all journalists. Now, I do want to make a quick point about Can, but Dave, do you want to make a point about IPOs because I've teed it up and I thought you had a good point? I put out, you know, I put out this tweet a couple days ago because I saw Michael Dell posted that 38 years ago today, Dell Computer went public. We raised $30 million at an $85 million valuation when they went public. So like a series seed. Yeah. Yeah. And so I said, you know, 80, that's around 200, I'm sorry, 84.5 at a $239 million market cap in today's dollars.

36:51

If Anthropic had gone public at the same price and grown to its current valuation of $965 billion, a $100 investment would be worth $1.35 million today. Wow. It sounds like you're a degenerate teen on a gambling app there, Dave, with wanting those types of returns. It's true. But the question is, and then I just said, like, bring back earlier public listings so that every, every person can bet on tech too. Right? So you do like prediction markets. I've made this point a lot of times, odd, but I think that it's interesting the responses that I got. I got a lot of people being like, there's definitely the same amount of growth from here on out for the next 30 years.

37:32

And I'm like, people are idiots. You're going to see, like, that would imply that we're going to, Anthropic's going to be worth $11 quadrillion. Yeah. Those are the people who put $100 in SpaceX when it goes public. Those are not real people. It's shocking to me the sheer amount of them though, Sam. Speaking of which, that $100, well, let's see. I don't think that $100 is worth. It's 96. It's shocking. But Dave, here's the thing. Here's the thing. I don't, like, I actually, look, it's funny because I'm the one who's pro-prediction market and I'm fairly libertarian. I'd be fine if there was no Sarbanes-Oxley and you had a much more open public market. I think that

38:04

Sarbanes-Oxley is like so stupid and like bad for everyone. But, but, here's the flip side is I don't know how you can believe that but then also be anti-prediction market because the whole point of all these regulations and accounting standards and why things aren't public and the cost, blah, blah, blah, blah, blah, all the reasons it's expensive to be public is effectively, if in theory, if not practice, consumer protection so you're investing in real companies at real prices, right? And so it's a really hard one, right? Because I think I'm more in your camp of like bucket, like there should be much lighter regulation. Everything is effectively gambling, right?

38:39

Like don't worry about accounting standards. Don't worry about what's real and what's not real. Have at it. But I don't think you can simultaneously be against the prediction markets and pro earlier companies going public with lighter standards. That's interesting. I'm willing to consider that. Because it's the same thing. It's the same thing. It's like here's some made up shit I made up that's worth 15. I mean, I don't know. It's not exactly the same thing, Sam, right? Like there is a something to going public and committing to report something to every man, right? Like, I don't know. We just had a company. I actually am proud of this. So TeamShares,

39:12

which I seeded originally at an even lower price than Dell. It was like, love that company. $400K on a 4 million post was my first check into that when they were getting started. How are they doing in the public market? They're public and they're up, you know, but like that. But there's SPACs. Yeah, there's SPACs. SPACs are everywhere, I feel like. A robotics company did a SPAC today. There's like SPAC every day. SPACs have got a SPAC. Because it's so hard and expensive to deal with getting public, right? And like the standards to it as well as just the like where you need to be scale-wise, the attention is so high. That's why you're seeing the SPACs.

39:45

That's the closest thing you have to that type of asymmetry. Like TeamShares being public is obviously much bigger than Dell going public, but it's not many orders of magnitude. It's like one order of magnitude more, not two, not ten, right? And so like that is, you say like, let things be public earlier. TeamShares just went out and I'm really proud of is a great example of something that goes out at a really low price where there should be 100x ahead of it, in theory. Like that's not advice, but like it's got the dynamics where it's not a quadrillion dollars, right, if it works. But it's really, really hard for companies to get out that way right now and expensive.

40:18

And you just have to either accept much lower standards and totally change the dynamics and like, or, but then you have to accept the fact there's going to be a lot more fraudulent companies that are public and people are going to lose more money. And, you know, like it even relates to the indices, right? Which is like, the extent that everyone's gotten into passive investing and like auto clicking on the whole market, well then it really matters what's in the market, right? Like versus it being individual stock picking. So, I don't know, it's a complicated set of issues. It's easy to agree with it intellectually, but when you get into the details,

40:48

it's just a shame, right? It's, well, yeah, it's also just like the rich got richer. Yeah, it's just a shame. Like the rich got richer. Like you don't need a public market when like 1% or even 0.1% of the people have plenty of money to pay for everything. Yeah, it's just, you know, I look at Dell and like a lot of people, like we all grew up in the era that Dell and Apple and, you know, all of these great companies that are still public today, you know, there was about a 3,000 X between when Dell went public and today. But, wait, but Dell also got taken private and then went public again and there's a lot of good things in the middle. Dell was very nuanced.

41:23

There were drawdowns. I mean, I think, I mean, I don't know if this is at all knowable. My bet is it's not. I'm very curious how many people actually put $100 into Dell at zero and they couldn't have wrote it all the way through or Apple and still holds it today. Like no one does that, right? So it's a very intellectual argument based on like, oh, in theory, if you had done this, whereas in practice, I know one of our neighbors actually who's not at all in the tech world is one of these guys who put money in Apple in the 80s and now kept it and good for him. But like, I think it's very few people that actually do that, right? But I think that's why the Anthropic example

41:57

is so interesting because it's happened in under five years. Like the Anthropic thing is so funny because that's not a natural design pattern. It's also not based on anything other than speculation, really, right? And like, I think part of the story there is like, like, here's a funny thing. There's a bunch of people who are going to make a ton of money on Anthropic purely because they didn't have the time to sell, right? Like this happened a little bit with Facebook, but it's definitely happening with these companies. We're like, if you had gone to a rational early employee and it had been public, hypothetically, and they went from having $2 million to $20 million,

42:33

they'd be like, fuck yeah, I'm out, right? They'd be thrilled. The fact that it grew so quickly is you have this whole class of people that rationally should never have been concentrated in that asset and never would have been had they had the option and had been public, but because they had no option to sell and it happened so meteorically fast, there's all these random people that are now going to be worth hundreds of millions of dollars by accident, right? Interesting. It's like an interesting counter-argument to this whole thing, which is like, there's random walk- This happened at Facebook. There's so many people at Facebook, right, in the early days who were very,

43:06

very junior engineers and got hilariously uber-wealthy in the first three years of their career purely because they literally couldn't sell and it appreciated so quickly, right? That would have never happened in the public markets because people have been like, oh, I made 10 times of my money. I can pay off my mortgage and pay for my mom. Fabulous, right? Yeah. That's an interesting point, but somewhat niche compared to the bigger point, but interesting. Well, it's just the point is that whether these things are public or private, you're locked up, you're not locked up, there's a whole bunch of random elements to this, you know, and there are trade-offs

43:39

to the whole thing. Okay, I'm going to end us by sending us over to the Riviera where none of us were. Britt, did you ever do can in your media days? Oh, I've done many, many cans before, lots of rosé on yachts. Is it con or can? Con and can if you're American and not can-ess, like Dave said at the beginning of this podcast. It's like paddle padel, paddle padel. Isn't it all relevant anymore? It was never relevant. Is this like now the least relevant thing in the world? Yes, it is. No, what happened was it used to be legit and then- No, Britt, it's always been an ad conference. Well, Dave, ads are legitimate. Ads are legitimate. It's a big business.

44:18

The business of the internet, actually. Yes, but people get very confused. They think it's the con film festival, which is actually cool and interesting. There are two festivals. There's the film festival and the media festival. The media festival is always about advertisers, creative industry, like creative people and media. And then, big brands came in. So, of course, Google and every big tech brand comes in. And then the whole thing does the same as South by Southwest did, which is Jump the Shark. And you have like Pepsi booths and all kinds of things. Get real. It's an excuse for a boondoggle. And so people still go. You drink a lot of rosé and a lot of yachts.

44:56

You go to a lot of parties. There's like speaker panels and it's fine. But it's really hot in France right now. It's like 112 degrees. And I'm really glad I'm not there. Here's the thing. Like I heard someone was like, the cool event is the UTA beach. And I'm like, I can't think of a company that is less irrelevant than UTA. So like what, like what planet are we on? I think my experience is a bunch of like people, like the people who go there, like there's a guy on my team who went. And I was like, I'm not paying for that. He's like, it's fine. I'll just go for fun and keep working. He's like, if I find a deal, can I expense part of it? I'm like, fine. If you find a deal,

45:31

what you're not going to find. So like, it just seems to me like a bunch of 20 year olds who like are underemployed or can work remote, like pretending like they're doing important things when they're not. We're complaining a lot of things. So Britt's totally right. There's this underlying like creativity, like a word component that is like totally trumped by the business. It's a sales conference. No, it's like the Super Bowl. It's like ads, which has been trumped by the just boondoggle conference. I am now on record as a fan of the boondoggle known as Davos. You are? What? Because I did a shit ton of business there this year. So I am like, I was so productive

46:10

in those three days that I said, I helped so many parts of my team by being there for three days. So. I should probably be less cynical. A boondoggle done well. It is good to hang out in person and get things done. Is effective. Better than Zoom. I will say though, my takeaway from, I always watch and Britz points out the media, the tech companies have all the money, you know, put up these fancy things and I'm always interested in their messaging and their messaging this year basically boiled down to AI is not evil. Now, Google was out there showing the pro creativity side of AI, which I believe in, by the way, but their whole thing was like, you know,

46:51

you can still be creative with AI. It doesn't kill creativity. And then Meta, which riffing off the glasses launch, which I don't think was there, their slogan is the future is for everybody, which I think is really weird, but is also like, including your AI friends. Right. But also like trying to say you too, person who doesn't feel like AI is for you, AI is for you, AKA don't regulate us out of oblivion. So my take was that the interesting threads to look at. We're sort of, we've talked about how tech companies could work to change the positioning around AI. And this is a small example in the creative. Guys, AI is evil.com is for sale for $100,000. Oh God.

47:37

What about AI is good.com? Well, AI is not evil might be available. We might need that. We can sell it at the con festival for millions of dollars to tech. Join the AI is not evil consortium. I can't believe you're not a member of the AI is not evil consortium. Oh God. Sam is going to hold people to the... See, I'm going to have Claude make that for me in two seconds and like go pitch it to all the heads of tech companies. They're going to cost like $12, not $10,000. Good luck with that. And dear listeners, well, I hope you enjoyed having the quad back. I hope that you heard us because these editors are going to have to do some major interrupting

48:13

cleanup from this episode. But I hope you take that as a sign of how happy we are to be back together talking tech, talking shop. And I maintain it's going to be an exciting summer. And you should enjoy the information's new vertical scrolling news feed because I kind of like it. I'm glad you made that change. And actually, I made... The biggest decision I made was to make shorts the default in the video. That's a great decision. Good job. That is really the only consequential thing I did. Compared to Instagram, there are far fewer bikinis in this though. And that is exactly where I sign off. Because anything else I say can and will be used against me in all aspects

48:54

of my life. And so with that, I say thank you, dear listeners. And we'll see you back here next week for another episode of More or Less. Bye. Bye. See you later. If you enjoyed this show, please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for at more or less at Dave Morin at Lesson at Jay Lesson and as for me, I'm at Britt. See you guys next time.

Reading tools

Type to find a passage

Appearance
Ask this transcript

Notebook

Export Markdown

Add a note