This Week in Startups - Just how frothy is the AI Bubble anyway? | E2195

Episode Date: October 21, 2025

Today’s show: *Is the AI bubble going to POP? And if so, how much of our economy is it going to take down with it?On a new TWiST, Jason, Alex, and Lon debate the size and scale of the AI bubble, whether poppage is imminent, and just who we can blame for the last few economic crises. (HINT: Not the tech industry!)PLUS why Jason STILL thinks the current AI cycle will be the biggest in tech history, how this morning’s AWS mega-failure might impact the company’s earnings, the spread of AI fakes in the wake of this weekend’s No Kings march, why LLMs may soon put your broker out of business, and much much more.**Timestamps:** (00:02:26) Is AI a BUBBLE? And if so, how BIG?(00:05:12) Are all the major bubbles and corrections always Wall Street’s fault?(00:08:39) Why Jason thinks this AI will be the biggest cycle in the history of technology(10:00) LinkedIn Ads: Thanks to our partner @LinkedInMktg. Launch your first campaign and get $250 FREE when you spend at least $250. Go to [http://LinkedIn.com/ThisWeekinStartups](https://www.linkedin.com/thisweekinstartups) to claim your credit.(20:00) Thank you to our partner @trustvanta! TWiST listeners automate your SOC2 and get $1,000 off at http://www.vanta.com/twist(00:23:01) Will this morning’s AWS failure impact Amazon’s earnings?(30:00) Thanks to our partner @Openphone! Streamline and scale your customer communications with Quo. Get started free, plus get 20% off your first 6 months, at [http://quo.com/TWiST](http://quo.com/TWiSThttp://quo.com/perks/twist)(00:32:49) People couldn’t spot AI fakes from this weekend’s protests… are we cooked?(00:49:54) Jeremy Strong shows off his Mark Zuckerberg impression… and Jason is impressed*Thank you to our partners:LinkedIn Ads: Thanks to our partner @LinkedInMktg. Launch your first campaign and get $250 FREE when you spend at least $250. Go to [http://LinkedIn.com/ThisWeekinStartups](https://www.linkedin.com/thisweekinstartups) to claim your credit. @trustvanta! TWiST listeners automate your SOC2 and get $1,000 off at http://www.vanta.com/twist@Openphone! Streamline and scale your customer communications with Quo. Get started free, plus get 20% off your first 6 months, at [http://quo.com/TWiST](http://quo.com/TWiSThttp://quo.com/perks/twist)*Check out Jason’s suite of newsletters: https://substack.com/@calacanis*Follow TWiST:Twitter: https://twitter.com/TWiStartupsYouTube: https://www.youtube.com/thisweekinInstagram: [https://www.instagram.com/thisweekinstartups](https://www.instagram.com/thisweekinstartups/)TikTok: https://www.tiktok.com/@thisweekinstartupsSubstack: [https://twistartups.substack.com](https://twistartups.substack.com/)*Subscribe to the Founder University Podcast: https://www.youtube.com/@founderuniversity1916All rights for the video shared and linked below belong to the original copyright holder, and we are using the footage under the principle of fair use.Listen here:Apple: https://podcasts.apple.com/us/podcast/this-week-in-startups-audio/id315114957Spotify: https://open.spotify.com/show/6ULQ0ewYf5zmsDgBchlkr9Overcast: https://overcast.fm/itunes315114957/this-week-in-startups-audioMore from us:Twitter: https://twitter.com/twistartupsInstagram: https://www.instagram.com/thisweekinstartupsOfficial site: https://thisweekinstartups.comSubscribe to our YouTube to watch all full episodes:https://www.youtube.com/channel/UCkkhmBWfS7pILYIk0izkc3A?sub_confirmation=1Subscribe to TWiST Clips for all the best moments:https://www.youtube.com/channel/UCS7tJlcUA6PzVHEMo-X7ddg?sub_confirmation=1#startups #entrepreneurship #investing #angelinvesting #tech #news #business

And if stocks go down, they have less money. So their concern is that we're going to take this kind of like a stick to the face. And I am curious what you think about how big the bubble could be today if you believe there is one. Yeah. So a couple of things. These economists have predicted 11 of the last seven recessions and corrections. So they are extremely good at this. This week in startups is brought to you by Quo. Quo, formerly open phone, is the number one business phone system. that streamlines your customer communications. Get started free plus get 20% off your first six months

at quo.com slash twist. Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Get $1,000 off for a limited time at Vanta.com slash twist. And LinkedIn ads. Start converting your B2B audience into high

quality leads today. Launch your first campaign and get $250 free when you spend at least 250. Go to LinkedIn.com slash this week in startups to claim your credit. It's October 20th, 2025. We got a full docket and let's get to work. There's a lot of debate about the AI bubble. Alex, Wilhelm. It's here again. Alex, your thoughts. Yes, sir. You know, this big debate. I saw that a lot of people are chiming in. A lot of people are chiming in.

Yeah, enough people have been chiming in, Jason, that I think the narrative inside both the worlds of finance and tech is, this is a bubble. The question is just how much when it will pop and what the damage will be. You shared a New York Times op-ed with us that was titled, Warning, Our Stock Market is Looking Like a Bubble, written by Jared Bernstein and Ryan Cummings, two economists, and they're both former members of the Council of Economic Advisors.

under the Biden administration. So they're in the game, if you will. Their argument's pretty simple. They think that AI investment plans are too lofty. They found Open AI's $1 trillion spend common to be a little bit incredible compared to its revenue. They also said that stocks today are too expensive

looking at historical price earnings ratios. And they say that a lot of those gains are AI predicated. So essentially the AI companies are doing very, very well, creating risks. And that if this bubble does pop, yes, many of the companies that are making investments are well capitalized. But if the market falls, Jason,

There's a thing called the wealth effect. And as you and I know, wealthy people own a lot of stocks. And if stocks go down, they have less money. So their concern is that we're going to take this kind of like a stick to the face. And I am curious what you think about how big the bubble could be today if you believe there is one. Yeah. So a couple of things. These economists have predicted 11 of the last seven recessions and corrections.

So they are extremely good at this. You can't go wrong, saying a correction is. coming because, yeah, there will be a correction. There always a correction every 10 years. So just thinking from first principles here, if you look at the current revenue footprint of AI companies and you know, you've got to think we're closing in on $100 billion in revenue. You know, if you have Open AI with 10, 13, whatever it is. And sorry, hold on a second, guys. Do we have a, are we projecting back the feed of the live stream into the Zoom? I don't know if we can not do that because it's on a delay, but let me just put me on

self-on speaker mode here. Yeah. Okay. Three, two, if we would look at the total addressable market, which we started our TAM project here, it's probably linked in the show notes. You know, we're pretty clear there's going to be in the short to midterm, short term, one, two years, midterm, two to five years, long term, you're five years out. But in that short to midterm, we will see a trillion dollars, two trillion dollars show up in revenue. That's my belief. It will come $50 billion a year consistently for five years, six years, seven years, and we'll just, we'll get right to that one trillion. It might take seven, might take five, could take ten. But there's a trillion dollars in Tim. I think we would all agree. Yeah, Alex, out there.

I very much agree, absolutely, if not more over time. And keep in mind, we're talking about this often through the lens of how much do people spend today on software for their staffers and thinking about that as kind of a proxy for the total available market. I think also we should consider the amount of labor, just the cost of humans, because that's what we're going to try to replace quite often with automation. And so when we think about the Tam Jason, yes, it is the entire software industry on one hand. But it's also, I think, the labor market on the other. And added together, that's most of the economy. Right. So we're building this bottom up and top down Tam, and we'll keep sharing that every week. It's important for people to keep that in mind. What is tweaking folks is the scar tissue from the Great Recession caused by financial malfeasance by Wall Street. Had nothing to do with tech. In fact, it had nothing to do with real estate. It had to do with financial manipulation. Now we look at the dot-com era. What caused that?

Was it actually the tech companies? It was not. It was financial manipulation by Wall Street that was taking companies public. And you remember, Henry Blodjeet got banned from ever working in securities. He went on to do business insider because he was, you know, on one side, the firm he was at was taking companies public. And on the other side, he was saying, like, this company's, you know, do-do, a dog, whatever, I think was the exact quote. This was like garbage, you know, on the internal communications. So the finance were the cause of that, too.

So I looked to this third bubble bursting, and I say to myself, well, what could call? Oh, and by the way, when we had the mini correction with Silicon Valley Bank and First Republic Bank, what was it, Alex? What caused that? It was a decision by the Silicon Valley Bank company to shake up its balance sheet to deal with losses that were derived from changing interest rate prices and therefore overpricing of debt that led to a confidence crisis that led to a run on the bank that was then bailed out so you described this as the people who cost it were in banking correct oh finance oh i see your point it was a finance driven correction it was not

predicated on tech companies yes okay now i see what you're asking yeah it had nothing to do with the tech companies who had you know loans from silicon valley bank or their deposits it looked like so we i just went through the three of our lifetime the three major ones now i also was, I guess, 17 years old when we had Black Friday, but I'll leave that one out. Just talk about the three in modern history. So, three for three were vibrant markets, but where the finance people started to get frisky. So this is where the two economists, who I assume have never really done anything in the real world, almost universally, these economists are giving advice, but they've never been on the field.

In other words, you know, it's like Stephen A. Smith talking about basketball, you know, when he's played like three games in his career and had like one point average per game. Like they're just talking heads. They do the best they can. They analyze stuff. Yeah, sure. They have some deep knowledge of markets, but they're not the one building the companies. As somebody who's in finance and company building, I can tell you, this always has the same pattern. People get excited.

People are building stuff that's obviously going to change the world. and provide massive value to the human species. Then bankers go, great. How can we help accelerate this? And then they start the acceleration process because they get paid. How, Alex? How do bankers and finance people get paid typically?

Fees, fees, fees on deals. And you know what? Why not do some more deals and get yourself some more fees? Correct. So one of the great things about our system here in America is we have entrepreneurs, we have capital allocators, and then we have these finance people.

They always come in at the last minute, and they F everything up. They take it too far, and it's literally like Wolf of Wall Street. So the question is- Can I ask you about that, Jason? About the five people coming in. Yeah, so in this fourth, the real question becomes in this fourth boom cycle here that we're talking about, are we seeing them run amok or not? But you have a question, go ahead.

Yeah, yeah. Well, your point about finance was coming in, I was just thinking about the 2021 kind of ZERP era venture capital boom that we all lived through and kind of the deflation that came afterwards. Do you think that Tigers arrival later in that cycle with an enormous checkbook and making a lot of quick deals is an example of the finance world kind of coming in at the last minute and causing more problems? Or is Tiger, you know, deeply integrated enough into the tech ecosystem that their massive increase

in investment was not an example of that? So an easy way to do this is, are they doing it to make quick, money. If they're doing it to make quick money, I would put them in that finance genre, as opposed to capital allocators. So when you have Masayoshi-san come in, or you have, you know, Japan came into Hollywood during a period of time. When people come in with big chip stacks, that's a sign of a top when they start doing this. And when finance people just do these kind of deals. And on the margins, we are starting to see people come in and say, hey, maybe we should, you know, spack this.

Maybe we should do these round-tripping things. Maybe we should do speculative things. And when you get to that rampant speculation by the finance peoples, that's where I start to pay attention and say, okay, yeah, we're top ticking here. Founders are constantly asking me, hey, where can I find my first customers? I need a customer. I'm talking about business-to-business advertising and marketing your product to other businesses.

And to do that, the greatest place in the world. world, hands down is LinkedIn ads. LinkedIn's advertising tools aren't just going to throw your message around to anybody clicking on random ads. No, these are not randos. You can pick decision makers based on the data that you know inside of LinkedIn works, the job title, the industry, how about the size of the company? Or maybe you're working on startups and you just want people in New York City and Miami because that's where your first two sales offices are. There's over a billion people on LinkedIn, and that includes 130 million decision makers and 10 million C-level executives. These are the big fish, right?

These are the whales. Business to business markers report 2.5 times higher returns on their ad spend on LinkedIn versus the other social platforms. It makes sense, right? So to start converting your B2B audience into high-quality leads and customers, we're going to give you $100 credit towards your next campaign. That's how confident LinkedIn is that you're going to get value. They'll just give you the $100 to prove it to you.

LinkedIn.com slash this week in startups. LinkedIn.com slash this week in startups. No spaces, no dashes to claim that hundi. LinkedIn.com slash this week in startups terms and conditions do apply. If you look at Sam Altman's strategy in just in the last three to six months, it's raise money from everyone for everything. That would be a sign that we could be his strategic approach, the last six months is a sign that a bubble is building when you say i want to deploy one trillion dollars in infrastructure and then you do a deal with amd with invidia and and did they do another one with another chip company broadcom with brockom so that those three to me with the round-tripping nature of these um and them wanting to invest money they don't have yet into infrastructure that is bringing

in, you know, whatever, 10 billion, 15 billion, something in that range is, you know, where they're trending or just passing with their run rate. That's when I would say, yes, we are starting to experience financial, unnatural financial acts. Unnatural financial acts. Do you think that part of this is also driven by, there's a lot of behind the scenes, startup and tech-driven use of AI? Like, we talked about that just recently, like the 50 top AI apps that are being used by startups, and it's an interesting list that we were going through. Yeah.

But I feel like on the consumer side, like even chat GPT, wildly popular, 800 million global users now, but only 5% are paying. And I feel like we haven't seen. I feel like that's what's fueling a lot of this perception that it's a bubble is there isn't an app everyday people are paying tons of money for. It's all still corporations and enterprise. and startups. And so from the public's perspective, they're not seeing everybody using replet or whatever. They're just, or lovable.

They're just seeing like none of my friends is paying for an AI app, so it's a dead technology. Yeah, so that's Lon Harris checking in here, our co-host. For those of you listening, everybody. Sorry, not introduced you there. So I think the corporate spend on AI

is very real. corporations are not going to invest in this stuff unless they have a clear path, a line of sight, we would say, in the business to it working. And we all know here, as we use these products and services, that they're working. And we watch startups use it when we see tax GPT, you know, make CPAs 10% better at their jobs, 20% faster at their jobs. And we're seeing this across the portfolio of startups we invest in from Founder, University and the Launch Accelerator, that they're actually making customers low double digits,

more productive, you know, every three to six months. And that is the reality of what's happening today. There's good stuff happening on the field. But the finance people are seeing that and saying, well, let's accelerate it. But there's a natural pace of technology getting absorbed into organizations. And you two gentlemen can experience it yourselves. You say, oh, I'm going to try to make some, I'm going to do some research using these products. And two years ago, it was like, well, can we even trust these hallucinating results? The facts are all wrong. And then all of a sudden, you skip over and you're like, wait a second, this stuff is great. It's actually doing it correctly. And we start using producer Claude during the show. And it actually is giving us the right answers

in real time. And that all occurred over 24 months. So two things can be true at the same time. this will be the greatest investment cycle in the history of technology. That's absolutely a certainty in my mind. So let me just state that again. This will be the absolute greatest returns in the history of the technology industry. It'll dwarf, you know, the PC revolution, the smartphone revolution, and even the internet revolution. I am 100% certain of that. So then it's just the timeline.

And are people investing in things that will be dollar. Are they investing in that too high over price? That's yes and yes. But will it crash the economy? No. Will the people who are losing money become destitute? Nope. It's not that amount of money.

The worst case scenario is that it takes three or four more years than the aggressive estimates of this technology landing. And if that happens, what you'll see is some companies just pause spending or cut these. aggressive spending programs by 50%. And that happened with fiber back in the day. They overbelled fiber. That's the analogy I've been using for two years.

And now everybody seems to be using that one. So it's not a bubble, but there are frothy moments occurring. I would say this is a nice frothy cappuccino. And it will become a flat white shortly where you don't put a ton of foam in it. You just have like just the warm milk or, you know, with the with the coffee. What's that trigger point? Because one thing that we often see is a moment in which the market gets shaken a little bit. And if you go back to early 2025, Jason, we had the Deep Seek R1 moment when the Chinese company

Deep Seek claimed that they trained a very good model for a very low price, really scared a lot of folks off of Nvidia for a bit. So for people that are watching, what might be coming that could kick off that Cappuccino to flat white moment. Yeah. So what would deflate all this foam is pretty straightforward. one of these data center projects will go belly up or one of these data center projects will have their valuation just get walloped so there are you know like there's a core weave right

and so let's say core weave has a lot of debt and they're buying a lot of capacity and they're building a bunch of projects there might come a point where they say you know what we have to pay back and we've got loans and we've got payments to make. Yeah, we're going to pause those two places where we broke around. And even though we are, I don't know, two billion dollars into building out these new data centers, we're going to pause them and we're going to sell them to Zuckerberg or Google. We're going to pack it in and we're going to slow down. Then people look at the stock and go, wait a second. You know, we were basing it on, you know, the price earnings ratio. What's the price earnings ratio of?

CoreWeave at this point, what's their debt? You know, you start looking at those aggressive neo clouds and hypers. Yeah, you just might have a come-to-Jesus moment. Well, Jason, yeah. The price earnings ratio is in A, because it doesn't make money. But if you would like the current price sales ratio for our friends over at CoreWeave, it's 14.81 on a trailing basis according to white charts. Right. So 14 to 1 going backwards is, yeah, that's That's pretty crazy. And you could see somebody like Corrieve say, hey, we're going to take a pause. Then you would see, you know, a Zuckerberg say, oh, we were, you know, I don't know what he said

at that Trump tech dinner with the tech leaders, but he said, I'm going to put 400 billion, maybe 600 billion, he was kind of spitballing there. Maybe he says, you know what? We're going to only spend, you know, 25 billion a year for the next two years. We're going to pause our spending. And then, you know, in video stock would have some headwinds. price earnings ratio is not insane, but it's robust. And maybe then Azure says in Microsoft, as the story broke the past week, that maybe Microsoft's

like, we don't want to build all this infrastructure for chat cheap. They should go do it on themselves because we don't want to overbuild. So that's really what's at the core of this, is the build out of the data centers. And then we talked about on All In this past week and you start Shemah tweeting about it, that there's not enough energy to build some of these data centers. So you had two or three data centers say, hey, we're going to take a pause because the people in these cities are concerned

that their electricity bills are going up and that'll be bad for our reputations if we cause consumer pain and suffering. So all of those things would, to me, the data center build out, Nvidia stock correcting, would be a sign that were overheated. Now, what does that mean?

Well, that means people who made 30% on their money this year and 30% last year riding this boom, you know, might have only made 7% the average, right? Or 15% double the average. But they didn't make four times the historical average on their money. The worst. We're doomed. No, we're not.

AI is revolutionizing the way people work. It's increasing our efficiency, obviously, and things are moving faster. But even when you're working at startup speed, big enterprise clients still expect you to meet all of their security and compliance requirements. That's why your startup needs Vanta. they're going to help you automate compliance, not just for SOC2, but also for HIPAA, GDPR. Plus, VANTIS AI is going to streamline your security reviews with automated questionnaires and a trust center that demonstrates just how safe and secure your offerings are to potential customers.

And that's just one way that Vanta is using AI to make it easier to track and monitor your cybersecurity. They've also got AI-generated vendor summaries that instantly alert you to, fresh insights and client needs, making sure nothing slips through the cracks. So here's your call to action. My listeners get $1,000 off just by going to vanta.com slash twist. That's v-a-n-ta.com slash twist for $1,000 off. Thanks, Vanda. It was, this was back in September.

Trump asked Zuckerberg publicly how much meta was going to invest in the U.S. in the coming years. Zuckerberg in the moment says at least $600 billion through 2028. And then later they caught him on a hot mic saying to Trump, sorry, I wasn't ready. I wasn't sure what number you wanted to go with. So, you sort of just made that. State capitalism, baby. I mean, listen, it's, I think people are trying to curry favor with the administration by saying they're investing in America, which is this administration's, you know, major marketing effort. their major way to say we should take out regulations is there's a lot of investment going on here.

To a certain extent, they're succeeding. Even if people are exaggerating the numbers by a multiple of 2x or 3x, it's still going to be a huge spend. And that's a good thing for America. So anyway, I'm not, I would say we're frothy and those round-tripping deals and specifically the behavior of chat GPT should make economists you know, sound an alarm, but I don't think this is like five alarm fire. This is like, hey, it's dry season here. Maybe there's some Tinder on the ground. We should rake some stuff up and maybe fill the reservoirs with water.

But I'm not calling like a DefCon 4 or whatever, DefCon 5 right now. It's just a little bubbly. And if we lose money, the people who are going to lose the money or the people who can afford to lose the money. All right. Yeah. Next up on the docket, AWS went down and it's still going down, Jason. a really big problem in the U.S. East One region of the cloud giant. This brought down companies including Coinbase and Robin Hood, Snap, banks in the UK,

and also Signal quite a mess to wake up to on a Monday morning. However, things have gotten a little bit better, so it's not a snow day here in tech. I wanted to bring this up because I'm super curious what you think founders should do in moments like these. If you lose access to your cloud compute, what should you do? And also, do you recommend that founders today pursue a multi-cloud strategy to help possibly mitigate some of the risks that being on just AWS might bring. Yeah. So this happens to every provider every couple of years,

which means we experience these type of outages, I would say on average every six months right now. Usually they're corrected so quickly that they fade from memory. This one seems to have hit a decent swath of important services. When your finances are on Robin Hood Coinbase and Venmo, and all three services don't work, that's a reminder. Maybe I should have multiple bank accounts with multiple ways to get to money. So this morning, Venmo didn't work for me.

I was sending a birthday present to a family member. And it was like, wait a second. And then I was like, oh, this must be part of the AWS thing. All of this should, yes, you're correct. Founders should have their systems able to roll over to clouds. That still is a bit complicated, but it's not impossible to do. and this is why sometimes people will, in fact, have their cloud spend across multiple instances and then use essentially a smart router to send half the traffic to AWS and a quarter of it to Azure.

But this is infrastructure cost and it slows down your production to build this kind of resiliency into your systems. But yeah, it's a good warning for everybody. and also warning for people to understand that projects that seem like they're decentralized, typically have a single point of failure. So as much as people want to believe that things are decentralized, and in fact, the internet is built to be decentralized. That's why we were all still using the internet and Zoom still worked. And if Zoom doesn't work, you can go over to, you know, Google Meetup or whatever they call

Google's product or Slack. Utters, Google Meet, yeah. You can just move around the internet, but if your service is tied to one cloud, that's no one-o. Just a good reminder for people. I would say early-stage startups shouldn't waste the time and money on this. They should have backups, but it is a lot to be multi-cloud. You need to have a team doing that. I wonder if any of these services did fail over successfully from AWS to their Azure or Google Cloud or Oracle instances.

And yeah, there are websites that track uptime. Yeah. And here's the thing. This could show up in Amazon's earnings. So we'll see. Make a note of that, Alex, when we cover Amazon's next earnings. They sometimes will have to give a credit to all those people for downtime. They get a penalty, basically, built into their contract.

So I wonder if we'll see Amazon stock get hit today or this week that people think, oh, they're going to lose 1% of their cloud revenue for the quarter or 2% or something. Yeah, Amazon shares are up. 1.3% today, which I think just goes to show how large of a company it is, that they can have this public of a fiasco in one of the most profitable entities. And everyone goes, well, the market's up. I do want to point out that producer, Claude, our dear friend from Anthropic, did a really good job explaining the reasons why U.S. East One matters, Jason, because why does one AWS region going down impact so much of the internet? Well, as it turns out, there's a lot of concentration

there because a lot of other AWS services are routed through U.S. East 1. And so if it breaks, many other AWS regions can have problems. So even though it seems like AWS is very distributed because it has different regions around the world, it's not as decentralized as you might think. And so the question then becomes, why hasn't Amazon managed to fully detangle its regions from one another to prevent one area having a problem impacting the rest of the world? But it doesn't seem to be the case. And this has happened in 17, 20, and. and 21 as well, according to producer Claude. So I keep in mind for people building.

So it happens really, if you were to average that out, if there's been four of these in what looks like an eight-year period, maybe my once every six months. Yeah, I was based on my once every six months across all clouds. So it seems like they're having a major network outage every two years at Amazon. That to me seems acceptable. I'll be totally honest. Really? Okay.

Yeah, only because we're building this infrastructure out at such a, significant pace and utilization is going up and we're still trying to keep up with to our first story the AI bubble all of this infrastructure being built out if it happens every two years I think it's kind of in the bucket of acceptable if you look at energy in California I would lose my energy in the peninsula gosh and you know San Mateo Palo Alto you'd lose your electricity five times a year six times a So I'm more worried about the electrical grid than I am about the data centers. The data centers, you know, they've done really well at distributing. And when something like this happens, it's typically some router.

A human makes a stupid setting. They have to roll it back. But because AWS is so big, you know, that's funny. So here's your meme. There's a famous image of a series of blocks that are stacked up in a very precarious fashion and held up by one single block. And often this is with an arrow pointing out of Jason saying, you know, random open source repo that's maintained by one person in like, you know, Sweden.

In this case, the whole internet is being held up by the block entitled US East One. And without that block, everything came a tumbling down is what it is. It's not a big deal. It's a good reminder for people that they can go do other things in life. Yeah. Another reminder for people. Like we do, of course, you can go to clod.a.ai slash twist and there's a discount there for you. We use quad. I think it's 50% off or three months.

What is it? That's pretty darn good. But we live and die by our AI usage internally. So thanks to our friends over at Anthropic. It is 50% off. Twist listeners get 50% off their first three months of Claude Pro. Just go to clod.a.i slash twist to get started. All right, Lon, you were in the group chat.

There was a really large series of protests. Yes. Are they protests? No kings. They estimate around 7 million people came out over the weekend. in various U.S. cities. We had one. It was not that large.

But we had one here in Austin, downtown. Big ones in Minneapolis. I think Providence, actually, I saw some tweets from Providence, Alex, that the largest public gathering in maybe the history of Providence, Rhode Island, over the weekend. Yes.

Well, I'll just say that. I don't think Trump's going to win Rhode Island in 2008. Wow. Every call you miss is a missed opportunity, whether you're hiring somebody or you've got to do customer support or you've got inbound leads so stop losing money and sign up for quo you know quo they used to be called open phone and they've got the number one business phone system in the world allowing your team

to collaborate on incoming calls and text in one unified inbox quo is making the most out of AI tools now every single call triggers an automated summary of course with recommended follow-up steps and their AI agent will even answer the phone when you're not around routing important messages to right person and letting you know when an important new lead has reached out. So see why over 90,000 businesses are already using Quo, including mine. We're not just going to start you off with a free trial, but we're also going to get you 20% off your first six months. And if you've already got a phone number, no problem. Quo will port it over for free. Just go to quo.com slash twist. That's QUO.com slash twist. Oh my gosh, what a great domain name. QUO.com.com.

com slash twist. This goes to, you know, what I said on the All-In podcast two episodes ago that got a little bit of, well, I'd say I've probably got more feedback on that episode since Trump was on or J.D. Vance or maybe the Larry Summers, Ezra Klein debate club, you know, all those episodes got a lot of feedback in my DMs, you know, people calling me, texting me. And, you know, I just gave a basic overview here. and politics a little outside of our remit here. But the 2.0 agenda, you know,

majority of Americans agree with those items. They shut the border, lower taxes, stop wars, people like that kind of stuff. A lot of the moderates who shifted over to support Trump and went to the right, they were part of that Trump 2.0. But the Trump 1.0 agenda, which is, you know, I'll just generally say chaos, lawfare, and maybe a little bit of violence and federalism,

where the federal government is taking on too much, that tweaks people. And I think that's what we saw this weekend. So it's perfectly timed with, you know, happening two weeks after me sort of saying, get back to Trump 2.0, or are you going to lose the midterms? And that seems like it's happening.

But there's another angle here, Lon. Tell us what that is. Yeah, the technology angle. Well, I was on blue sky over the weekend, as I so often am, checking in on what. everyone was saying. And I saw an image that kept being shared and was setting off a lot of discussion and debate. Let's take a look at it. So this is, I feel, a very obviously AI-generated image that's kind of playing off of all of the costumes and the funny shenanigans we were seeing

from all of these protests. If you look closely, you don't even have to look that closely. There are a number of tells. The flagpole is sort of weirdly bent. The, if you look at the flag, the bottom of the flag does not actually line up with the pole. That unicorn is sort of standing in an odd position that no human would stand. The frog has three eyes. Not all of the signs in the background makes sense or are legible. Even that big peace and joy banner precariously swinging between two buildings. There's no strings or anything holding it to either of those buildings.

So to me, at a glance... Lon, is this a Gemini logo right here in the corner that people might have, you know, Noticed. Oh, my God, it is. That is a VO Gemini. Obviously, this was an AI-generated image, and yet it was being shared very widely on blue sky by a lot of people who did not seem to notice that it was AI. In fact, Kara Swisher, friend of the pod, she shared this on Blue Sky.

Karen Swisher, former friend of Jakehouse. Without any indication that this was an AI-generated image. In fact, I shared some of these in our group check. at some of these people were reacting like, well, apparently that image is AI generated, so I took it down. And it's like, to me, this is really worrisome. Like, if you can't look at this image and immediately figure out this is animated,

we are cooked. I really feel like that's a very bad sign. We got, we're one shoted by. Yeah, people are just no longer, they're no longer being discriminating. It really made me think, you know what Sora is doing, where anytime you output a SORA video, it has that watermark. And here there is a Gemini watermark, but it's not doing any good.

I feel like these companies need to be more serious about making sure that any output that you create. The word Gemini should be on it. It should say Gemini and it can have it in the same. Yes, because SORA is, it says S-O-R-A and has the logo, I believe. Right. And now I have seen there is another app that people are already installing on that will remove the SORA logo. So there's always going to be workarounds.

I mean, that's just the nature of the internet. There's always a game of wackable. But 95% of people using these apps would just leave the water market place. And it would go a long way towards helping the Seth Abramson's and Kara Swishers of the world who want to share every image that appeals to them that comes across their feed to be like, hang on a minute, this is fake. Just make sure you know before you post it that it is in fact fake. People are so not smart at times.

So I saw this image on X and it immediately tripped all the hairs in the back of my neck because it appeared overly, centrally focused as if the entire straw was staged, which AI loves to do because it's making something for you, not showing you reality. And then I zoomed it on one of the signs and went, yep, AI and moved on. Lon, that's why I was so surprised to see it today in our chats because apparently everyone else got it wrong. So Jason, in a world in which open source AI models from China are able to do nearly everything that the Gemini is and the Saurus can do, will adding those watermarks actually help the problem here? Or is there not really a solution in a world where open source AI is free

range and powerful? You know, there's a Michael Crichton novel. I believe it was Disclosure, which was made into a major motion picture. And the premise of it was even back then. And producer Claude will tell us what year that book came out and give us a little summary of it, specifically around videos. And the premise of this excellent book, and Michael Crichton was just such a great author because he would make these real page turners based upon some new technology

and how they might in a procedural, you know, thriller, a procedural slash thriller. He kind of worked in both those genres. He nailed it because they were like, oh, my God, this video is, I don't want to spoil it for people, but I think it's been out long enough that I can, Lon.

I don't know what the rules are. I think I can spoil it. So basically, it actually also predates the Me Too movement because the premise is Demi Moore is harassing Michael Douglas

and they have video of people coming in and out of people's offices and it's been manipulated. And it was the entire, premise of the film was that you could see a video and it was not true. Now, this is 1994. That's 30 years ago, 31 years ago. He saw this coming. That video could be edited and it could change reality and people could go to jail, lose their jobs based on it. You know, and the obvious place this is going is, you know, you could start riots based on these videos. If somebody were to release a video of Rodney King or a Rodney,

King-esque video. That video came out. Los Angeles burned. And so we need to be really thoughtful about this. Probably where we're going to get to is nobody will trust social media. And nor should they, nor should they ever have. And so that's where this will come from. Social media, citizen journalism, journalism by organizations that don't have insurance and more than, you know, 100 or 200 employees and something to lose, they're going to become more valuable. The institutions of journalism, if they can let go of their bias and stop looking at themselves as the resistance or, you know, the retribution group on both sides, and they can actually say, and they could literally make their living, just saying, this is fake, here's the source of it.

In other words, Snopes, the function of Snopes, which was to find Jimmy, you know, the kid from the peanut butter commercial didn't die from pop rocks or whatever it is, the Wheaties commercial kid. Like, it was trying to get rid of urban legends, you know, that existed for a decade. Now we're going to need the equivalent of Snopes or community notes, which I think Instagram and threads have now are also working on. We're just going to need real time fact checking sources. Yeah. And people are going to need to stop retweeting things that, fit their narrative. That's what's occurring here. People are so tribal that the first thing they do is retweet things that fit their narrative. And so I've stopped retweeting the stuff. Or you'll see in a lot of my recent tweets the last year, I'll say, if this is true. I've seen that. If this is true, here's my opinion. And I literally did it for those robots. Remember I showed the

robots of the port? There was a Chinese port video of like robots going around. And I said, And if I said, if this isn't AI, and this is in fact China, we should really go this. And then somebody immediately said, oh, yeah, that's a port in the United States. And I was like, that was the port of L.A. I don't think it was. I don't know. I don't know. Well, welcome to hell, everybody.

This is where we are. I still don't know. And this is the sloppy issue. This is the same thing that happened to me when we were having debates about the southern border. And they were like, people are streaming across the southern border. And then people were sharing videos of people coming across the border. And it was like, well, wait, that was a border crossing under Clinton or under Obama. And I saw people correcting and saying, no, no, here, the community notes was pointing.

That's from this story. So people were just finding videos. And the problem is there are incentives, like getting followers and getting revenue share on social media, getting followers, and then putting ads on. And it's not just X. It's all the platforms. They would reward you for publishing this stuff. It doesn't matter if it's correct or not. It just matters if it trends.

So this is a really difficult spot run. Now, it turns out a friend of mine, Elon Musk went to the border. And I was texting with him. And he's like, it's true. People are streaming across the border. It's nuts. And I was like, and he showed me pictures before he publicly tweeted them. And I was like, yep, people are streaming across the border.

And that was before the border crossing data had come out. And we were looking at it and like, this doesn't make any sense. Like, look, it's been the same. the same, the same across multiple presidencies, multiple parties. And then it's the same and same for Biden. Then all of a sudden it's spiked up. And I was like, I don't know if this is true or not. And I gave this disclaimer.

Turned out in that case. It was true. People were going across the water. There are some good projects that are happening, Jason, to help establish ground truth. One of them is, are you familiar with USAFax.org? I have seen that. Is it partisan in some way or not? No, it's not partisan.

Indeed, it's run by, it's founded by Steve Balmer, the former Microsoft CEO and multi-cagillionaire. Also, it's a great place to go to learn stuff because if you click on government spending, we talk about that on the show quite a lot. They put together this awesome chart. And they really do a lot of just really fantastic information on what's going on. So here's it, immigration channels into the U.S., authorized and not. And I do think that this type of thing is going to become incredibly important.

The problem is this is entirely predicated on government data. So it still has to lean on some other authority to have it standing. So as long as government stats remain uncooked, then we're good. We'll still have some ground truth. But when it comes to, you know, no offense to boomers, but boomers and AI slop, I think the ship has sailed. So maybe we're just going to have a post-truth senior voting block lawn. And for everyone else, we'll just have to look at the charts. Yeah, I mean, I think it's sort of, it's sort of done.

I feel like to me, there isn't the will. Like, people don't care if the thing they're posting is true or not. they just want to post it. They want to be right. They want to win the argument. They want to share this thing that they're ideologically aligned with. I think we've reached a moment in our culture

where the idea of being fact-based is sort of over. And people are like, I just want my side to sound good. I don't want to be, I don't care about whether it's real or not. Pull up this producer, Claude of Disclosure. I just want to read the synapses here. Sure. Disclosure published in 1994 is a, a techno thriller set at a fictional computer hardware manufacturing company in Seattle.

The protagonist, Tom Sanders, is the head of advanced products manufacturing at Digitom. He expects to be promoted to run the Advanced Products Division after the company's merger, but the position instead goes to his ex-girlfriend Meredith Johnson, who has recently relocated from the company's headquarters in California. When Meredith calls Tom into her office to ostensibly to discuss CD-ROM drives, discuss a CD-ROM drive, She makes an aggressive sexual advance towards him. When Tom spurns her advances, Meredith angrily vows to make him pay.

The plot then concerns Tom's struggle to prove that he was actually sexually harassed by his female employer, and she falsely accuses him of harassment. Instead, the novel incorporates elements of computer industry sabotage, corporate mergers, stock options, and technology to help Tom clear his reputation and save his career. The book explores themes of power dynamics, workplace abuse, in a high technology of corporate setting. It was based on a true story of a male employee being, actually housed by a female executive with Crichton reportedly basing it on a 1980 case.

I don't know if you guys remember this, but it was also about VR. Like in the movie, especially, they really play this up, that the CD ROM they're developing is a virtual reality thing that allows you to like go into your computer and look through the files and organize things like as if you were there in like Tron. So it was, it was, it was always trying to be a little ahead. Like it was really trying to be cutting edge. But I mean, just that description is like 20 and 30 years later, all that stuff is, yeah, big part of our dialogue and pop culture and in the news. There was also something going on with AI. Speaking of AI, somebody made a trader that's beating everybody.

The holy grail of AI is, of course, to build an AI that can beat the market. And that was the same thing with machine learning or with quants and algorithms. Hit me with that one because that's fascinating. Yeah, I found this. Now, it's only been running for a couple of days, Jason. But what the company in of one has built is a really fun game that pits a lot of the leading AI models against one another. So it gave each AI model $10,000. They can only invest in crypto perpetuals on the hyperliquid exchange.

And their objective that they were given is to quote, maximize risk-adjusted returns. So who do we have here? GPT-5. We have Claude Sonnet 4.5, Gemini 2.5 Pro, GROC 4, DeepSeek, Chat 3.1, and then Quinn 3 max. And then there's just Bitcoin's performance. Now, we only have a couple of days of data here, Jason, but you can see that Google and OpenAI and Quinn are losing. And GROC, Deepseek, and Sonnet 4.5 are in the black, and they're all beating Bitcoin. But the idea here, according to N of 1, is that AlphaGo, I believe, was the model that Google trained a deep mine to beat Go.

and Chess, their idea here is do we need to create something net new to beat the markets or can out-of-the-box algorithms today do better than, you know, just the overall market? So I'm watching this because this is the coolest thing I've ever seen. This is so much better than Claude plays Pokemon. I'm not going to lie. Like, this is a nerds dream. What's so great about this is there is a real world, there's real world stakes here, and it plays off the competition between the.

the LLMs. And I'm assuming they're all been given the same instruction set. So this is a really interesting approach they've taken. What would be very interesting is if people could enter their own horse into this race with $10,000 and make this available to anyone. So the three of us could come up with an instruction set because the prompt and the data in the context window is obviously got some major part to this. So how cool would that be if we could say,

you know what, we have our own instance of GROC or DeepSeat 3, or we're going to use the one that's available, you know, whatever cloud service. But we have our own proprietary instruction set we're giving it. And then you have to reveal your instruction set when the final, you know, time period ends. So if you, and that would actually be an interesting one, you could, you could make it a race to see who can return, who can double their money first. I like that.

You know, so a time-bounded challenge. But that does incentivize risk-taking, Jason. Sure. You're going all in. Yeah. Oh, I see. So you want to know which AI not only is the smartest, but also has, what do we say now? We don't say Cajon is anymore.

You can say, yeah, sure. That's, okay. Oh, right. I'm literally the twist. PC police. There's no more. Polis. I think, you know, that's

yes. All right. Well, keep an eye. If you want to take a look at that, the website is, I think it's the inof-one.com for the background, and then you can take a look

at the challenge yourself life. Fantastic. All right, Lonnie, Donnie, you're up. All right. We've got to talk about this. So there is a sequel. Aaron Sorkin is currently working

on a follow-up to the film, the social network. Of course, I was directed by David Fincher. Sorkin is directing the sequel. himself. It's called The Social Reckoning. It's coming out next year. Set 17 years after the social network, this one's going to be about the Facebook files. But Jeremy Strong, a succession bet was amazing in The Apprentices, right home last year. He's going to be starring as Mark Zuckerberg

aged up from Jesse Eisenberg's classic performance in the social network. He did an interview on a red carpet this weekend. I believe it's for the Bruce Springste movie that Jeremy Allen White, who's also in this one was in. I'm pretty sure. Don't quote me on that. It might have been an academy event. But Jeremy Strong was on a red carpet. He was asked about playing Mark Zuckerberg in a film.

Okay, let me hear it. And here is the video of his response. Now, Mark Zuckerberg. Tell me, how do you even begin? Begin to play Mark Zucker. You know, I'm approaching it the way I approach everything else. Empathy, with objectivity, with care. What made you want to do the role?

I'm utterly fascinated by the material. It's one of the great scripts I've ever read. I've made two films with Aaron. I mean, I'd like to put my feet to the fire. Will you reach out tomorrow? No comment. No comment. Are you on Facebook?

You know what? I'm not on any social media. You know, no secret account. So for those of you don't get the joke here, he's doing Zuck right in that moment. Yeah, he's got some red carpet. Yeah, that was the academy. There was an academy, motion picture academy benefit over the weekend. You could tell by the carpet behind him, it's got Oscars on it.

So this was definitely- I think that's Variety's video. This is a shout-out to Variety. Yeah, that's a variety on the red carpet at this Academy fundraiser over the weekend. But yes, we pulled a few videos of Mark Zuckerberg speaking recently, and you could see that clearly Jeremy Strong has been studying the videotape and trying to master Zuckerberg speech patterns. I think the one of Zuckerberg on Theo Vaughn,

if we could pull that one up, is the best example I could find. Here's the real Mark Zuckerberg answering the question. About being so funny on. It's hilarious. All right, here we got, Theo Vaughn. You know, it's an interesting question. I just think that there are a bunch of factors here

that you need to peel apart. I think someone can be. I think socially perceptive and understand kind of what is going on in social dynamics and what is have a lot of empathy and care about other people while still being quite awkward and how they communicate and oh my god so that's what you guys to do was to really wanted you guys to do these like five seconds back and forth so when we cut this into the actual show, just go back and forth five seconds. Right.

Well, let's play the other one now. This is Mark from his Horizon World's recent presentation at the 2025 developer conference. Here's another clip for Mark Zegger, maybe we can try to go back and forth. It has also been really neat to see how many people are using Quest to watch video content. You know, it's just a lot more immersive. So we think of this story. The pauses and the so. Yes.

There was that one line where it goes, we think people are changing the way they watch a video content. And that's what Strong is doing. That like weird pause in the middle of the sentence thing. Like I'm going to start my thought. And then I'm going to finish. But you're not sure if I'm really done. And if I'm at the end of my sentence.

I'm going to, you know, Yeah. Think about the end of the sentence. Mark Zuckerberg is clearly a man who is not accustomed to being interrupted. Because if you leave gaps that large in your speech, normally people jump in. Yes. But if you're worth several hundred billion dollars, apparently, you can. Just do it like this.

You know, I was thinking about... So some people might... If we're reading from the docket, you know, I think, look. You know what? SNL sketch where they were doing the delicious dish, the NPR ladies, and they were just like, all of everybody was like, so in my opinion, the needs you need for a great barbecue are meat, bread, and sauce.

You know, like, you're just like very thoughtful, very long pauses between everything. With the sew and an, um, go back to Jeremy Strong. Give me to Jeremy Strong one more time. After this, I'm going to run my Zuckerberg defense strategy here. Go ahead. Give me the Jeremy.

Let's play Jeremy. Let me do Jeremy one more time. Let's skip to after the question. All right. We're not. You know, I'm approaching it the way I approach everything else. Everything else. With empathy, with objectivity, with faith, care.

What made you want to do the role? I'm utterly fascinated by the material. And I don't think this is how Jeremy Strong normally talks. No. He doesn't. Also, who knows Jeremy Strong was three big tall? He does not talk like that. He's trolling the interview right now.

He is in full character. This is why the guy, the other two or three stars from Cessation couldn't stand this guy. Because he wouldn't break character. Yeah. And I can imagine when you're trying to read the notes or talk about the shot or, you know, trying to change the lighting. If the guy won't break out of Zuckerberg. Yeah. Brian Cox was going to choke this guy.

Brian Cox, the other star of Succession who played the dad, he famously had an interview where he was like, Jeremy's a brilliant actor, but he's impossible to work with. And it's very frustrating because, and like there are, I've seen Jeremy Allen White, who also is in this movie coming up, the social network movie. He's the guy from the bear. He's Carmi from the bear. He's playing Springsteen in that new one. And he said there are some actors, he talks about this on Hot Ones this week.

week. There are some actors who on set, they could just be gabbing with you between shots, well, how the kids do and what you do this weekend. And then as soon as the camera goes on, boom, they're in character, they're ready to go. And there are other actors that they need to like warm up and they need to be in that mindset all day to make it work. And I guess it is hard for those two kinds of actors to work together. But Jeremy Strong is definitely one of those. Like he's with a boy all day on the set. Don't talk to him unless you want to talk to Kendall. Who is Jeremy Allen White playing in the social network? I got to know who he's playing.

He's playing Peter Thiel. Here's the capsule review, the summary we have. A young Facebook engineer named Francis Hogan, played by Oscar winner, Mikey Madison, from Anora, teams up with Wall Street Journal reporter Jeff Horwitz, Jeremy Allen White, on a risky mission to bring attention to Facebook's biggest secrets. So they're playing. Mikey Madison is the whistleblower. Investigative journalists.

Who leaks the Facebook files to a reporter. Jeremy Allen White is the reporter. And then Jeremy Strong is Zuckerberg. So they're going to have some like heat level De Niro and Pacino scenes together. I'm predicting. Correct. Where he's interviewing him. Love it.

I cannot wait for this movie. I haven't watched the social network since I saw it. I don't know if it holds up or not. Oh, it holds up. It's really good. I think, yeah, his, Sorkin's dialogue is a bit, yeah. Purple? No, it's like, it's almost a bit too dramatic for my taste.

He's a playwright, and he comes from that world, and it's very theatrical. Like, a few good men was Aaron Sorkin's big breakthrough, and it's like, all of his stuff has that, like, you want the truth, you can't handle the truth. All the truth. I've never been in a conversation that sounds, that's literally, and I've been involved in a lot of amazing conversations in this life. And none of them have ever reached Sorkin levels of dialogue. Yeah, it's, it's, it's, it is the adjunct. But I mean, the West Wing has that quality, too.

I was going to say, is it very realistic or is it what we wish Washington was like? And I feel like that, VEP is the realistic version of Washington. West Wing is the aspirational one. If you. want to be a mensch, you want to be a super munch. You know, everybody always says to me, Alex, Jake out, God, provide so much value for startups, you know, programs that you do,

found a university, you're always available for a meeting or, you know, the podcast itself to get people inspired. Is there anything I can do for you? And generally, no, unless you can get me a couple of extra ski days or Nick's court side tickets, I'm good, and I can get those things myself. But what I can't do is I can't go write a review on Apple Podcasts or post really thoughtful comments on Spotify. But you can do that.

You can send us a picture of it and you can email it to Reviews at This WeekIn Startups.com. Email us at Reviews at This Weekend Startups.com. And we'll give you a shout at the end of the show. Thanks so much and we'll see you all next time. Bye bye.