This Week in Startups - Twilio Bets Small, DeepSeek Scares & the GameOn Fraud | E2076

Episode Date: January 24, 2025

This Week in Startups is brought to you by… Fitbod. Get 25% off your Fitbod subscription or try out the app for FREE when you sign up: fitbod.me/TWIST Squarespace. TWiST listeners: use code TWIST to save 10% off your first purchase of a website or domain: https://www.Squarespace.com/TWIST Horatio. Visit https://www.hirehoratio.com/twist and get $2,000 off your initial set up. Today’s show: Jason and Alex cover Twilio’s team size compression, DeepSeek’s scary good AI for only $6M and the GameOn fraud saga that sounds too crazy to be true. Timestamps: (0:00) Alex and Jason kick off the show (6:27) Rundown of potential topics (7:10) Twilio's team size strategy and AI's impact on workforce (9:40) Fitbod. Get 25% off your Fitbod subscription or try out the app for FREE when you sign up: fitbod.me/TWIST (14:52) Entrepreneurship in the AI era and automation concerns (19:30) Squarespace. TWiST listeners: use code TWIST to save 10% off your first purchase of a website or domain: https://www.Squarespace.com/TWIST (22:49) Meta's investment in data centers and surplus compute power (26:36) Technology’s potential to solve humanity’s problems (30:12) Horatio. Visit https://www.hirehoratio.com/twist and get $2,000 off your initial set up. (32:01) Empathy for workers displaced by automation (34:11) DeepSeq's role in reducing AI industry costs (37:27) Game On startup fraud case and due diligence in startups (53:13) Venture capital's role in startup oversight (55:36) OpenAI's Operator announcement and AI autonomy risks Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.com Check out the TWIST500: https://www.twist500.com Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp Follow Alex: X: https://x.com/alex LinkedIn: ⁠https://www.linkedin.com/in/alexwilhelm Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis Thank you to our partners: (9:40) Fitbod. Get 25% off your Fitbod subscription or try out the app for FREE when you sign up: fitbod.me/TWIST (19:30) Squarespace. TWiST listeners: use code TWIST to save 10% off your first purchase of a website or domain: https://www.Squarespace.com/TWIST (30:12) Horatio. Visit https://www.hirehoratio.com/twist and get $2,000 off your initial set up. Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Check out Jason’s suite of newsletters: https://substack.com/@calacanis Follow TWiST: Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups Substack: https://twistartups.substack.com Subscribe to the Founder University Podcast: https://www.youtube.com/@founderuniversity1916

So meta has a plan to build a data center that's so large. It actually comes with its own map. So I don't know if you saw it's on threads, but here is Mark Zuckerberg showing a superimposition, his future two gigawatt data center on top of the Manhattan map. And Jason, just for folks out there, how much of Manhattan do you think this is covering? 35, 40%.

Yes, yeah. I mean, it's all of Central Park. It's, yeah, it's giant. It's going from 120,000. 5th street down to 34th maybe? I mean, this is gigantic. This weekend startups is brought to you by FitBod. I'm tired of doing the same workouts at the gym. FitBod will build you personalized workouts that help you progress with every set.

Get 25% off your subscription or try out the app for free when you sign up now at fitbod.me slash twist. Squarespace. Turn your idea into a new website. Go to Squarespace.com slash twist for a free trial. When you're ready to launch, use offer code twist to save 10% off your first. purchase of a website or domain. And Horatio. Customer experience can make or break your business.

Horatio provides top-tier outsourcing solutions to support and delight your customers so you can focus on growth. Visit hire Horatio.com slash twist and get $2,000 off your initial setup. Hey, everybody. Welcome back to this week in startups. My name is Alex and I am one of our two co-host. I'm Alex over on X and I'm joined today, as always by the other This Week in Startups co-host. It's Jason Calacanis.

Jayal has been under the weather, but he is out and he is back. And he is ready to rock. Jason, how are you? I'm back in the game. I'm back in the game. You know, sometimes you push yourself a little too hard, Alex. And at my age, sometimes the world pushes back. And you're just like, whoa, you can get knocked on your ass.

So I went to the inauguration, which for a never-trumper, people know, you know, it wasn't always the biggest fan of Donald Trump. But, you know, I always root for whoever wins for the presidency, just as an American, I believe should. You know, I went to this inauguration, and it was quite nice. You know, the Republicans, they have a very wide tent. They don't like criticism all that much. You know, these all political people are a little sensitive about criticism. So, you know, having J-Cal around is a little bit sensitive.

Like, oh, my God, am I going to say something critical of the president? Oh, I called the hot swap of Biden, you know. I mean, I call balls and shrugs. Anyway, long story short, I go to all these parties. I have a great time. but you know I had just gotten back from Japan which was a last minute trip that I wasn't planning on doing after I had been in you know Tahoe for the holiday break and so this is like three trips back to back to back you know Austin Tahoe Naseko Hokkaido back to Austin to D.C. to back to Austin I get to like you know the day after the inauguration there's all these balls going on. candle like this, whatever, starlight gala, that.

I'm invited to everything. The invites are flowing, famous people in my DMs, whatever. You know, selfies, you saw it on socials. I just wake up in the morning. My voice is gone. This is, I guess, on Tuesday of this week. I mean, it's even a blur. And I just pass out.

I'm in my hotel room 24 hours straight. Oh, wow. Yeah, sweats, everything, chills, body aches. felt like I get hit by a truck. And it's like 19 below zero in D.C. is the most miserable place you've ever been. DC's not my favorite city. Let's leave it at that.

I mean, I did like a four, I did four days there. That's enough for four years, okay? To my, it's a little bit like Vegas. Single digit days, get out. Yeah, 48 hours, I'm out. Absolutely.

I think that's exactly right. I can do 48 hours there. Don't need a hotel room. I get back. My wife says, you know what? I'm sending, Because we just moved here, I don't have like a primary care doctor yet.

Need to get a recommendation. I had, you know, the whole thing dialed in when I was in the Bay Area. I had the, you know, like a concierge type doctor. So if you get sick, you know, the doctor comes to your house. It's quite nice, right? I don't have any of that. So head to here. I go to urgent care.

Shout out to urgent care. Just mad apreci for urgent care. Mad apreche. It's so efficient in Texas. urgent care. There's like 17 of them between me and my house. I go to urgent care. Urgent care has the testing thing right on site. So there's no guessing for influenza strep. I love it. COVID, everything. So they're sticking things in my nostrils down my throat. And yep, within 15 minutes, they're like, you have influenza. Then they give me, there's a new

flu medicine. I get the new flu medicine. and it worked like a charm. It's one pill. I'm going to tell you the name of this. I just want to give a mad apreech to Zofluza, X-O-F-L-U-Z-A. You know, usually take the Tamiflu, and that's like a bunch of pills. I don't know if you ever taking that. But I get this Zofluza, and, you know, another day of the hot sweats and the whatever shakes

and my throat's terrible. And then, you know, I got, yesterday I was 80%. Today, I feel like 85, and I'm back in the game. I wake up this morning. I'm about to get on air. I have my coffee. I go turn the shower on. Hot water's now. So that's where I'm wearing my Knicks cap. Sorry, folks. You don't get to see the mullet today. It's a bit of a mess. But I am long story. I'm back in the game. Glad to be home. I've missed you a little bit because, you know, you've been traveling. So we've been doing some, you know, different types of shows, different setups. I love consistency, though. And so I'm just like, I'm going to be healthy on Monday. You're going to be healthy on Monday.

Let's keep down. Once you get back from SF, we'll be back to normal. So yeah. Because there's a couple of things going on in the world. It's not like we had a slow winter, you know? No. I mean, the tech news keeps popping. And then it seems like a lot of what's happening in this new administration is around

tech and business and, you know, tariffs and taxes and crypto. I mean, it's just going to be nonstop. And I guess the pace at which they're going feels to be, you know, they're kind of shot out of a canon, as they say in the business. business. But what's on the docket today? Let's get, let's get to work. So we had a, we had a number of fun things. One, there is a new case of fraud involving a startup name Game On that I might want to touch on because it's one of those fraud cases when you're like, what did you think was going to happen? We also have notes on static team size from Twilio and how I think that theme might

actually need to be tweaked to be a little bit more extreme. We have notes on Trump's EOs on AI and crypto. We have notes on building the next supercompetal. computer, a lot of big hardware announcements, not just from Open AI, not just from XIA, but also meta's in the mix. And then finally, what about R1 and Deepseek, the Chinese company that has taken the AI benchmark by store? That's pretty interesting as well. So, Jay, some, dealers choice. What do you want? You know, I really like this Twilio static team size plus plus. Let's say, what's that about? Because I know return to office mandates are starting to kick in this year. A lot of people said, hey, 2025 is the year. We get back in.

in the office five days a week, and there's a lot of hand-wringing about that. Sure. But this one sounds even a little more extreme. So, yeah, so Twilio, for those who are a little bit out of the loop, was a well-known startup, went public back in 2016. It's the company that was famous for helping developers plug into SMS networks and just telephony in general. They provided the bridge, essentially, to get you off the internet onto phones if you needed that. And it became a really big company, Jason.

Jeff Lawson. It's been on the program many times. I was just about to say. Jeff Lawson was the guy. And then in the post-COVID world, growth slowed. Jeff Lawson's strategy kind of fell a foul of investors. And eventually he was, I'm going to say, from what I can see, defenistrated, just kicked out, ousted. And they brought in some different folks who we've been tracking this company.

But I was going through their investor pitch because they dropped some notes last night. And their stock was up 20% today last time that I checked. Actually, now 21%. So I'm going through the documents, Jason, just read and trying to figure out what are people so excited about here? And then it turns out the company has put together a plan that says the following thing. Okay. So, yeah, usually when you see a 20% jump like that, it's either because of earnings or getting fit in a riff, right?

That would be the two possibilities or maybe both. So what the company said was it dropped this number right here. This is their non-gap operating margin, target for between 2025 and 2025. Essentially, they're shooting for more than 20% in a non-gap operating margin, which means for those who don't track this, essentially they have 20% more revenue than they do operating costs. Sorry, that took me way longer than it should have. This is better than investors expected. People are very excited. The thing is, Twilio is doing this with a workforce that they say has declined by 40% since Q3, 2022. So they are not keeping

their employee size static. They've cut 40% in the last couple of years. And I don't think they're about to start increasing their cost basis because this entire pitch is about profitability, cash flow, and operating margins. So to me, Jason, static team size, this thing we talked about all last year, might just be too polite. It might be smaller team size than static team size. That's my pitch. All right, folks, we're a couple of weeks into 2025, and I have a question for you. How are your New Year's resolutions coming along? For most of us, getting healthy and exercising more, well, they're at the top of the list, right? Right up there with sleep. And you know, these things are all correlated. That's why January is the perfect time to build habits that stick. And I've been using FitBod to keep track of my workouts. It is awesome. It's like having your own personal trainer right in your pocket and on your wrist because the Apple Watch version is so tight. It designs custom workouts for you tailored to whatever your goals are. But it also knows your fitness level and it knows what equipment you have available. Plus, it knows what workouts you've already

done. So if you did leg day, now you're going to go to chest and you're going to do core and it will keep leveling up your fitness using machine learning. See, this is the magic of FitBot. They combine machine learning with exercise science to create a workout plan just for you and it tracks things like your muscle fatigue and your recovery so you don't overwork or underwork any muscle group. And they will keep your workout fresh with over 1,400 videos to guide you through every move. And that's what I need help with. I want to get the technique right. Because from what I understand, if you don't do the technique right, you're not getting but half the benefit. No, gym, no problem. You tell FIPAA what you have available, and it's going to build a workout instantly. Even if I don't have anything, if I'm just in a hotel room,

it's going to give me core exercises, floor exercises, all that good stuff to do. And the Apple Watch integration is amazing because I don't have to take my phone out between each set. I beep, beep, And I just log everything right on my watch. Easy, peasy, lemon squeezy, kick off 2025 strong and take the guest work out of working out. Just open the FitBod app and start making progress. Get 25% off your FitBod subscription or just try it free right now. You can none to lose. When you sign up at FitBod.

Me slash twist, that's fitbod.m.m.m. I guess this is directly related to AI. And this is something that, you know, gets lost in a lot of the. I guess celebrations about AI. We saw this Stargate announcement, you know, high-fiving with President Trump and Larry Ellison, Sam Altman, and Masayoshi-San. And listen, it is great investment. Hey, we're going to have a hundred thousand jobs created to put in this $500 billion supercomputer. Okay, great. What I did see in the replies to a lot of those comments, the replies were all, okay, 100,000 jobs are being created here, how many are going to be destroyed by AI? Now, that sounds like a bit

cynical, right? Of course. But there is some truth to it because, as we've seen, you know, we do it here. Okay, we want to research a topic. How much faster are you at researching a topic now than you were three years ago? Or prior to a chat chippy-tie? Oh, I mean, for me, probably 20%, just thinking not a lot. A starting point, eight good things to look into, then I do my own digging, but it does provide some early directions, yeah. Okay. I would say for somebody who's in their first couple of years of work,

it probably makes them 200% faster, right? Absolutely. You know, between these two numbers is the truth, right? So for an elite person, it makes them 20% faster. Great. For somebody who's a neophyte, it makes them 200% faster. Okay. And that's in year like three of this. where will we be in year four, five, and six? And I think that's what you're seeing in this Twilio number.

Now, all big tech was fat and we used to always hire for where we would be in two years. So that was always the plan. But if you want to make your stock price go up, there's a pretty quick way to do it, is to just keep expanding earnings. And then you'll expand earnings by getting rid of middle management. getting rid of the DEI department, getting rid of the PR department, getting outsourcing HR, instead of having eight lawyers, having six, because the six remaining lawyers are, you know, 50% faster, or 20% faster. You yourself said you're 20% faster. So if you were at TechCrunch right now or Mattermark back in the day,

and there were five of you and you each got 20% faster, that means you either got a sixth employee or you can get rid of the weakest of the five members of that team. This is moving at a quicker pace than I thought. So static team size does need to get updated to a new name, which is, you know, slipping, sliding, diminishing, cutting, falling. Well, we'll find the right word for it. Yeah, compressing, consolidating, something in that range. But there is some compression going on here, like you're making a zip file where, like, I think we're just compressing talent.

And you're going to be able to do more with less. It used to be called right sizing. But this is really more, it's almost like making elite units. And so if you're not elite, if you're the bottom third of employees, man, you got to really get your skill set up. And if you're in the middle of the pack,

I think you do need to wonder, you know, if you're next kind of thing. I think you could draw an extrapolation in your mind of how fast AI models are improving, how fast we're seeing, agentic AI go from kind of like hype meme to think people are playing with to, I presume, in six months thing we all use and really begin to worry about the bottom third feels right for 2025, bottom half for 2006. I do worry a little bit about folks who are struggling to pay

student loans and mortgages and so forth. But I really think that Klarna gave us an early indicator of what was going on. For those who don't remember in middle of December of last year, Klarna said, and I'm quoting here from Insider, Klarna's CEO says the company stopped hiring a year ago because AI, quote, can already do all of the jobs. And they've also seen their team size come down as the profitability has gone up. Business people love to talk about operating leverage, Jason, which is essentially just keeping your costs flat while having more revenue, more gross profit, more operating income, etc. I mean, if AI lets everyone have operating leverage, that's going to create a lot of shareholder value. Huzzah.

But I do worry about the overall economy sometimes. people are going to have to do is say, hey, what else can I create in the world? So it's going to force people's hand to be entrepreneurial. If you were part of the group of product managers or middle managers at a company like Google or Facebook, or you were in the DEI department or the PR department or that fifth lawyer, you know, at one of those companies that, you know, did this consolidation or eliminated roles and said, hey, we're just going to be focused on profitability for some time here. Well, look at your previous company or companies, look at what they did and say, hey, what are they not doing that we always wanted to do there? What was a business opportunity

they didn't pursue because the core business was so great? If you were at Google or Facebook, there were some products or services they worked on that you can now say, you know, hey, wait a second, if they can do more with less, maybe three of us could go do something more with less and we could make a new product or service. And so, more entrepreneurship is what's going to be necessary, I think, to become sustainable. And then you have this work from home standoff going on concurrently. A lot of those folks who are being exited also probably want to work from home because, hey, that's dope.

You and I are at home right now. I am too. Go quit. Create a product or service. Use these same AI tools that replaced you at Cloner or replaced you at Twilio and make a competing product or service, or maybe it's like a little nook and cranny that, you know, Facebook or Google or Twilio says, you know what, it's not worth focusing on that. It's just too small of an opportunity, but it might be a $10 million a year opportunity or a $5 million a year opportunity, which for four

people is a pretty great life. Yes. Lifestyle business is only a pejorative if you're trying to achieve a venture level outcome. Otherwise, lifestyle business means awesome lifestyle from your business, which is, you know, no sin whatsoever. Who's the guy who makes Overcast, the great podcast player Marco. He does a podcast as well. Marco Arment and I forgot the name of his podcast, but it's very good and I catch it once in a while and he does overcast and I was just listening to his pod and he was talking about whatever beach town he lives in. I won't docks him, but I don't know if he's public or not about it because he didn't say, but he decides he's going to buy the local restaurant and his podcast mates are like, no, you're going to screw up the podcast

we need you on the pod. And he's like, yeah, you know, I just want to do stuff in the real world and I make enough money and I'm good. Yeah. This all goes back to the countertrend. So, you know, you could be a victim or you could look at one door closing and two more opening. And that's going to be the opportunity of AI. So I agree a lot with that, Jason. Before we move on, I just want to say that for everyone who is technologically savvy, this is

going to be an awesome era because you can build apps faster. AI models are getting so much better so quickly. We're seeing increased connectivity with Starlink and Project Kipers launching and, you know, et cetera. So I see the opportunity point for people who are ready and able to go after it. But what about folks who like drive semis? Because you and I both think that in 10 years, that's going to be entirely automated. And there's, there are going to be folks that don't upscale into the stuff that you and I understand.

I don't think we've an answer yet for them. And I think we probably should think about that as a society. All right, founders, let's kick off the new year by talking about your website. If you're launching something new in 2025 or you need to refresh your brand, and I bet you do, you need to use Squarespace. Squarespace makes building a stunning, professional website ridiculously easy, whether you're selling products, offering services, or just sharing your ideas. Squarespace gives you the tools you need to make it happen.

And you know Squarespace is at the cutting edge with AI tools that make building your site easier and faster than ever. For example, it's design intelligence tool. It's just like having a world-class designer at your fingertips. You ask it a few questions and their AI takes it from there, building a fully customizable website that's perfectly tailored to your brand in just minutes. We're talking professional layouts, on-brand visuals, and premium content all ready to go, day one. So here's your call to action.

Start the year strong. Check out Squarespace.com slash twist for a free trial. And when you're ready to launch, go to Squarespace.com slash twist to get 10% off your first website or domain purchase. That's Squarespace.com slash twist. I think there'll be an opportunity for some number of them in, let's say, trucking. Trucking is going to be like a two-stage process. They're going to keep the driver in this automation scenario to be a supervisor, and they'll probably have two or three of these

things, daisy chain together, tractors, and they'll just run them for 24 hours. So you'll be on a 24-hour run, or you'll be on, like, let's say, three, eight-hour runs with a two-hour charging break in between or a refill break in between or one hour. So the travel time to do this, like, let's say, 20-hour run is going to be like seven hours an hour break, seven hours an hour break, whatever, something like that. So you'll still have the person in the cabin. They can just sleep. So they're going to be like a supervisor. They'll be able to take over, like, let's say inclement weather happens. And the AI, they don't want. to the AI driving the car in sleet snow or there is a traffic jam or there's an accident or something.

So I think it's going to move to that kind of mode. But I wouldn't take that job in, I wouldn't go into that career now. Just like I wouldn't start the career of being a coal miner now. Absolutely. Not a great idea. Absolutely. But so here's the thing that I think about, I think we take a look at old models, old examples of times of economic and technological change and how that impacted work. But I think we've been taking people and consistently say, all right, let's automate the physical stuff and move you up the value stack towards information work.

And if we do begin to change that broad progression by making information work less valuable and more automated, I do wonder if it's going to be harder to balance employment in the society in the near term. In 50 years, I bet you were still at 4% unemployment. But in the meantime, people have mortgages. So I- Generation Tool Belt, right? That's our other trend to keep an eye on. So we'll see a lot of people going into trades and artisanal stuff or arts. And people might enjoy it more.

Like I have to say, like the whole concept of like going and doing TPS reports and mind-numbing labor in front of a screen creates its own set of downsides compared to work in the land and enjoying being in the fresh air. I know that that's not wrong. And speaking of people who are out there in the fresh air building things, It seems that every company in the world has a new push, Jason, for even bigger data centers. We have a graphic prepared for this. We're going to pull up right now. So, Meta has a plan to build a data center that's so large. It actually comes with its own map.

So I don't know if you saw it's on threads, but here is Mark Zuckerberg showing a superimposition. Wow. His future two gigawatt data center on top of the Manhattan map. And Jason, just for folks out there, how much of Manhattan do you think this is covering? 35, 40%, 40%. It's, yeah. I mean,

it's all of Central Park. It's, yeah, it's giant. It's going from a hundred and 25th street down to 34th maybe. I mean, this is gigantic. That's why I asked the New Yorker

how big this is. I was hoping to get streets. So thank you for that. Yeah. A couple of crazy things about this. It's not just the fact that it's very large that is incredibly important.

It's also what Zuck is promising. So he is saying that this year, in 2025, Meta-AI is going to be an assistant that serves a billion people. That by itself is a pretty big claim. He claims that Lama for their next open-source AI model will lead. And then he says we're also going to be doing a lot more of our own coding development with AI. So we're going to build this two gigawatt data center, one gigawatt this year, 1.3 million GPUs.

And here's the headline number, J-KL, $60 to $65 billion in investment in Kappex this year. So if you're thinking that Project Stargain is shooting high, so is meta. And then over in India, Reliance Group, a company we don't talk about because it's not in our remit. It's talking about building

a three gigawatt data center in two years. And then also Anthropic, of course, is working with Amazon on all of their computing needs. Holy crap.

This is kind of crazy. It's more money than even I expect it. There is a cash surplus sitting around. And we've talked about it for years. Like, why does Apple have

the largest, like private equity fund in the world sitting in their bank account. What do they do with that? Like Microsoft, just all this money piling up. And we're talking, there's going to be more and more companies like this where they're going to just keep having the earnings go up, cash sitting there. What do they do with the cash, right? If they don't spend it, they pay tax on it, right?

Okay. And when you remember all this repatriation of cash coming in, Apple had to deal with it, coming from Ireland. And it was like all a bit of a negotiation with the government, like, okay, well, we could leave it overseas or we can bring it here, you know, and, well, now they finally figured out something to do with it and something productive to do with it, which is, if you build these data centers, maybe you solve problems that humanity appreciates. The question becomes, well, what if they solve all the problems?

Like, what do you do with all this compute? And that's where it's going to get super interesting, I think, because the model sizes might get smaller. the things that humans need to get done might get resolved and in their day-to-day basis there's this concept of like surplus and you think about your average

Mac Mini M4 or your laptop of the potential compute power it has how much are you using? Well I'm using Chrome right now so the answer is 110%. Okay so you've opened up using Chrome 2%. Exactly so you've got like 75 windows open and, you know, they've got this amount of it. But the truth is, you know,

of all the desktop computers out there and the phones out there, we're using a very small percentage of what they're capable of. The surplus is extraordinary. I wonder, and why is that? Well, because there's nothing else to do on your computer. You could fire up 20 programs, but you don't have a problem to solve and you want to go take your kids to gymnastics or get a burger or go ride a horse. So humanity has a certain number of problems. to solve and right now we're backed up on these language models. I do wonder if we solve cancer and we solve energy, we solve X, Y, and Z, like what's left to solve? And that's going to be a very interesting place to be. It feels like we're building so much compute here. It'll be a dark

fiber moment where we have so much fiber and what are we sending over it. Yeah. You can only consume so many movies, right? We filled that up with higher quality videos. But When we think about the demand question and your point of like what happens when we solve things, we're at least not there yet or even close to it. This is a quote from the Wall Street Journal talking to the CEO of Anthropic. And he said, I'll just quote him, the surge in demand we've seen over the last year and particularly the last three months has overwhelmed our ability to provide the needed compute. And then he says later on they're going to have a million ships powering the Claude family by

2006. All that's today, as we think about these massive compute projects, they are predicated on, I think an anthropic level of increasing compute demand. And Jason, I think frankly, this brings us to DeepSeek and it's R1 model. And the fact that they managed to train the underlying V3, Deep Seek V3 model for something like $6 million, which scared the crap out of a lot of folks. Because if you can build something that's that good for $6 million, you don't need to spend a billion dollars. And therefore, you probably don't need a hundred thousand, five. 500,000, 5 million Nvidia chips.

And I'm curious to see if the deep seek breakthroughs that they've shown, which Mark and Dresen called a gift to humanity because they are of open source, if that is a apparition from the trend or if it is the new trend. And I don't think anyone knows yet the answer to that. There's going to be a refocus on quality very soon because it does feel like we're getting to the point where it's answering questions, there's enough capacity to answer, you know, a large number of questions

or provide a lot of services. Let's just say for customer support or legal. Then it's going to be like, hey, can we slow down for a second and just get, make sure the answers are correct? Oh, I see. You see what I'm saying? Like, yeah.

Yeah, get back to five-nines. So if you were doing storage and let's say you were building S3 in your Amazon, okay, yeah, How much can we store? Okay, important question. Okay, great. What's the throughput of that search?

Okay, great. Another important question. Okay, we can store a whole bunch. We can store a billion photos. Okay, great. And how fast can we get them off people's phones? Okay, we can do that in a day. And what's the resiliency of that?

And how often do they break down? And, you know, how often do we lose photos? You know, and then you start like, okay, we need to build a raid array now. I'm just excited that we can be seen. sitting here in 2025 and have an industry that was trying to figure out like, I don't know, how to like split bills and how to do photo images and filters. You know, you and I were covering this as journalists for a little bit and building it in my case and investing in it.

It just feels like we're going to solve some pretty awesome problems. In today's market, your business is just. judged by its customer service, right? We all know that. One bad experience, that's it. Caput. Customers gone for life. And you know what? They may start saying negative things about you online, and they have the power to do that. So stop thinking about customer experience as just a department. It's actually your entire business, right? That makes sense. And Horatio is the trusted outsourcing partner for hundreds of startups and fast-growing companies looking to deliver exceptional customer experiences. Horatio provides dedicated teams who act as an extension of your brand, delivering

world-class support 24 hours a day, seven days a week. Whether it's chat, whether it's email, or even on the phone, Horatio will make sure your customers are happy, loyal, and coming back for more, right? You want those customers you've already got. Those are the best ones to continue to buy your product or service. And when you have Horatio on your team, that's going to give you the ability to focus on running your business and scaling it. Why fast-growing brands like Built Rewards, Bark, and Dr. Squatch rely on Horatio to provide exceptional customer service. So here's your call to action.

Horatio isn't just outsourcing. It's seamless integration with your team. And the best part, their solutions are cost-effective and completely tailored to your business needs. Visit hire heratio.com slash twist to learn more and get $2,000 off your initial setup. That's higherheratio.com slash twist. It's awesome to me if we can just blow everything up and start over and solve these problems. I hope we have empathy, deep empathy for displacement of humans.

And I'm going to go off on a little emotional tangent here. We have to remember that when we make this stuff, it has an impact in the real world. right it does and we all benefit from it we all make incredible amounts of wealth tons of money just flows to whoever hits the the gusher right and listen i've been there when a gusher gets hit and i had my bucket out and how hey whoa you know we hit oil you know it's like uh there will be blood right i was this what i was thinking up that's a great movie by the way if you haven't seen i mean let's watch it i mean such a great film um let's all watch it. And you know, you drink everybody's milkshake.

I drink your milkshake. Yeah. But let's get, let's just remember that when that person is crossing the Rio Grande, right, and they're coming here to this country to have a better life. Or you automate all those truck drivers or the cashiers, you know, give weight to toast. And we all, you know, the ordering service. And hey, you know, all the venture capitalists and the CEOs and everybody gets nice and rich, you know, and oh, wow, this is great. We solved a big problem. Just remember your fellow human beings who help you build the last business, who are you convinced to come join you on that adventure to build the last one that just got this place.

They need a place in the world. Yeah. And let's have some thoughtfulness about that. I'm glad you said that. I'm very optimistic about us as a species figuring out some good path forward if technology improves at the pace we're thinking. But at the same time, there's an interesting kind of irony

to this DeepSeek story because a lot of VCs have put tons of capital to work, backing, anthropic and open AI and owning shares in meta that are spending billions of dollars. And then Jason, this comes out, and this is a set of data points

from artificial analysis. which I use a lot to track models. And if you look at where Deepseek R1, which is their Open AI 01 equivalent ranks, it's darn close to exactly as good as Open AI's 01 thinking model. It's slower, but critically, look at how cheap it is compared to 01. It's like 8% of the cost or something like that. It's insanely affordable.

And that to me just says that, look, all the money that, you know, Coastler or whoever put into Open AI better get to work because apparently $6 million in some nerds in China are out there changing everything. But here's the best part. For startups that are not open AI or Anthropical or one of the giants, this is great news. Do you want cool tools that cost less and are open source? Here you go. Have fun. So I'm going to restate my empathy. We need to have empathy for the venture capitalists and for the CEO who have deployed hundreds of billions of dollars only to be wiped out by open source. These poor venture capitalists and their LPs, they made the wrong bet.

What's funny is the LPs. Can we have some grace for them too? The LPs are us, man. The LPs are like my mom's pension fund and like, you know. Here we go. Listen, it's all going to work out. As Alex always says, buy an index fund. Some are going to win.

Some are going to lose. We could sit here and tell you all day long that we think Microsoft's going to run away with it or XAI or Open AI. Nobody knows what's going to happen, right? It's so true. It's so true. And, you know, if we were sitting here in the 80s, looking at the PC era, which is it going to be Eagle computer? Is it going to be compact? Is it going to be

HP? You know, it's going to be Gateway. Is it going to be Dell? Oh, gateway. Sure. I mean, there was, there were 50,000 people making PCs in their in their dorm rooms and somehow Michael Dell figured out how to make it into an actual real business. You know, and then in the web era, oh, which search engine is going to make? Is it going to be Lyco? It's going to be Magellian. Is it going to be Info C? Info C. I don't know. Maybe this new one, Google. Yahoo. We'll see. Anyway, this is the moment we're in. And so it's awesome to have competition. And you know what?

It's not just competition from the United States. It's going to be all countries. All of this comes back to energy now. When I look at it, I'm just focused on two things. Energy and the quality. The quality of the results, right? Like the resiliency of the data. And is it actually just like a dumb teenager that regurgitates?

And I think a lot of cheating is going on on these tasks. you know how when they benchmark them. I think the benchmarks are all being fed. I call me a cynic. But I think these teams are incented, like the engineers, to game the benchmarks. I know this because, like, it's the same group of people that would, like,

send their kids to, uh, get coaching on, um, IQ tests, which is like the most unethical thing you can do. Also, isn't that self-defeating? That's like just putting risers in your shoes before you get measured. Like,

it doesn't actually make you more. No, but it gets you into the right school, which then gets you, the right tutors and teachers and then gives you the confidence and then maybe it's self-fulfilling prophecy. So what do they call it when you, what is it, red jerseying or red, red shirting? Red shirting, you know, like you keep your kid back a year and put them into hockey and then they can you know, check the other kids on their asses. It's kind of like that kind of thing, you know,

well, I got my kid into a better school with the better teacher. So now, now we'll see who's smarter. On the, what people are building and how fast, if the venture capitalist will need to bail out. I'll just say that this post on Blind has been going around Reddit and a lot of other places. And essentially the gist here, if you're listening to the audio version, is that this

blind report, which we cannot verify by definition, because it's on Blind. But it does come from a- Explain what Blind is for the folks who don't know. Blind is a quantized social network for employees of companies. You register with your corporate email address to ensure that you are actually

at the company in question. and that provides a reasonable layer of BS detection in the system. That said, I am trying to caveat this by saying that everyone agrees this is true and real and actually what's going on, but we don't have nine sources on this. The point is, the headline reads, MetaGN AIOrig in panic mode. It started with DeepSeek V3, which came out last year, which rendered the Lama4 already behind in benchmarks, adding insult to injury. They did it with a small budget.

And essentially, meta is trying to dissect what this company has done. And this is the same company that just pledged $60 to $65 billion in hardware investments this year and has really been leading the charge for quasi open source AI domestically. And I would say internationally to boot. So if they're worried, okay, because $6 million, Jason, is nothing. Like to meta, that's literally nothing because they have tens and hundreds of billions and trillions in market cap. So like it's, it's an impressive accomplishment. And I think a lot of people are going to be, a lot of people have spent a lot of money. They're not going to get back.

sometimes you lose money on a business venture and you learn a whole bunch and you take those learnings to the next one. So, you know, I can, all of these like startups where you invest 125k, 500K and it blows out and then the founder comes back to you for the next one, it blows out and then the third one, you know, winds up being, you know, like the big winner. So I always like to bet on the second or third time the founder's up at bat, right? Because, okay, here we go bring all those learnings with you to this game and boom now we're going to the championship we're going to the finals so i think that's what's going to happen but you had mentioned there was a fraud i'm always here for a good fraud okay all right so i'm going to have to to back up a little bit

so this all comes from uh essentially alleged fraud maybe i put a legend just a pre i'm going to preempt the uh the fraud here alleged this is an alleged fraud a massive brazen alleged fraud evolving a company called Game On that was based in San Francisco. And it had raised a, I think about 20 million before the fraud begins. And the fraud kicks off per the indictment in September of 2018 through 2024. And Jason, this is why I wanted to have it on the show. It's funny because it's catastrophically sad. But also, the company went on to raise another $60 million while committing fraud.

So this is actually a bad actor that is doing quite a lot of damage to VCs, back to bailing them out. So they built an AI chat business. Essentially, if you're a league or a sports team, you could sign up with them and they'd provide a chat bot for your fans. And this was a business that people were excited about. You know, people back in 2016, 2018 chatbots had a moment.

AI, of course, has had several different peaks. But then came all the lying and the fraud. So let's move on to the next slide here and talk about what they did. The alleged lying in fraud. Yep. The alleged lying and fraud. Thank you. See, lies allegedly,

included Jason, revenue that never existed, and cash balances that were vastly inflated. So the company also lied at the beginning by saying the founder, Alex Beckman, said that he sold a company for $100 million, give or take. It was actually two. That even by Silicon Valley standards is a bit of a unfortunate stress point. But then they went into all out absolute fraud pretty quickly. So they did things like stealing the money. They used corporate funds to pay for their children's private school tuition. They bought a $4.2 million house in San Francisco with corporate funds.

They paid their personal credit card bills, please. I mean, yeah, keep going. Oh, it gets funnier. They spent hundreds of thousands of dollars on jewelry and luxury retail. They bought Teslas. They, I believe, used corporate funds to pay for their wedding to, allegedly. if I recall the full indictment. And the lies were really insane.

So the amount of money they claimed that they made as a business had absolutely no bearing in reality. In 2022, to pick one example, one year, they also lied about, allegedly lied about 2023. Game On said that it had nearly $70 million in revenue in 2022, which, Jason, for a startup, means that you're a year or two away from an IPO, if you want. It's a huge number. and they had $54 million in gross profit that year. In reality, their revenue never cracked a million dollars in any year. Okay.

According to the indictment. So was there a board of directors is the first question. Was there an audit committee? Would be the next question. Who was the CFO? Who was the outsourced accounting firm? And who was the lead investor in these rounds? This all goes to the lead investor should have these controls.

in place. What we do with our very fledgling young startups is we share with them the diligence you'll get. Like, here's a diligence list for Series A. And we'll run people through it for their seed round with us. And we'll just say, you know, this may seem silly because you're going to have a lot of n slash A's, not applicables in this list when we go through this checklist. You know, you go through this whole thing and this is what founder. will learn over time is, you know, these controls are in place for a reason. It's really to protect the founder and to protect the shareholders, to protect the employees so that, you know, you don't have a bad outcome.

Yes. The only people will ever object to this, it's like literally like clockwork. When somebody is like, what do we have to do this, you know, all of a sudden, red flag comes out, ooh, and Molly and I had like a joke. We, I have to get your red flags. I don't know if you remember. But we used to just hold, she would hold up a red flag or I would hold up a red flag.

Yeah. Because this was happening so often during Pigserv that we would just laugh at each other. It just, you know. No, it's, it's, it's,

I've seen actually, uh, clips of that I remember for episode 2000, we went back and watched a lot. So I, I've seen lots of historical twist. Yes. And the red flag,

when somebody, I'll just abstract. Please. One or two stories for you. You know, I had a founder who didn't want us to have our board seat after agreeing to it. We own, you know, low double digits percentage of the company. Let's just say 10% for argument's sake. And, you know, they don't want to start board meetings. They don't want us on the

board. You know, we're having like all this pushback, but we've already made the investment. And, you know, we, we start to like, okay, yeah, no, no, but companies like trending towards a million in revenue, you really need to have a board in place because we need to go from one to five. And to go from zero to one, like I've seen founders do it like just sheer force of will. You know, like a great founder just barrels through some brick walls. And they can get to their million.

They get to a million, but to get to a million five, kind of need a plan. And you can't just, you need a team, right? Superman can do certain things on his own. Sometimes he needs the bat. Sometimes he needs Green Lantern and Wonder Woman.

Whatever, there's just too many things. Even the guy under the water. Aquaman, you know, can come in handy once in a while. I believe the collection of this is called the Justice League. Yeah, you need the Justice League. You need the X-Men, whatever it is. Everybody's got their purpose.

Each mutant has a special power. This is going to blend a couple of different stories. That's totally fine. I appreciate that. And let's just say, like, the person is, there's like a weird entity getting paid money every month. And it's like, who would have thought that there was something unto war? They didn't want to see found.

It's like, well, what's this entity? And this entity is like one of the founder's entities that's a consultant to the other entity getting paid like whatever hundreds of thousands of dollars a year and but for what and there's no contract it's not documented but there's you know whatever tens of thousands of dollars a month let's say it's 25,000 a month going to this entity and then well we don't have any insight to that in these books and it was like well this founder didn't feel they were getting paid enough and they had seeded the company and i'm like well why aren't they on the cap table if they're a seed investor and they're like well no it was a loan i'm like okay well where's the loan document and say well there is no

loan document. It's okay, where's the bank transfer? Well, there is no bank transfer. Well, what did they loan? Goodwill? Was it like, oh, yeah, well, they gave their services for free. And now you're like, whoa, my head's spinning as to what's going on here. Also, if you try to pin something down at every time you do it, move two inches away.

It's like a shell game, right, exactly. And I feel like I'm on 42nd Street playing three-card Monty. Like, I know, trust me. Yeah. I'm from Brooklyn. I'm wearing a New York Knicks happen. in a New York Knicks shirt, like, there's a briefcase over here.

Yeah, exactly. And so, I just said, like, guys, timeout. You can't hustle the hustler. I've been here before. The briefcase is on the shelf. You don't need to know what's in the briefcase. Just let's go back to first principles here.

Whatever's going on here, stop. Clean it up. because and then sure enough like there's another investor finds out about it and now we're in the clean up game and I'm like the founder goes oh man I should have just been straight with you from the start I said yeah we could we could have clean this up it's very simple you you had an entity you felt underpaid we've raised your salary you paid your taxes you stop paying for your apartment your loft your back pay if you had a back pay issue great we'll put it into your

forward pay. There's a way to do this that would be tax advantageous and fair and fair to the other co-founders. Just be straight. And that's what happens in these fraud cases is there's no controls in place. And, you know, Paul Graham, you know, I don't, I mean, just to summon Paul Graham here for a second. He always puts these VCs or like the enemy kind of like the VCs, killed the companies of VCs. And I get it. Like there's bad actors in VC, da, da, da, da, and he's the champion and only he is good for founders. and he's the sole person and everybody else is, you know, in the ecosystem is anti-founders. But Paul and YC are the only ones who are a pro founder. You know, it's kind of their schick.

But the truth is, yeah, there are some bad actors. There are also some bad actors who are founders who will do crazy shit like this. And that's why controls are in place. That's why you should start a board. That's why you should have an audit committee. That's why you should have a plan. When you get to a certain level, he's right when it's a tiny company. when you hit a million, two million, you raise three or four million, have a board with one board

member and one lawyer and do four meetings for one hour a year and learn the best practices to clean this stuff up. All of that is fair. By the way, the only actual pro founder of VC is the one that I'm usually talking to. And every other founder, every VC they know is terribly anti. Yeah. No, I mean, I joke. But these people, so it wasn't just Alex Beckman.

And it was also his then-partner-later spouse who was a lawyer who helped with a lot of this. And they did not just do, okay, some back pay, some shenanigans. They were doing, allegedly, whole cloth fraud, inventing documents, showing, inventing email addresses from people that may or may not have existed. Wow. And I just want to highlight a single story from this indicting because it goes, Jason, I mean, it is, it is. Game on. Game on indeed. Game on.

Be careful what you wish for. From page 26 of the indictment, delivery of bank statement to financial institution two. And I'm going to gently paraphrase this. June of last year, Investor 1, a partner serving on Game On's board of directors, arranged to go with the co-founder Beckman to a bank in Chinatown near their office. And the partner, quote,

wanted to see a statement directly from the bank to know Gameon's true account balance with financial institution one. So they concoct a scheme to have his, I believe, then wife, take a document to the bank that they then dropped off so that way Beckman could bring the VC to the bank, have that document given to him, and so he could give it to the VC as if it was from the bank to show fraudulent information. But this also... How would the bank

agree to do that? That's deranged. I'm trying there's a lot of... Yeah. I mean, this sounds like... Here it is. This sounds more difficult

than building a $10 million business. Thank you. Dude, the amount of the work... This is like people who have like, like, two families. And I'm like, oh, how do you have... I'm trying to be a good partner to one. You know?

Yes. Anyways. the fake account statement that Beckman and investor picked up had been planted by Lau his spouse. Earlier in the day, Lau went into the same branch and met with bank employee one. Lau sent the false and fraudulent game on account statement described to bank employee via email asking the bank employee to print it

even though she knew was false and fraudulent. Then Lau had them arrive and then they picked it up. I'm not entirely sure about the mechanics of it, But that's the scale of fraud here. So they were not just incredible inflating and tinkering. This was fraud on a level that I don't understand. Because did you think that you could steal all the money

and not build a business and not get caught? That's the thing that I don't get. What's the end game for this? Yeah, I think the criminal mind is, this is why I wanted to be an FBI agent. Because I just, I, I, I became, and I wanted to go to John Jay criminal justice. I was going to be a cop and then become an FBI agent.

My brother Josh went into the New York Police Department, as I've told the story here. And I went to Fordham and then I was going to go to John Jay and do criminal forensic psychology. I was fascinated by the criminal mind because I do think there's like just people who just commit crime. Like they're just sociopaths. And then there's people who chase it, right? They think they can unwind the fraud at a certain point. And then they realize they can't. This is like the stage where, you know, they know they're going to get caught at some point.

And they're just trying to extend that period of time. But they're not even smart enough to realize I should just sell everything and leave. And leave and go on the lamb. These guys were like, go on the lamb territory. They should have been selling the Teslas and the jewelry and the house getting cash and going to South America and disappearing. Maybe we shouldn't give, now that I said. They had allowed criminals ideas. Maybe they're just not good criminals.

I mean, at a certain point, the jig is up. The problem is they were insufficiently good at fraud. And that is why criminals get caught. But no, I want to bring up a different point, though. VCs are not fraud detectors. VCs are backers of people who are building things aggressively. And I think that there's always going to be some of this that happens. Yes.

And I just think that we're, seen now companies that during ZERP, as you mentioned earlier, eventually implode, either for normal reasons, lack of revenue, lack of growth, losses, too high, whatever, or frauds are coming due. Because one thing that you can do to check the peak of a market is when fraud picks up, you know you're at or near the peak. It's basically just consider it like credit card, one to two percent fraud. Yeah. That would, I would say one percent fraud is probably baked into the system or cheating or whatever it is. And so, You know, we, we, you can't slow down the industry to put in so much constraint.

So you have to trust people, you know, like it's, we're giving a 25 year old, a $4 million seed check. Well, what if they go do something stupid? It's like, what if? When? Yeah. You're going to do something stupid. What you're trying to do is train them on how to be stewards of capital and how to steward the ship, you know, but you can't learn how to pilot the ship without taking the controls

the ship and going out in the ocean. And you know what? Most ships don't make it to the new world. And some amount do. This is what the queen of Spain learned. She's like, you know what? I'm going to fund a lot of like adventures here to go on the spice route or to the new world. And I'm going to give them 20% carry. And the reason I'm giving them 20% of whatever comes back is because I know a lot of stuff's not coming back. I want to send them to bring something home. And if they bring, whatever you bring home, you get 20% of you know what? that's where the term carry comes from. The concept of carry came from ships. All they cared about is what came home.

What made it to port? And whatever makes it to port, the venture capitalist, the captain of the ship, they get 20% of, the LPs get the other 80% all hell the queen. Well, by that analogy,

there's a whole lot of startups, boats that are outside of port, liquidity, Then VCs wanted to get the cargo carry off. And that's the hope for this year. I do want to hit just really quickly, Jason, a couple of things that we're not going to get into entirely. One, did you see OpenAI's operator announcement? I saw the announcement.

I didn't watch the video yet. I saw Sam and had a brief conversation with him when I was in DC for the Adag oration. All right. So they've announced a thing called operator. What is it? It's their first agent. Just in case people are curious. AI defines an agent as an AI system that can do work for you independently.

This is currently rolled out to pro subscribers, so the $200 per month Open AI subscription. It is cool. I think it's cool. I don't know if it's better than Anthropics computer use, but to me, this is more of a step on a longer journey. I'm not trying to vet this as the BL end all, but Open AI is dropping more cool stuff. And so as we talk about AI, just keep in mind that not every AI system that people have built will survive, if the big model companies coming out and potentially crushing them. Developers have these, you know, on their desktops and are playing with them for some time.

They show them to me all the time. And the reason they're not out in the wild is because when you give control of your browser, you can do things like, I don't know, open up your password manager, you know, track your keystrokes. I don't know, buy stuff on Amazon. I don't know, go search for things on the internet and post things to your social media. that you should would never want posted. Go do a fishing account. Send a thousand emails to a thousand people that are personalized based on your history with

them while you're asleep. That caused chaos in your life. Like literally you could just tell chat GPT. Cause chaos in Alex's life. Open up his email box. Read all his history with his boss. And then send a spicy message that would cause the most damage. Like literally that's where we're at, folks.

that's why these things aren't in your hands yet. I'm trying to figure out what email I would send you that would make you the maddest. Jason, I'm joining the AI Safety Group at the American Communist Party in the Office of Equity. I've decided to take a DEI role at the... Careful, I probably will at some point. I've decided to lead DEI at Google. more. That job sounds dope.

Yeah, exactly. I'm going to. Twist is back on Monday. Jason's going to be in San Francisco. We are back on our live stream and cadence. So make sure that if you are on YouTube, you hit the bell and the subscribe button. All the things.

Just click all the buttons. We got so much coming this year. IPOs are going to come back. MNA is going to come back. We've lost to be a great year. It's going to be a good year. It's going to be a good year.