This Week in Startups - Anthropic wants to slow AI down and Bernie wants 50%: JCal Reacts | E2297

Episode Date: June 6, 2026

This Week In Startups is made possible by:Grasshopper Bank https://grasshopper.bank/twistVanta https://www.vanta.com/twistRender https://render.com/twistPlaud https://Plaud.ai/twistToday’s show:Anthropic wrote a blog post calling for a global AI slowdown. Meanwhile, Sen. Bernie Sanders wants the government to seize 50% of every major AI company’s stock. Find out why JCal is reconsidering universal basic (or even high!) income policies, and why he thinks the 2028 presidential election will likely come down to AI policies.PLUS a live ComfyUI demo from founder Yoland Yan. Find out why the free-to-use open-source node-based platform has become a crucial part of millions of designers’ and VFX experts’ workflows, and how their tool has been used to create everything from “The Wizard of Oz” at the Vegas Sphere to those viral Coca-Cola holiday ads.GuestYoland Yan: http://x.com/yoland_yanComfyUI: https://comfy.org/AI Models and ToolsIdeogram 4.0: https://ideogram.ai/models/4.0/Stable Diffusion: https://stability.ai/LTX Video: https://github.com/Lightricks/LTX-VideoLoRa: https://huggingface.co/docs/diffusers/training/loraGoogle Veo: https://deepmind.google/models/veo/Relevant Links:Anthropic: “When AI Builds Itself”: https://www.anthropic.com/institute/recursive-self-improvementBernie Sanders: “The Public Should Own Half of the Big AI Companies”: https://www.sanders.senate.gov/op-eds/the-public-should-own-half-of-the-big-a-i-companies/Bloomberg: “Sam Altman-Backed Group Completes Largest US Study on Basic Income”: https://www.bloomberg.com/news/articles/2024-07-22/ubi-study-backed-by-openai-s-sam-altman-bolsters-support-for-basic-incomeTimestamps:0:00 Guest 1: Yoland Yan, ComfyUI — live demo intro2:06 Plaud: If your work depends on conversations — interviews, meetings, calls — you need a Plaud NotePin. You can check it out at https://Plaud.ai/twist and use code TWIST for 10% off!4:34 Guest 1: Yoland Yan, ComfyUI — live demo intro9:47 Grasshopper Bank - Time is money. Don't waste either. Go to https://grasshopper.bank/twist and get an exclusive $500 cash bonus just for opening an account.20:05 Vanta - Get $1000 off your SOC 2 at https://www.vanta.com/twist22:24 What is Outpainting?30:01 Render - Find out why 5 million developers are already using the all-in-one cloud platform, Render. Go to https://render.com/twist and apply for the Render Startup Program to get $500-$100,000 in free credits, depending on your stage and backers.32:13 Jason's insider sales team advice38:42 LA mayoral race: Bass vs. Pratt42:25 Anthropic wants AI to slow down?48:45 Will Sen. Sanders' argument resonate with the public?59:39 Why 2028 will be the AI jobs election1:05:32 Brian Chesky's new AI lab1:15:21 Jason's "Mandalorian and Grogu" review1:18:53 YouTubers take over the box office1:24:16 Dean Potter vs. Alex HonnoldSubscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.comCheck out the TWIST500: https://www.twist500.comSubscribe to This Week in Startups on Apple: https://rb.gy/v19fcpFollow Lon:X: https://x.com/lonsFollow Alex:X: https://x.com/alexLinkedIn: ⁠https://www.linkedin.com/in/alexwilhelmFollow Jason:X: https://twitter.com/JasonLinkedIn: https://www.linkedin.com/in/jasoncalacanisCheck out all our partner offers: https://partners.launch.co/Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarlandCheck out Jason’s suite of newsletters: https://substack.com/@calacanisFollow TWiST:Twitter: https://twitter.com/TWiStartupsYouTube: https://www.youtube.com/thisweekinInstagram: https://www.instagram.com/thisweekinstartupsTikTok: https://www.tiktok.com/@thisweekinstartupsSubstack: https://twistartups.substack.com

Anthropic. They want to slow down global AI development. What? Wait a second. Why are they doing this now? Didn't we just all go through this that they withheld mythos, acting responsible? Now they're calling for a global pause? Shut down claw then if it's too dangerous. If not, what are you doing? You're the ones behind all of this development. Bernie Sanders released a video. This is a completely insane, deranged concept. He wants to seize 50.

percent of the shares of the AI companies. 2028, the number one issue in the election, it's not going to be inflation. It's not going to be wars. It's going to be AI. This weekend startups is brought to you by Render. Find out why 5 million developers are already using the all-in-one cloud platform, render. Go to Render.com slash twist and apply for the Render startup program to get $500 to $100,000 in free credits depending on your stage and backers.

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 Grasshopper Bank. Time is money. Don't waste either. Go to grasshopper.

com slash twist and get an exclusive $500 cash bonus just for opening an account. All right, everybody, welcome back to Twist. It is June 5th Friday. 2026 with me again, DeLan Harris. How are you doing, brother? Hey, I'm all right. I'm doing okay.

I looked a little banged up this week because I got a new dog and she clawed my face, but I'm feeling all right. We'll talk about in an off-duty. We'll also talk about the Knickerbockers. The New York Knickerbockers, yeah. And my experience at game one

here in San Antonio. But let's get started. There's a lot in the news. I hate to, you know, too much small talk at the beginning. We put that at the back, like a mullet. Business in the front party in the back. I like it.

Let's get started, Lonnie Donnie. I think before we go any further, I think we should applaud our friends at Plod and point out we're both wearing our plod pins right here. I've got mine. You've got yours. A million, a million uses for these plod pins. That's what you discover new ones all the time while you're wearing them. Like I just, I adopted a dog named Barry this week. And the whole time that I was talking to her previous foster mom about, you know, what she likes, everything.

I recorded the whole thing with my plot and then on the way driving home with her. I jotted down some more notes. I captured some more thoughts. So now I had a whole Barry adoption action plan written up by Plaud just because I was recording it. I didn't even think about that. You go to see your doctor. You see a vet. You're doing something important.

Trusting human memory, not a good idea. Trusting human judgment? Great idea. And this is augmenting human intelligence. I found a new use this week for my applaud. I had it on my wrist. Today I'm wearing it on my shirt here. I had my plot pin.

You can also get the one that goes in the back of your phone. I was at the liquidity conference, and I just left it on for all the programming. Wow, yeah. Now, after doing four hours of content, I have the entire live straight feed in my phone available to me. What they did at the conference was a very unique thing. It turns out plot.

It's such a good experience at this week in startups that they went and did a partnership with All In. The liquidity conference was our LP conference this week. they recorded the conference. Then every night, they sent out the same link that you and I would use when we're sharing plod conversations that have been recorded, mind mapped, summarized action items, all that stuff. They sent it out to the 550, 650 people there.

Yeah, that's amazing. So it was a group plot. Once again, we applaud, plot. I met the founder. He's awesome. Millions of users using this. It is a tremendous life-changing product. Give them the code.

So if your work realized on, conversations, we urge you to get a Plaud Notepin S. You can check it out at Plaud.a.ai slash twist. And if you use the code twist, you're going to get 10% off your new Plaud PIN. 10% off an already unbelievable, reasonable, reasonable prior. Really great deal. All right, let's get started. Let's get started.

We got our first guest I want to bring up. We're doing an interview and a demo. It's Yoling Yen, the co-founder and CEO of Great Tool that I use very often. Come for you. It's amazing to be here. Yeah, great to have you. Are you going to show us a little demo here to kick us? off. What I usually tell people what comfy is is the polar opposite of what a, you know,

a chatypte or a mid-jorney prompt is. Over there, you have this black box of typing out the prompt, generating an image. Maybe it's not, it's what you want. Maybe it's not what you want, but you change one word and everything changes. Comfey, on the other hand, give you a node-based interface. It's very complex. It's something that people actually take a decent amount of time to learn and now becoming a professional skill required in many studios such as Netflix. And what it gives you is the maximum amount of control that you can change every single parameter. And what I'm showing you over here for folks who are

listening is this type of individual nodes that connects to each other and with all of the parameters exposed, such as the noise, with height, and then including the prompt itself, which... So we're seeing like a TikTok story here of a gentleman, you know, skateboarding. And obviously the dimensions are that of a TikTok story. But then you have multiple boxes connecting to each other,

and then you have all these variables, as you're pointing out, not just the dimensions, but I'm assuming each of those objects or layers, I guess, was how Photoshop did it, explained to us the concept here at a high level. Yeah, yeah. So the prompt, the text itself in here describes the entire image you are trying to generate. This is actually using a very new state-of-the-art open-source image model called Ideogram v4. Shout out to Ideogram V4. It's an open source. And is it out of China? the U.S. Where is it from?

It is out of a lab in, I believe, Canada and U.S. It's actually fairly... I-D-E-O-G-R-A-M. Ideogram. Yeah, they just dropped the new model this week. If you haven't taken a look, definitely check it out. And then the prompt itself is describing what the final image is. What this model is actually really cool in terms of doing is you can set bonding boxes

to say like, hey, I want the image to be generated exactly at this portion of the place. So over here, there's a, you know, the description is a massive 3D fluff, white typography spelling of comfy. So as you can see in the end, so there's exactly generating around this. And it looks like balloons and it's in the air.

So this is what you don't get. When you give a prompt lawn to stable diffusion or any of these other Gemini banana, Nano banana. Yeah, it's up to them to decide where everything's going to go. This is much more granular. Saying, hey, I want the logo here. I want the person here.

I want this font. And you're giving it a much more granular, I'll call them micro prompts within a mega prompt. So you have micro actions in a mega prompt. One analogy that Yolen and Comfey have used a bunch that I really like is like it's it's like a slot machine. When you're engaging directly with the model, it's like 60% of the time, 70% of the the time you're rolling the dice, maybe it's going to get what I look like. But the more granular, you can make the prompting in the early controls of the weights and the, you know, adding on

lauras and all of these other sort of add-ons with comfy, the more precise you can get, you can get what you want the first time without having to keep pulling the lever. Well, and the lever takes, it would be like a slot machine that takes 90 seconds to tell you if you watch your money. It's pretty annoying as a slot machine. That's correct. And actually, one thing that people keep on emphasizing a lot of times is that for a lot of these other models, you know, that is behind a single prompt what you usually happens, even if you were to use exactly the same prompt, you don't have the reproducibility. That's because they actually hide most of the parameters away, such as the

initial seed. So in comfy, what you can do is actually you set a fixed seed and this image and this generation would always be exactly the same. And that's huge for creatives that in the end, they are able to reproduce exactly the imagery they were expecting or other people have produced for them and reliably using that in, you know, production environment or in kind of scaling up the product. So I actually just queued up the job again, just to showcase that two of the image are exactly the same. But if you were to change the seed a tiny bit, And then in that case, scenario, you're enabling the model to, you know, reuse a new noise to start the generation from the ground up. And then in that case, scenario, you'll actually get a brand new image in here.

This is kind of a live generation demo of like how it's actually showing a pretty cool diffusion process of. Is that prompt written by human or do you give an audio prompt with a bunch of details and say, make me a prompt? When you're a founder and you're moving fast and you want to close a big deal, a traditional bank can kill your momentum. Long approval processes, clunky outdated systems, not intuitive. I mean, some of these systems look like they were made in the 1970s and have never been updated. Not all banks are designed to work for startups. We all know that. And that's exactly why Grasshopper Bank was created.

Grasshopper is a real federally charted bank, but with the kind of financial tech stack that you'd build for yourself. They're going to open your new account in just minutes, not days. That's a key pain point. An integration-friendly platform with real-time transfers and AI tools that are going to save you even more time. All without training your budget on a bunch of unnecessary fees. That's where they get you. And because you're a valued twist listener, Grasshopper will drop a $500 cash bonus into your account when you use the promo code Twist, TWIST.

Start building without the slowdown at Grasshopper. dot bank slash twist. Terms and conditions apply. Grasshopper is a federally chartered bank member FDIC. Yeah, you can do many ways. You can send, actually, there's more complex workflow in here where you can say, ask a cloud model to generate the prompt using this kind of structure.

So in the end, it will, like, you can give very simple prompts and you don't have to try right this entire thing. And then afterwards, it will give you a more structured version of this type of. This is something people miss, right? That you, we were talking about on All In, David Sachs, in month three of us doing this, like, daily report on what topics are being discussed in the news that would align with, you know, the besties interest. We had trained a model.

Here's all the besties interest. Here's what we've talked about previously. And then it went out and found new stories. Okay, great. But producer Nick had asked it, make me a prompt. And then Sachs saw the prompt. He's like, oh, my God. It's incredible because it looks so structured.

It looks so like a developer with, you know, OCD wrote it. This is like an OCD level, gorgeous code. And it was like, he's like, no, no, no, no. I described the prompt I wanted. I got back the prompt. And nobody seems to know this. Lon, if people don't understand that the number one job of AI is to write the prompt, they're using AI like it's three fucking years ago.

It's insane. Yeah. And this is, it's one of the big things through, too, talk about with different models. Like some models are great for things like prompt writing. And then when you take that and you feed it into another model, that model can perform so much better because the previous model helped you fine tune your prompt. So that's why like operating it. You know, everybody gets sort of like, oh, this is my favorite model. This is the one I use for everything. They're all good at different things. Like Claude is great at writing prompts,

but not the best image maker. Then you feed that into your nanobanas or your ideograms or whatever. And like it's, that's where I think a lot of the magic comes from that people miss who don't use these tools is like the combination of models. Yes, very important. Yoland all this that you're building here, how do you run it? How do you run the prompt? Or does it change the prompt? Like if you said change comfy there and make it orange, if we wanted to make the comfy in the sky orange for orange and blue skies for the Knickerbockers, how would you do that? And then when does it change? Like does it refresh in real time? How does it work? So you, one, this is a very interesting thing for folks who don't know about comfy.

Comfi is both open source and can run in a local environment. So for anyone who wants to just use their computer, whether it's a Mac or whether it's a Windows machine that has a decent GPU, they can completely download this for free and run in a local environment and use your own GPU. Wow. So you're the hosted version of Comfi. This is the hosted version. But I could download the open source version, pay you $0.00.

Use my MacBook M5 with 48 gigs of RAM. I'm assuming that's enough. It's enough for some of the models. I'd be surprised how big it would be the model of performance requirement is these days. Got it. So you probably need to have that. I think there's a MacBook now with 256. I don't think they've gotten to 512.

Do you know? I would actually recommend using Nvidia chips for running a lot of these models. It's much better performance. So what do you run? I would say, you know, just get as comfortable as you can buy a high-end GPU from Nvidia if you were to run it locally. But if you don't, you can use cloud. And in that case, scenario, get access to a state of the art graphic.

Yoland, I'm unsure of this, but do people buy an Nvidia GPU, like a card that would go in a PC for gamers and put it in its own enclosure and then attach it to their MacBook pros and have that as like an offloading point. I did see some enclosures when I was surfing around the internet. Yeah. I think that's definitely a possibility. But a lot of times in those case, narrow, they rather just rent or, you know, use a cloud-hosted service as opposed to just doing that.

There are actually studios out there that do use comfy in a kind of enclosed environment. And sometimes it's because of privacy and air gap requirement as well. Got it. Okay. So you could be like literally running this on the flight to Japan if you had the right laptop. Wow. And so look at the detail level here. You've got, you know, 15 different boxes and arrows going in between them, connecting them. What are the arrows that we're seeing connecting all these together? Because you're just making an output of a silly TikTok video. But this looks like you're launching a rocket ship. Yeah. Yeah, the complexity definitely can scale up. So what I just did was going, taking. one of the nodes that from the outside and then entering into what we call a subgraph,

which is essentially a component that encapsulated a lot of the functionality and actually help some of the users to abstract the complexity. So what are we seeing in here? Like you had that megaprompt and this is kind of what's going on under the hood. This is the detail level? Yeah, this is the detail level. This actually give you control over everything. here you can decide on what model you're loading, what type of weight type you're loading it into,

because there's all sorts of different mechanisms you can utilize the CUDA GPU, and then including the certain type of configuration. So for example, this is a parameter called CFG, which is on the high level, kind of giving you how much of guidance the prompt itself is, is injecting into this model so you can, you know, tune up, tune down that model itself. Now, I don't have to do all this level of granularity. I'm just basically looking at the granularity behind the prompt. And that's what's abstracted in current products like chat GPT. But when I use Proxy Computer and it's doing these crazy searches, I don't know if you saw

the interface I built on like Next Unicorns line. I just decided like I'm going to make my own Next Unicorns. database of like who could potentially be a unicorn a year or two from now. I hope we're on there. Yep. Well, we'll see. I may have to adjust the prompt here and insert you in there. But it has all, when you're running it, you and I ran it for two hours last night, it had

to collapse everything, but I could click and expand it and see what it was doing under the hood. So this is the same equivalent. Yeah, Yoland? Yeah. Okay. So show us changing comfy from white in blue skies, a white, white. logo to an orange logo.

Yeah, I just edited it out here earlier, and then now I run it again using the same configuration, and then the job will kick started. The model itself runs fairly fast. Let's see if in the end it actually does what you said. Here's a moment of truth. Demo or die here on this week in startups. Let me see it, but there are more blue. Ah, okay, there's vibrant blue.

Playing around with a tool like this, Jason, It is a great way. Like, I think I got a lot better at making AI generated images because I worked with comfy UI early on. Because you learned about, like, what does every word of my prompt do? And, like, how do these models, how are all of the different variables and factors going in to producing the final image? And, like, you know, even like the setting, he was just saying, like, how much fidelity do I

want the final image to have to exactly what I wrote in my prompt versus how much freedom do I want the model to have to like abstract a little like you'd be amazed at how different your images will turn out just by tweaking a few of those kind of weights and things it's a great way to like see well there's the orange there's long filibuster there's the orange in the background though thank you looks like we put the orange in the wrong part of the prompt um let's uh move i think we understand how the product works yeah yeah this is industrial strength um levels of prompting oh i see you have one here of me there yeah yeah Yeah, 100%.

This is definitely like if there's one takeaway for the audience, this is basically a production level of a product that enable people to give that fine grain control and quality for AI. Hence why the Coca-Cola ads, the recent like AI Super Bowl ads were all generated using comfy. I actually wanted to ask about that because you guys worked on the Wizard of Oz sphere of Vegas show as well. How is comfy actually like what part of the process is comfy being used on like, how do you integrate into the process of making something massive like that? Yeah. We are the tool builder.

We did not directly participate. I cannot pay credit over the Wizard of Oz on sphere. Accusing you of ruining the Wizard of Oz, Yol. Don't worry. Oh, I didn't even know people were saying that. The AI revolution isn't just creating new tools and new products. It's reinvented the entire landscape when it comes to compliance. Startups face an intricate patchwork of overlapping rules and regs from the EU, the United States, federal agencies, everywhere.

It's coming at us from all angles. And if it sounds like a lot for you to track while you're also trying to build your business, well, you're right. And that's why you need a reliable, trusted partner like VANTA. V-A-N-T-A. Vant-A-I-powered platform automates your entire compliance, Whether you're preparing for a SOC to running an enterprise GRC program or doing an audit, Vanta's going to worry about the security. So your team can focus on building great products. That's why some of our favorite companies, like Ramp and Ryder, and they're both doing great,

are spending 82% less time on their audits by working with Vanta. Whether you're a fast-growing startup or a global enterprise, Vanta is here to help you automate your security and compliance and earn and prove trust. So get started today at vanta.com slash twist. That's va.mta.com slash twist. It was very cool because we were able to kind of be fairly close with the studios that ended up working with the Google V-O team and generating all of the results. So a very kind of simple scenario, like is using the V-O model and the company together

for upscaling. But you cannot just take a. a, you know, a 1920s color film into full 32K video for a sphere screen. Because a lot of times where if you scale, let's say, this image into a sphere environment, what you're going to get is Jason's face being completely warped around the globe and it just looked good for the audience. So in that case, they had to outpaint. They had to create imagery that never existed before in this film. So that's a case scenario.

They utilize Comf's power in here to run all sorts of parameters to figure out, hey, what would be the best option of outpainting all of these images? And sometimes I have to be generating characters reacting to the main character talking and all of those things. And you just got to make sure that the highest quality and control is there. And hence why Comfey is utilized in that environment. All right. So what are we seeing here?

You took a screenshot of me talking or is this from a pre- previous episode? Yeah, yeah. This is from a previous episode. And I actually wanted to show, because we were talking about the outpainting process, this will run a while. I can rerun it again, but this is an input video of Jason talking. And this is a TikTok type of video itself.

But what comfy was able to do in the end using the LTX open source model is that it looks very complex. I'm not going to explain the whole thing. I actually, funny enough, like, I don't even know the whole thing. Like, so I just excuse to say I'm not going to explain it. What happens in the end is this ended up outpainting the entire video based off of the input. So it's guessing your arm, guessing all of the background, what happens over there. So this might look like the original, you know.

It isn't, but you filled it in, you're saying? Yeah. Yeah, this is a generating. Based on the TikTok image, they guessed at. what is on the outside edges of your screen? There's a couple of accounts, long, that do this on the internet for, and they make tall versions of Star Wars

or Goodfellas or Sopranos. I don't know what they call that meme, but it's to basically soprano's in vertical. Right. Yeah. Well, there's in painting and outpainting. In painting is when you put something in the image that wasn't already there,

so like we could replace your microphone with like a dog running around the table. And then outpainting would be extrapolate beyond the parameters of the original frame, like what's going on on the sides or in the background, or like Wizard of Oz at the sphere, like what would the rest of Kansas look like other than that little window of Dorothy that we see?

Yeah, and it's not perfect. Like it put me in a short sleeve shirt. Obviously, I'm not wearing a short sleeve shirt, but that's okay. It's something you could change, right? Yeah, that's a part of your problem. Exactly. You can change all of the parameters in here.

Yeah, and then all sorts of crazy things you can do. You can utilize the model, And this is actually using a Chajibiki image model to take this input video and then transform you. Oh, this was single to my. Yeah, first thing, right. Here I am. hilarious. Senor Calacanius.

Yes. So, Yolen, where is all this going in terms of as a business? You know, you are building on top of open source models, other people's frontier models. You're building these tools. You have open source. Explain to me the contours in the audience. of how you're architecting strategically your business in a world in which people are saying,

like a neophyte investor might be like, isn't this just a wrapper? What if like Open AI does this? And then explain, so there's the wrapper criticism you'll get, not that I would give that, because obviously we can see there's a level of fidelity here that we don't think Open AI or anthropically to get to an interface and tools and user base.

And then the second piece is, you know, after it. So first, being a rapper. Second, open source as a weapon or as a liability. Those are the two things I'm curious about. Take it in whichever order you want, Goland. Yeah, yeah, yeah. These are great questions.

I would say from the first perspective in terms of whether Comfie is a rapper or where is the business going fundamentally. comfy actually, there are many layers in here. company underneath the hood, the reason why we are able to run these open source model is comfy. At the core is a inference engine. It actually do the calculation. So there's a very deep technical layer in terms of how we can efficiently run these model.

And for folks who actually understand the training and the inference difference in here, there's a huge gap in terms of what inference. engine has to do to unlock these models capability and value and speed. And those are the area where we say, like, we would say it's actually fairly difficult to build an inference engine, especially the most popular diffusion model inference engine. As an open source tool is actually the top, I believe, 70 most popular GitHub projects out there by, you know, kind of the amount of stars ranked. And then another component in here is at the core,

Comfee has this open source and extensibility nature. Anyone can buy a custom code. Inside Comfee, people can bring their own code and connecting with different components. Whatever people don't actually have in the BATIM version of Confit, they can easily just use Cloud and write a version of that code and bring that into it. And that's a huge, huge string. especially in an agenic coding role, what you're seeing as people be able to easily build their own component, customize their pipeline. So at the core company, just want to feed that foundation

layer of a, what we call operating system for generative AI, visual generative AI. Have you raised a ton of money for the company? And where are you at in your journey? I mean, obviously, AI has gotten very hot. This is an incredibly verticalized tool that people who are in the arts and who are making social media clips all the way up to, you know, whatever's going on at the Wizard of Oz or not, and Coca-Cola and Super Bowl commercials. This has got a large group of people who are interested in using it. So tell us about the pricing for the product, how that's going,

consumption-based, proceed, yada, yada. And then how's the fundraising going? Pricing itself is we have the local. It's completely free. You can actually use local while using some of the close-source models, using our API platform as well to speed up your generation and that you can just pay us as you go. And then on the cloud version, you have the basic tier of $20, $35 and $100, depending on different usage, like a different kind of demand, you get different amount of usage in the end.

And then on the enterprise side, we do offer kind of more customized solution in terms of enabling enterprises to get, you know, the tier of features and security they need. And those are kind of based off of sales-led growth. And you're, you know, raised 30 million from my pal David Sacks over Kraft? Yes, yes. Not that he did it because I know he's been pretty busy. The U-I-Zar, David Sacks, Nelson. U-I's-Zar, David Sacks. Mim Zar. Yeah, I was fortunate to meet him once. So, but yeah, we raised from crafts the most recently, a 30 million round at 500 million post. Amazing. Yeah.

So I think on that end, we're pretty excited and looking for cracked engineers and, you know, operators to join our team. All right. Yohan, great job. Everybody can go to comfyuI.com. Do I have the domain correct? Where you got? It's comfy.org or comfy. Oh, love it.

And so it's comfy. And, you know, sometimes the founder has first name at the domain name once in a while. Big cloud providers may offer you cheap compute, but you'll end up paying the difference in engineering costs and hiring extra developers. You don't want to waste time configuring virtual networks. None of us do. Or your access policies. You want your team building your product. So it's time to look at render.

Render is the all-in-one cloud platform for developers that allows you to deploy scale and secure your apps and agents with zero ops. Most cloud platforms ask you to split your focus between product and infrastructure, or they force you into platform constraints. You know you're going to grow in six months, but just connect your GitHub repo to render and you are live. Web services, cronjobs, the whole stack in one platform. It's time to find out why five million developers are already using render. Five million. Go to render.com slash twist and apply for the render startup program. You'll get anywhere from $500 to $100,000 in free credits, depending on your stage and who your backers are. That's render.com slash twist.

Who do you need to hire right now? I see you got a bunch of people actively working in the set from office space. Is that the film? What was the film, Lon? I mean, I think you're thinking of office space. That's the one about engineers, and they're in Austin, and the guy gets... The red stapler? Yeah, that's office space. The mic. And you're not. And you're not. Nondescript office with the drop ceiling. Is that a video you made with comfy to? Or is that the actual office? That's the actual office.

Where are you guys based? We are right next to the bridge. You can't see it. The Bay Bridge. Yeah, the Bay Bridge. Amazing. You want to be right in the heart of it all. Yes.

We'll work for Yoland. What are you looking for? What's the top, hardest position you need to fill? I think all across the board, definitely on the most urgent need on our end is scaling up our sales team and also our front end development team. So you have a go-to sales motion

that you need to get on? Yes, we are actually only able to process above 100K inbound requests so far because of the share amount of demand. You're trying to do this in San Francisco? Bill the sales, big mistake. It's just you're up against so many lunatics right now who are going to overpay salespeople because a lot of these sales executives have families.

And a lot of them have worked at Google or Uber or Salesforce or Oracle, you know, like major comp places. And so they tend to, I'm not going to say entitled, but they have high expectations. This is my best piece of advice. Is you a first time founder or second time founder? You're on? First time to get to this stage for sure. Got it. Okay. So congratulations.

Number one. Number two. Thank you. What a lot of savvy founders do is they open a remote office in a place where other people have already done so historically and added salespeople who are grinders. What's an example of that? Somebody like HubSpot, Spotify, Yelp. And I remember the place they love to do it, Lon was a place where you can golf, you can

raise your family. It's got great weather. And there's people who are college educated, but not so educated with graduate degrees that they want to be the CEO of the company and take your job, Yoland. That place is known as Phoenix, Arizona. Oh, interesting. Or Salt Lake City, Utah. Or, you know, pick up Austin, Dallas, Houston. There's tons of salespeople, grinded out salespeople who are awesome. And instead of wanting to get paid $300,000 and one percent of sales, they would prefer to get 75K and 8 percent of sales. In other words, they'd rather bet on themselves than force you to pay a huge salary because their home costs $500,000 and it's already paid off. That's my best advice is to explore that. And the way one of my companies did this, the way

they did it. They told the story here on this week in startups. And now that we have AI, we can go into the 2,000 episodes and find it. What they did was, it's the most brilliant thing I've ever heard. They put out ads in four different cities, four salespeople. They put a range of positions. They put a range of salaries. Then they set up in the airport Marriott. And they said, hey, we're going to be in town at this day. I'm just picking like aeropartic, you know, the easy airport. So they just landed there in the morning all afternoon into the evening. They did recruiting where they, for the previous couple of weeks, said, hey, we'd love to interview.

You come anytime on this Tuesday and Wednesday. The next day, so let's say they started in Salt Lake City. Next day they went to Boulder. Next day they went to two cities in Arizona. Next day they went to Austin and Dallas. Yeah. And then they just collectively looked at the resume flow and who they fact, who they met and they said, you know what, this town, and I think it was an Arizona town, is the most

amazing fit because Yelp was paying people a 50K base with a 50K package, which 100K, and this is 15 years ago they did it. It was one of the most brilliant ideas I ever heard. Most CEOs who are not of action would be like, I'm going to hire a consultant. This one was like, F it. Two weeks of time, sent the COO, parachuted it in himself, said, we have to get this working, give me five cities, two days in each. And I want to see the salespeople. I want to get the vibes. I want to understand how hard.

And they lost their mind. Don't tell anybody this technique, Yoland. And don't spend more than... Yes. And it will spend more than 48 hours in Phoenix, not recommended. That's the max. It's max out of 48 hours.

No, Vegas was one of them, too, actually. And my friend Tony Shea, rest in peace with Zappos, he did a similar thing where he was looking for where to play Zappos. And he found all these people who were... as customer sports agents were completely thrilled with a $40,000 salary. This is in 2005. Right. He had that whole plan.

He was going to build a whole tech hub in downtown Vegas, I remember. And rest of peace, Tony Shea, but Tony Shea was a genius. And this was his big secret. He wanted people to be happy. And what he realized was people were profoundly sad in a lot of cities that had extremely high cost of living. When he found downtown Las Vegas, yeah. He found an abandoned city

next to the greatest home developments, downtown leisure, gambling, shows, restaurants, a boomtown. And in the middle of boomtown, was a just dilapidated old town with these gorgeous buildings. And he bought the old city hall and put Zappos into it. Yeah. Genius.

It was So close, this was the genus of Tony. He knew how to make people happy. He optimized for the happiness of his staff, not his own. He didn't want to live in Vegas. He was a billionaire. He could live anywhere he wanted. He could place the home and live in Kauai in Honolay Bay.

Instead, he said, where can I make 40, 50K customer support people happy and give him a 10K bonus and have them be like, whoa, this is significant, Las Vegas. All right, everybody. Yoling, great job. Yeah, really excited. I'm really excited. This is a type of tool on that makes me want to stay up all night and try it. This is the problem with this moment in time.

That was my exact experience with comfy UI when I first started using it was, yeah, you suddenly realized because you could type things directly into stable diffusion and see what you'd get. It was not a great experience, especially in 2021, 2021, 2022. But as soon as you start using this tool, you start playing around with it, oh my God, I could start to get better images. I could start to get more fidelity for what I'm looking. board. It is. It was that kind of tool that you would stay up till 4 a.m., like, playing around with just to see what you could do with it. Hey, what's the Polly Market for today? Because you and I did

a hit on Jesse Waters last night on Fox. You did. You're helping me prep and getting me up to speed. Yes. On what's going on with all these elections in my, both of our previous hometown of Los Angeles. Shout out. You want to take a look at Los Angeles mayor of the election odds on Polly Market. I just want to understand what's going on here because they had the results of the runoff primary or whatever they call that? It's the primary. Yeah, the California governor race, there's the mayoral election. This is what I want to know. I want to know if they're going to let Karen Bass burn the city down or Spencer Pratt's going to have a chance to save it. Well, I mean, things are looking not so great for Spencer Pratt right now. I'll look up the most recent Bass Ramon. This doesn't make sense to me.

I'm going to go put money on Spencer Pratt. Just because if he wins, I'm going to have a great sense of joy that an outsider one who actually cares and who had his home burnt down. I like the narrative. This is Karen Bass, who burnt the city down. What I'm seeing right now, the most updated numbers, now that's the polymarket. The actual numbers right now. And the polymarket says 66% chance of Karen Bass. Of Barras eventually becoming L.A. mayor with Nithia Ramon at 31%.

Spencer Pratt at a measly 4%. Now, the actual numbers right now, Bass has the, lead 35.1% of the vote, then Pratt, 29.4%, and Ramon at 23.4%. The thing to remember about California elections is that their laws are designed to get the most number of ballots in possible. So there's all kinds of like, you can mail your ballot and as long as it's postmarked election day, it counts. So think about like the mail might take an four or five days to get that ballot to the counting center, they still have to count it. And there's other rules, too, like if you, if there's a signature mismatch, like when they're

looking at the ballot and your signature doesn't match right, they don't just throw that ballot away. They send you a notification and you then have 10 days to resubmit. So all of these rules end up pushing back the deadline. That's why it takes so long to count votes in California and why things seem like they're so back and forth. I got to tell you. Go ahead. Long.

Tell me. Four percent chance. Spencer Pratt. I might have to put five dimes on this. I think we were talking about this. Here's like when you're talking about California, you're talking about it's so heavily weighted to Democrats. There's so many more Democrats per capita that Republicans have to do exceptionally well. Like they have to go way over the top and be nice.

But he's a Democrat. He's a Democrat. I mean, he's been appealing to, I think, you want to know what I think the other big problem for Spencer Pratt is, is that a lot of the neighborhoods where he has the most appeal, they don't actually get to vote for L.A. mayor. Like, if Santa Monica and Beverly Hills got to vote, we probably see a lot more support for Spencer Pratt. But they're not part of Los Angeles, the city. Technically, they have their own mayor so they don't get to vote. So I think that would make a big difference too. Listen, I can tell you about California. There's like a, it's the. It's the. opposite of, it's the inverse of Texas. We have like a little drop of blue called Austin or purple. I always like, and then it's like it just drip orange county is a drop of red. So it turns like orange. But everybody there is a Democrat. Texas politician Rick Perry describes Austin as it's the blueberry in Texas's tomato soup. I was like that. All right. That's disgusting. But I would say it's the blueberry in the, somebody dropped a blueberry in the in the strawberry rhubarbine.

Right. Let's lightning around some news here. In big news this week, we're going to talk about this. Anthropic in a blog post on Thursday, they said they want to slow down global AI development, an odd thing for Anthropic to say. What? Yeah, their heads of internal research and policy wrote that slowing down the pace of global AI development would end up quoting, likely be a good thing. They warn that AI models are showing the capacity to self-improve, recursive self-improvement, as they call it, without human intervention.

Here's another quote. We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. Wait. Listen, I've been traveling,

Nix games, liquidity, Napa, Yombeil, French Laundry. I've been living the life of J-Cal for the past week, which is a crazy life. Your photos from Bouchon, I'm very jealous. I'll take it. It's incredible. Anyway, I'm a bit exhausted. I saw this go by my Twitter feed.

I didn't understand exactly what was going on here. Wait a second. Why are they doing this now? Didn't we just all go through this that they withheld mythos? They're doing their thing. They're acting responsible. Now they're calling for a global pause. Do you realize that Bernie Sanders is going to make this into a campaign ad

to combine with his desire to seize 50% of the equity in Anthropos? The timing is bizarre. It seems crazy. It's also they're on the verge of like an IPO, right? Like, part of me feels like if Anthropic genuinely thinks that AI is too dangerous and we need to slow down, it's like, well, you first, buddy. Like, you're leading the like shut down Claude then if it's too dangerous. If not, what are you doing? You're the ones behind all of this development.

It just seems like you... I don't know how you square that circle. I really don't. Okay. They have the ability to slow down, and they did. And they did it correctly, apparently. You know, we interviewed Nikesh from Palo Alto networks at liquidity, and he said they were one of the early companies, obviously, cybersecurity space.

They looked at it and were able to, in their using of mythos, they were able to find a lot of internal attack vectors and close them, he said. Right. And I said, is it the real deal? And he said, absolutely, this was a level that was of performance to find bugs that we had not seen before. So they did exactly the right thing. They don't need to call for a global thing.

Now, please combine this story, Lon, with the other story. I asked you to put on the docket. I don't know if it made it because we had a couple of package shows. We had a few missed shows. Yeah. Do you remember the Bernie thing that came out? I believe on Monday that I comment. commented on. Bernie Sanders released a video. This is a completely insane, deranged concept. He wants to seize

50% of the shares of the AI companies. And he doesn't say like these three companies. It just says AI companies. So does that mean, you know, everybody, Nvidia shares, whatever. Let's play a little bit of this on 1.5, 1.5 speed, please. We got to speed up. Bernie. San does because he wants, if you put him on 1.5 speed, that's the speed that some of these entrepreneurs talk at. Hello everybody. I probably don't need to tell you that artificial intelligence be the most transformational technology in the history of the world. It will profoundly affect the lives of every man, woman, and child in our country and in fact throughout the world. It is already bringing significant changes to our economy,

our democracy, our emotional, our environment, and how we educate and raise our kids. There is also a very real fear that as AI become smarter than human beings, it could eventually function independent of human control with potentially catastrophic consequences. The question is not whether AI will change the world. It will. The question is who will own and control that future? Who will benefit from it and who will be hurt by it? Will AI be used to make life better for working families?

Will it help us eliminate poverty, extend life expectancies, and help us solve the climate crisis? Yes, it will. Or will the future of humanity be determined by a handful of billionaires who have developed AI to become even richer and more powerful than they are? That here's the fundamental choice before us. Yes. And let us be clear. AI was not created out of thin air. The foundation of AI is our collective human intelligence.

Let me repeat that. Training data. The foundation of AI is our collective human intelligence. Where do they get it from? Songs, artwork, journalism, computer code, scientific research, videos, conversations, images, images, and ideas spanning generations. As Sam Altman himself acknowledged, AI models were trained on our, quote,

collective experience, knowledge, and learnings of humanity, end quote. Yes, that's true. The reality is that big tech oligarchs have fed this knowledge into their AI models without permission, acknowledgement, or compensation. In other words, the creative work of millions and hundreds of millions of people has been stolen by the wealthiest people in the world.

The time has come to reclaim what was stolen from us. How are you doing that? Since AI is built on the collective knowledge of humanity, the wealth it generates must benefit humanity, not just Elon Musk, Jeff Bezos, Mark Zuckerberg, Larry Ellison, and other billionaires, or the venture capitalists and Wall Street firms who see AI as the next great wealth extracting machine. Interesting. And that is why in the coming weeks, I will introduce the American AI Sovereign Wealth Fund

Act. This legislation would give the public a direct ownership stake in the largest AI companies in America through a one-time 50% tax not on profits, but on stock. It would do two extremely critical things. First, it would give the American people a direct role in determining the future of this technology. Like the invest America, yes. No longer would the future of AI be dictated by a handful of big tech oligox while the rest of the world sits back and watches them do what they want. Secondly, it would guarantee that the trillions potentially created by AI are used to improve the lives of all of us, not simply to make people.

the richest people on earth, even richer. And I have to tell you, this is not. So basically, I advocates for 50%. This is one of the great arguments that I think will play with the American public. Why will this play with the American public? Everybody knows that AI companies stole people's IP to train their models.

How do we know that? There are an infinite number of lawsuits and settlements that have occurred and are still underway. Yeah, and you can see it. Like, if you ask, you know, Open AI, give me the plot of Wuthering Heights, like, it knows it. It can, it read that book and knows what's in the book and can spit it back. Or, or it read a summary of the book from the open crawl, et cetera. Right.

So the fact is, this has already been done. Now, the New York Times has a big lawsuit. Other people have had lawsuits. Settlements have occurred. So he has the high ground on this one with the American public. The American publics understands this. And if you're going to scare the bejesus out of us and say over and over again, we're taking your jobs with AI, for him to combine it with not only are they taking your jobs,

but they stole all of humanity's knowledge in order to steal your jobs in order for personal enrichment, I'm not saying I agree with this argument. What I'm saying is the public's going to agree. with this argument, and it's going to have an, it's going to get consensus from the left and the right and some moderates. Yeah. Because it's plain as day that they took everybody's content without permission. That's a true statement. Now, is it legally permissible to train your eye on somebody else's content without their permission? The courts are going to adjudicate that issue. They have started to adjudicate it. It looks like some amount of training on people's content is going to

allowed. Right. That's where the courts are signaling. So far, the precedent seems to be you're not allowed to steal copyright material to train AI. So like, you know, Anthropic found like that vast archive of like bootlegged books and then they downloaded it and used it. That's against the law because those were bootlegged and you're supposed to pay for content. But if you legally access copyrighted material, you're allowed to train your AI on it. So if you bought the book, ripped it apart, scanned it, you're allowed. that expanse you can. You're allowed. And it's the output. So far, that's what courts have determined is. If an AI output infringes on a copyright, that's actionable. But if you trained it up front on a

copyrighted material and then it reconfigures it and comes out with something original, that's allowed. That's what so far. I mean, there are certainly a lot more lawsuits working their way through the courts that could change that. But that's the precedent so far. This is, this appeals to people who feel the rich aren't paying their taxes. It appeals to people who believe intellectual property should be protected. This appeals to people who feel like jobs are going to go away in a large part in order to enrich other companies. And it's going to unite America first, America only, you know, the sort of most MAGA of the MAGA. Right. But I think it's kind of a separate movement now. I think it's going to

appeal to them. And I know it's going to appeal to the left because they are elitist who make content. There are, there's a real split going on in the political right. On the left, basically, everybody is against AI. On the right, it feels like there's this very direct split between like the pro business, the big tech side of things. Your JD Vance's, your Peter Thiel's, your David Sacks. They're very pro-AI. They're pushing this very accelerationist agenda. And then you do, you have the Steve Bannon sort of more traditional magicide, and they do seem to have a lot of doubts, both sort of ethically, morally, and, you know, the same kind of stuff that Bernie's talking about. Like, is this just going to take everybody's job and funnel money exclusively to the wealthy

and sort of kill the free market? I got a prediction. I'd love to hear it. I think anthropic and open AI are going to sit down with Bernie Sanders and say, we're willing to put 10% of our equity. into invest America account, Trump accounts, or dedicated to public service in some way, job retraining, et cetera. And that would be a non-trivial amount of money. It's trivial in a $40 trillion of debt for the country, right? If you just seized all of these companies combined,

it would equal $25 trillion. Yeah. Right. If you took every, all the top public companies, just literally seized SpaceX, seized Tesla, seized Nvidia, seized Google. communist, you know, just like taking over. It wouldn't even be 20 trillion. So you cut that. Yeah, if you nationalize it, you take 20 trillion. And of course, that would crater the stocks. So you wouldn't have 20 trillion. But anyway, it's, it's ridiculous. It's not going to, it's insane. But it's going to play. I think it's directionally correct. And I think that's what most people get like. And I think this is an opening salvo, you know, like Bernie's opening at 50% of the stock.

I think it's a negotiating position. You open at 50, the other side says 10. Well, now we're in a dialogue. But I think directionally, I think most people agree this is something Americans would like to see. And I think it has the extra benefit of Americans are so negative about AI. They're so dubious about AI. There's so much anti-AI sentiment. Well, if everybody started feeling a little bit more of an ownership stake, if we felt like if AI companies do better, I get more money or my local school.

district gets more money or my city gets more money. I think that's, that makes it a lot more compelling for people to have a, to have a little exposure to it, a little taste of their own. And here's the challenge. If you were to go through the top three AI labs, Anthropic is going to be in Bernie's camp to a certain extent. They're not going to agree. They, they, they, I think, correct me if I'm wrong here, Anthropic has done licensing deals and settlements with publishers, before, I believe, Open AI started doing some. So they have tacitly, through their actions, agreed with Bernie Sanders' position that it was trained on copyright materials without permission because they've done settlements and they've done deals and they sort after deals to get AI,

at least for the output. I mean, in fact, I got paid $2,500 for my book, Angel, I believe, by Microsoft licensed it for four years or something. So that's underway. Open AI has a nonprofit that's going to own some percentage of their trillion dollar market cap. I don't know the exact percentage that they own, but they actually have the vehicle to do this. They could just say, okay, Bernie, we have this Open AI vehicle. We're going to take X percent of our revenue and we're going to put it into job retraining, universal health care, UBI, some version of that. Sam Altman ran a UBI test, a long-forgotten UBI test.

He funded with Y Combinator in Berkeley. So you have Sam Altman believes in UBI and in fact did the largest test of it, I believe, outside of like, say, Alaska giving money to their citizens. So he did the most recent, most, strongest test run of it. Then Elon has tweeted, universal high income by a checks issued by the federal government is the best way to deal with unemployment, by AI. AI robotics will produce goods and services far in excess of the increase in the money supply, so there will not be inflation. Okay. So this is Elon Musk's belief. Elon's belief, Open AI's belief, and Anthropics belief is there'll be job displacement to some extent, right? They're all predicting it. Right. So who are you going to believe? And these tweets and

these positions are well known. Bernie Sanders is going to be able to say, Anthropic believes in job displacement is going to be massive, and we should pause it. Open AIs founder, Sam Altman, ran a UBI test. I asked for that. Please pull it up. Open AI, NYC, under Sam Waltman, ran a UBI test. He believes that we're going to need to get universal income. Elon believes that jobs are going away and they'll be optional and they'll be a universal one.

So here it is. Sam Altman puts $14 million into studying universal income and what's the date of this story? This is from May 7th, 2026, so just a month ago. Now says it's really useful, but not what we're going to need for this next phase. He did this study through open research in 2020. There we go. I've got the details. No, I got the details.

The three-year study gave 1,000 low-income adults $1,000 a month, while a control group of 2,000 people received just $50. Researchers expected to see whether steady cash would reduce motivation to work. Instead, they found something different. participants actually reported valuing work more. Belief in the importance of work rose slightly, and many agree with statements like, quote, work is a duty toward society and people who don't work turn lazy. At the same time, people have worked fewer hours on average, but they did not stop working.

They used the financial cushion to make decisions about their careers, like going back to school, pursuing certifications, switching into jobs that have more long-term potential. One participant said the extra money allowed her to take a temporary pay cut for a role with better growth opportunities. So the research ended up concluding it's not about laziness the reason that people worked a little bit less. It's the flexibility of being able to make larger picture decisions about how and where they wanted to work. So what Sam Altman's study proved, and it was a small study, but not an insignificant one, was if you give people a small amount of money, not enough so they can quit their jobs, but a small amount, $1,000 a month that might cover half their rent, I'm guessing.

That gives them more flexibility. And they might make a career choice that is more long term as opposed to short term. They might be working as a waiter or bartender at a great restaurant where they make 300 a night, but they're never going to have a career, but they're going to make this $75,000 a year. Whereas if they joined, say, an associate and training. program. Right. If we run here, they might get paid 60 or 70.

They might be 10K less. Yeah. To start in year one as a researcher, but you're learning and you're in the room where it happens. This is all to say that there is going to be a reckoning. And the reckoning is going to be 2028. Because by 20208, that election, I'm predicting now, the 28 presidential election will be the AI job displacement election.

Let me state it again. 2018, the number one issue in the election in 2028, the number one deciding issue. It's not going to be inflation. It's not going to be wars. It's going to be AI. And is AI a net positive for Americans or is a negative? And the attack vector for the 28 election will be these companies became worth trillions of dollars while jobs went down.

A small number of elitists in Silicon Valley benefited venture capitalists, the workers at these companies. of course the founders and co-founders. And the way that we're going to resolve that issue, I predict, is with the ability to take all of these social programs and have a grand bargain. I believe it's time for a grand bargain. What should the grand bargain be? I believe the grand bargain should be anybody making under $100,000 a year, $75,000 a year, maybe it's under 50 pays nothing, under 75K pays a $750, 1% token amount to the government, flat rate, no deductions, no nothing, just send your 750 in, you're good.

Under, you know, and then something similar, you know, from 100K and under, maybe it's like $2,000. You pay a token 2% of your salary to the government. If that were to happen, that's one concession on the K-shaped recovery. The rich doing better in the top of the K, the poor, doing worse, the non-equity holder. Second, we already have one solution, which is Invest America, aka Trump accounts. That has said, hey, your kids are going to have a stake in this equity. Then another grand bargain emerges. And this grand bargain is the hardest to do, but could be revolutionary one. If we said, here's what all the social programs cost. Food stamps,

unemployment, et cetera, et cetera, et cetera, sure. Give me that number. Whatever that number is. Anybody making under 50K a year gets a percentage of that pool. Everybody making under 40K a year gets a percentage of that pool, 30K a year gets a print of pool. Is it UBI? Kind of, but what we're saying is get rid of those programs.

Stop operating food stamps. Just give people the money. Yeah. Just give them the money. I think you get Democrats and Republicans to agree. on that as a principle. I don't think anybody's married to these elaborate bureaucratic nightmare systems that we all have had bad experiences with or mostly.

Everybody knows somebody who's been screwed over by one of these agencies. We know how wasteful they are. I mean, in Los Angeles, you hear about all the time the groups that are supposed to be helping the homeless, but it's just graft and they're just like losing half the money. I think most people would understand just if you gave everybody a check, you wouldn't really need to worry about those programs. I think we get so caught up in the sort of merit-based idea of it. You know, every time Democrats want to pass a social welfare program, it's like, well, it only goes

to peeve, you have to make less than this, and you have to apply this way, and you have to meet these five criteria, or just think about when you're paying your taxes in California. Do you qualify for any of these 2,000 extremely complicated exceptions that nobody knows if they actually qualify for? If you forgot all of that and just gave people money, I think that would be an obvious solve for. a lot of these issues. It makes total sense to me. Snap. SSI. All these means-tested programs is, I think, the fancy way of

saying it. There's housing. There's unemployment. Medicare, Medicaid, everything. And it's all very complicated and convoluted. And the people who need help don't necessarily get it. Other people are defrauding the system or taking advantage of it. If we just got rid of those systems that government just passed AI

profits on to the American people, I think that solves a lot of problems. And it simplifies everything, which is obviously better. Here's my big idea. You take all these social programs. They're all means tested. They're all complicated. They all take a ton of paperwork and they take a ton of administrators. Millions of people are figuring out how to give billions of dollars to Americans. There's a much simpler solution. If you didn't make any money this year, if you make no money, you get a percentage of all those social programs. If you made 20,000 or 10,000, 20,000, 30,000, 40,000, maybe it's up to that number, 40,000, let's say, maybe 50 max. You just get a percentage of that pool. And so we take that pool and we just give

it directly to people in a simple UBI check. But if you make more than that amount, you come out of it and it's done on a scale. Then, in addition to that, you would be able to get rid of the 30% of waste, fraud, abuse, and then probably another, 30% in administration, I'm guessing it's less than 50 cents of a dollar that actually gets to people. Yes, of course. Those two things. Right. And that's what, oh, anytime you go back and pick apart like a well-intentioned social welfare

program years later, that's what you figure out. It's like, oh, we spent 70% of the total budget just on the people counting up who gets what and dealing with the paperwork and operating the office that needs to be open. Like, you know, if you think about like, I don't know if you've ever had to go to the social Security Administration Office. I had to get my... No, it's a joke. I had to get my card replaced one time.

And it's hours of waiting. There's 18 people working there. It's like, what are we doing? What are we doing? I'm doing the SNL Tucker Carlson. What are we doing? Anyway, so anyway, I think this is kind of a very interesting moment in time. Yeah.

We got Kevin O'Leary agreeing to shrink his data center project. It's kind of similar to what we just talked about. And then we've got Brian Chesky's new AI lab that he's. found. Can me hit me with Brian Chesky. All right. So Brian Chesky, of course, Airbnb co-founder and CEO, he's going to remain on as

the Airbnb CEO, but he's also founding a new lab that is going to develop AI models. The focus, at least for now, is user interaction and design. He won't serve as this new lab CEO. He's going to stay on at Airbnb and just sort of dip a toe into this. So here's the big argument. He argues that AI for travel specifically is going to, going to need a richer interface. It's going to need better designs. He thinks people don't just want to go to like a chatbot and have a conversation about their trip. They want a more bespoke designed,

customized experience. And so it looks like that may be what some of these models are going to sort of play into. Of course, Chesky notably was a design major in college before founding Airbnb. Yeah, we had a company early on, even before I think Chat ChbT3.5. and it was called roam around. It was in batch 28. And if somebody can find... I think we looked for that before and it's already gone.

We did this in a previous episode and I don't think roam around still exists. Let me finish by Zend it. If you check YouTube, you'll find it because we have the demo day of it. And I just did a search for it. So we had a company, roam around. This was an AI travel planner.

It went through our accelerator, batch 28. I believe it was before, Chat Cheapety 3.5 came out. So this is when you had, I think, 2.5 and you could use it as an API, but not as the public. It was an incredible, incredible pitch. Everybody loved it. It was like, number one. And the team, I think, wound up giving, like, memory storage credit. They gave us the money back because they were like, oh, you can just do this in chat cheptie. And I was like, you know, I still want all the great parts of this. Let's see if we can scrub to the moment when they demoed the product here. And this is a very tactical and practical, important lesson we're going to

talk about. When you're a founder, should you persevere or not? Should you persevere or pivot? Here is the Rome startup. That was in an early accelerator class pitching. What date was this long? This would have been September 20th, 2023. Roll the tape. Hey, everyone. My name is shy. I'm the CEO of Rome Around. Romaround is an AI concierge in your pocket. So you ever feel like you need a vacation from planning a vacation? Introducing Romeround. Rather than spending days planning a trip, enter two days in Barcelona, will create a complete and actionable itinerary for you. If it's not perfect, you can say something along the lines of, it's our anniversary. And we'll iterate the itinerary to recommend... It was so impressive in 2020.

walks, hot air balloons, etc. With all that time saved, you can say add a beach day. You can see day three and Costa Brava has been added. Now that it's perfect, you can request it to be translated to Spanish. And then, once it's done, you can share via SMS, WhatsApp, email. Multiplayer mode. Oh, so much. You're probably asking yourself, aren't you just a chat chip of T-Rapper?

Well, leveraging our first mover's advantage, we've accumulated millions of lawn. This was like me going back. back and forth with the founder, begging them to keep spending our money, to raise money. And I just said, you know, yes, you're going to be accused of being a rapper, but don't let that get in your head. There's so many things around the model, and there'll be so many twists and turns on this journey that there's a business here in doing this. And now, Brian Chesky is like, you know what, there's so much of a business in AI would travel, I need to make a dedicated model. Rome around probably would have made their own vertical model based on an open source one.

or they could continue to build features, and a wrapper's not a bad thing. The Corvette or a Cadillac, the Cadillac is a wrapper around the Corvette engine today. Right. Yeah. It's the Corvette's engine in their high-end Corvettes. Right. In the high-end catalogs. Does it make it a wrapper or not?

Like Jaguars and some other cars have other components in it, it's a wrapper, but it's also a brand. And it also symbolizes something. It has a look and a feel and a texture and a brand value. Yeah, of course people use chat CBT, but what if there's other features in here like a human in the loop? Right. Beg them. Please, please, please don't give us your money. Please be going.

Because it was such, you know, it's like we're talking about 500K. I'd rather see the founder go for it. But they did what I think is also a mature decision. If they don't have the want, desire, stamina, belief, and they have a different belief than me as the investor that, hey, it's just not worth it. And it's not worth my time and the opportunity across, I have to say, okay, it's not my choice. But anyway, this was heartbreaking for me. Yeah, it's kind of the same thing I would say before. It's the magic for me, I think, is multi-model.

Like you use different models and different, you know, tools for the different parts of the trip. And so one service that could pull all of that together would probably be the magic rather than just going to chat, GPP and asking what else should I do while I'm here in Rome or whatever. That being said, if I'm Airbnb and I'm a shareholder, I'm on the board, I'm an institution with a large holding. I'm going to ask Brian Hey can Airbnb put the first 100 million in

and buy 10% of the company in the case you figure this out and have a perpetual license to it for Airbnb to use But then the question becomes for Brian And this is like, you know, no conflict, no interest If this actually works and it's great And he put his cycles into it

And Airbnb didn't collect it Or Expedio wants to use it it or Uber wants to use it. And one fine stay wants to use it or Hilton wants to use it. Okay, are they going to let Merritt and Hilton use it if they're competitors or not? So just all kinds of conflicts, all kinds of interesting things will emerge. And he's got to get some skin in the game for the Airbnb shareholder, I think. But he's super smart.

He's an incredible designer, a design first LLM, frontier language model, whatever it's going to wind up being a verticalized one. I'm here for it. Go for it. I mean, if it's even a language model. I mean, we don't even know if it's an LLM or if it's something else. I mean, there's at this point, I think we're reaching the point where a lot of people are feeling like the LLM has sort of, we know what it can do and we've sort of worked out most of it.

And now we're sort of picking around with world models and other kinds of models. So I don't even, you know, I don't know for sure if this is an LLM or something else. Yeah. Layla, a travel discovery and booking app purchased roam around. There. Yes. That's right. No, RIP, I'm around.

I go mental every time I go down in history like this. This is my gift and curse as an investor and supporter of our founders. I believe in them so much that when they give up or they fall, they collapse shop, you know, which is their decision, as I said, is my decision. I always think, oh, God, have we just persevered for six more months or one more year. Travel is so tough. I feel like that's one of the tough. Everybody wants to do a travel app.

But they were added in 20, 23. this was three years ago long. We had them in the accelerator. Can you imagine what they roam around would have done in three years? If the founder came back to me tomorrow and said, can I come back to the accelerator? I have another spin on travel.

I like that guy, Shea. He was great. I remember talking to him in the parking lot at Bucks after our closing dinner, closing lunch, had Jake Howl's Onion Rings. Of course. Might have even predated Jake Howl's Onion Rings being on the menu officially.

It might have been off menu item and I'm just talking to him and talking to him in the parking lot And he's struggling with this, right, this tension. And I was like, just stay the course, stay the course. You know, like, if there weren't questions, there wouldn't be an opportunity. So I'm just on a tactical and practical moment here as a founder, persevering in an important category is, I believe, the correct decision 90% of the time. If it's a big market, if you got a lead, if you got cash in the back, you can do pay.

of course, but I think you should persevere and pivot your way to success as opposed to give up and giving the money back. I always appreciate a founder's like, hey, this isn't here, we're going to give the money back, or these are my next two ideas. Would you vote in favor and be supportive of me pursuing either of those or both? That's this moment in time. If you raise money, and this is a message to founders and investors, if you've invested in and the founder has raised money for one vision, and you believed in them for that vision. And they have the maturity to say, that vision, I don't believe, is achievable. But I want to pursue this one. I believe that is how great startups are built, which is trusting the founder's instinct, letting them collect information

and let them make a hard pivot or a micro pivot or a complete. We're shutting this down, and we're going to try this with the money that's been invested. Why, the money's already in the company. By the time you return it. Everybody's getting back 10, 20, 30 cents on the dollar. The money's already been Pock committed. Let's play the hand. Let's see if the founder can figure it out. Yeah. I love a great pivot. So many great companies are built around them. So that's just my little tactical practical summary. There you go. Rome around. RIP. We'll miss you. Should we go? Should we go off duty? I feel like it's time to go off duty. All right. So there's a ton. Jacob wanted me to ask you your thoughts on Mandalorian and Grogu.

I know you wanted to talk about this HBO series you've been watching. Take your pick. I got a dog that beat me up. You can take your pick where you want. Mandalorian and Grogu was a great season of excitement and fun without lightsabers, without getting into the Skywarker saga or the Disgratziad sequels, episode 7,89. Just a lot of fun. A lot of great characters.

A lot of great Easter eggs. with the fans. Just great summer fair. You bring your family, you know, a six, seven, eight, nine, 10 year old's going to love it. And, you know, the Gen Xers are going to love it, everybody in between. Incredibly well done. John Favreau is a genius. And the guy who took over Faloni. Dave Filoni. Dave Filoni. Yeah, he's also a genius. And I love these characters. And it's a lot of fun. Sigourney Weaver piloting an X-Wing is just like incredible. Now, some people might think it's like a season of Mandalorian. And they're correct.

It was literally the season of Mandalorian. They made it into a movie. I don't mind it. My bigger idea is that I think all these series should come out in movie theaters, in extended versions or two weeks before they're on. So if you have a nine-episode Mandalorian series, I think you should be able to watch it two days before it comes out on the street. streaming service. So if it comes out on Monday on the streaming service, Friday, Saturday,

Sunday, you should be able to go see it in theaters and then Monday you can watch it. This would create a way of having community. That's what I want to see them do. I would imagine if they did that for the Sopranos or a peak television show where you could go buy a $15 ticket and go see Breaking Bad on Friday, Saturday night. And then it came out Sunday night. Incredible. People would pack the theaters. You have done some stuff like this. Stranger Things, you could go see the finale and Stranger Things Netflix played it in cinemas. Before or concurrent?

It was the night. Like the night that the series finale debuted on Netflix, that night you could also go see it in your local movies. Yeah, I got a better idea. Two nights before. Oh, there you go. And the community would go crazy. So what I think you do is you play the previous episode

with like five minutes of bonus material. You can only get in theaters. Then you play the new. new episode, five minutes of bonus material, you can only get into you getting 10 unique minutes. So you can show up 60 minutes into this double header. Yeah. And watch the previous episode with a little more in it or maybe an interstitial between the two episodes. And then you watch the new one.

It would bring back theater because, you know, secession, industry, all these ones have great communities. And people want to go share their fandom. Severance. People would have gone to see the severance finale in theaters for sure. No, but do it for every episode. and add five or ten minutes. Well, I like where your head's at, because I do,

I mean, part of my whole thing has been you save movie theaters by mixing it up. Like, it's some first run things. It's some classic things. You can show all kinds of stuff in movie theaters. It doesn't just have to be original first run movies. And I mean, we're already seeing like if you have YouTubers that start to make movies, those can also become huge hits, backrooms. Okay, well, take me off duty with that.

My understanding is there's a movie, there's two movies in theaters that, what blew past Mandelorian and Grove. The first one is Obsession. I actually saw this one the other night. It's by a guy YouTuber named Curry Barker. He and his friend Cooper Tomlinson, they used to do sketches.

They were doing YouTube and TikTok sketches, and then they started doing these horror shorts. They did one called Milk and Serial, like serial killer, S-E-R-I-L, that was pretty well received. And they just built up it, and then they made this $100,000 movie Obsession

or maybe a little bit more than that. I'll have to look it up. That's making, you know, Ted's the number one movie at the box office. It's become like a, you know, cultural sort of phenomenon now. I want to see it. It's really good. It's really well done. The concept is very simple.

This guy finds in a magic shop, this, like, device that you can make one wish on. He's in the friend zone with this girl who's in his friend group that he's got a crush on. And so he just offhandedly makes the wish. I wish she was more in love with me than anyone else on the whole planet. And he breaks the thing. Oh, no. Then it turns out it's real. Obsession.

It really has magic. And she becomes terrifyingly over the top, obsessed with him in a psychotic and frightening way. And the actress Indy Navarretti, who plays the girl who he makes this wish about. It is an incredible performance. I mean, she goes to places that are genuinely very scary. And then the other one is this movie called Backrooms, which is this is a, going back a few years now, YouTube,

one of those like creepy pasta online things that this guy, Kane Parsons made, who's now, I believe, the youngest director in history to have a number one film at the box office. He's like 21 or 22.

So he just made a big screen version of this online meme trend. It's about people who get caught in this sort of liminal space, the back rooms where it's yellowed, and it's just sort of endless hallways, and it's kind of creepy and unsettling in a way. And, you know,

I haven't seen the movie, movie, I'm sure there are monsters and things that pop up. But yeah, taking these ideas from the internet, you know, like the idea used to be IP. We always thought about IP, but I think our thinking about it was very narrow. It was like, well, Transformers is IP and Star Wars is IP and severance is IP. But what I think especially backroom shows is IP can be a popular YouTube video if enough people have seen it and feel enthusiastic about it. Like, Gen Z will show up to a theater to see something they like, but it doesn't have to be this like very pre-established IP like we think of them.

Yes, I mean, listen, this is going. We're going back to Easy Rider days of the movies, which is a young generation who has not been at the helm of movies before gets the helm and starts delivering. And their contemporaries say, you know what, I'd like to hear some new stories. This could be the revival of the movie industry. If these YouTubers can make something for a million and have it generate a hundred million of revenue, which is, is that what Obsession is going to do? It was $750,000 was the budget.

It's so far grossed over $166 million globally at the box office. So now I want to know what their take is of that because you can self-distribute that crazy It's Blumhouse. So they're sharing it. They're sharing it with Jason Blum, yeah. Okay, so he made 50% of every dollar, I'm assuming. And so this is pretty amazing. Blumhouse has now figured it out.

A-24 has started to figure it out, which is there's a place for one to $20 million budget movies. I think their largest one may have been $50 or $60 million. Yeah. This is the future. And you don't need to get everybody to go see it. There was a product called Blair Witch Project. Back in the day, that cost $100,000, $200,000 to make and also did something similar. incredible 166 times the production cost this feels like a business i want to get into i would love

to start a movie studio i've talked about that before i have a syndicate i have my own uh money i would love to independently distribute a film like this yeah i mean that that would be the interesting thing because the people what's the one of the crazy clown oh terrifier terrifier terrifier terrifying my understanding is terrifier self-distributes art the clown art yeah the art the clown right that they're a small production company that sort of has been, they started doing like road show things and then they grew it out sort of organically. But they control the means of distribution and own that terrifier Arthur clown, which is disgusting.

And I tried to watch one. It was just way too gross. Oh, yeah. No, they're brutally, the gore is like the whole thing. They're really brutally gleefully violent. I mean, it's like beyond gore. It's like it's unrealistic, but it's just too much. Yeah, it's really intense.

Yeah. So Damien Leon, that's the creator, writer, director of these movies. and yeah, he has his own independent production company, Dark Age Cinema. But they distribute it and their idea is to go directly to the movie venues and say, hey, we'll come up with a deal that's bespoke for us. They don't have to go to Blumhouse or Paramount or wherever you would go to. They partner with Screenbox, which is like bloody disgusting, that website, that's their streamer. But that's still, I mean, they're a small independent company compared to like the Blum houses,

which has a universal deal, you know. There's a great documentary called The Dark Wizard. Have you watched it yet? I was going to try to watch it last night because I knew you wanted to talk about it, but then I have this new dog at home. You can play it in the background, the trailer while I talk over it. It's about an artistic, autistic guy who climbs and does free soloing. You all know Alex Honnold.

This was Alex Honnold before Alex. Alex Honnold. And he's a challenging character, not always likable in his pursuits. And the trailer has him jumping off of a mountain while free soloing. You'll notice he has a backpack on when he does that. And then they show later in him doing base jumping, which is when you jump off a clip. So he's free soloing, he's base jumping, and then he also does slacklining. He's a true artist. He's a sensitive guy. And he's obviously like Alex Honnold probably on the spectrum,

which makes you able to do things like hang off a cliff without a rope. And it gets into the competitive nature of these two individuals, where Alex Honnold comes out and he starts taking all the routes that the principal character in this film, I'm forgetting his name right now. Dean Potter. Dean Potter. So Dean Potter's 10 years,

a senior, basically. Dean Potter has been doing all these things that are making people love him and sponsor him. And then this guy, Alex Honel comes in and takes all the things that Dean Potter wanted to do

and just crushes them. Like, it's just unfairly stealing all his records, doing all the things people predict Dean Potter is going to do, but Dean Potter is not the athlete slash performer that Alex Honnold is. So this kind of overemphasizes their competition, which they were friends, but it does show commerce coming into art. Right. And making people do unnatural acts and competition leading to suffering and something I've always talked about with Alex Honnold, which is, I think,

a lot of what he was doing would be just as notable if he used the top rope or, you know, some amount of safety gear. There's no need to do this without safety gear. Well, and I guess people have the gimmick, yeah. It's like, that makes it exciting because maybe he'll die, you know? I don't like that as a gimmick because life is precious. Just like I don't like these people, Ocean Ramsey, swimming with the Great White sharks and the bull sharks in Hawaii and like taunting them in my mind, riding them. They're not meant to be ridden, O'Shea, and you're not meant to chum the water and jump in with them. Yeah, I understand that this is the hook, but I don't like the hook personally. I like to see them use some basic, basic

safety. Pull this one up too, Jacob. Speaking of commerce sort of coming in and intruding here, I saw this from pump. Fun. They've got a new pump dot pump, you know, pump. fun. They're the meme coin sort of service. So they've got a new thing called Go and it's basically you can pay people to do crazy things anywhere around the world with this service.

I compared it to like that movie Moneyplane or like Squid Game where people are betting on these people doing dangerous things or unwise things or crazy things. Yeah, it's a little bit like rent a human but to put people in purposefully disturbing

or difficult or crazy circumstances, do things they wouldn't normally do with the offer of these payouts. It's dystopian. I feel like this world feels this world that we're creating now, bounties for, you know, doing things in the real world that you might not want to do. It, it unsettles me. It really unsettles me. All right. So pump fun is letting people put bounties on behaviors that are silly and or extreme. An all-encompassing. bouncing bounty platform where anyone can create or complete bounties for any task for unlimited rewards. It's the any task part that strikes me as like the funny, the weird middle.

This is kind of running man, Black Mirror. And is it anonymous? So like because it's a crypto group of people, you can say, I want you pay you $10,000 to hang off the Golden Gate Bridge. And nobody knows you put the bounty there. Yeah. I mean, I'm assuming like they don't say specifically, Yes, how to create a bounty. Connect your X account and wallet. Create your bounty, set a description, time frame, and deliverables. Submit your bounty payment into escrow, pump fund reviews, bounty submission. So that's it.

There's no public face disclosure. It's just if you have an X account in a crypto wallet, you're in. Listen, we have things like Fiver or the other places where you can hire people upwork, etc. Rent to human. And there was task. agent can rent a person and have them do something? There have been other platforms that do this. You have Fiverr, you've got Upwork, you have rent to human, which we had on the program where you can rent a human to have your agent do different actions for your check work. Great. What's challenging

about this one is I know which way they're going. They want people to do disturbing things in public. Now, if it was streaking, okay, that's illegal or I don't know, you run around in your, or Mancini and you do something like Barat or Borat, like, okay, maybe it's funny, whatever, but it always goes too far and then there's always somebody who gets hurt in this equation. So the people who make these platforms have to be very thoughtful about it. And it does dovetail with Alex Honnold and Netflix having him climb a building without a rope. I thought this was negligence on Netflix's part because nobody in their right mind wants to see him die. yet everybody's tuning in and they're promoting it because he might die.

We don't want people die. The human spirit, I mean, the human life is so precious. Don't do these crazy stunts. What they should have done with him is they should have had him do it and had a net that would say 10 floors below him that just rose with him. And they say, listen, he's doing it without a thing. But if he falls, he's going to take a 10-story and he's going to get caught by the net. It's still exciting.

It's still exciting. And you know what? Then he can fall so he could do more challenging things. Would that be the worst thing in the world to watch him scale one of these crazy, you know, buildings and make it even more extreme? Have three people do it and have it be a race and have them fall and the net is following them up the building. Yeah. And then if they fall, they can jump back on the building and they're five stories behind.

I do think part of it is as sick as it is. I think there is that human impulse to like, like people looky-loose on the freeway if there's an accident. Like I think people are more curious because there's a chance this guy might actually bite it if he falls. I think Alex Arnold even, he's got kids now. So I think that last I heard. All right, listen, this has been an amazing episode. This week in Startups Twist. We'll see you on Monday.

Bye, bye, bye, everybody. Thanks for watching this week in Startups. If you liked this episode, check out more. If you're a startup founder, Founder University cohort 13 kicks off this fall. It's a 12-week program that provides guidance on building your product, launching to real customers, and pitching to investors. Top startups receive $25,000 or $125,000 in investment. Apply now at founder. university slash twist. Already have traction?

The launch accelerator invests $125,000 and connects you with 500-plus investors to help you raise your next round. apply at launch accelerator.co. If you're an accredited investor looking to gain access to quality deal flow, apply for Jason's angel syndicate at the syndicate.com. We find two to three deals a month. And check out this week in AI, Jason's experts-only roundtable with top AI founders and operators every week. Find it this week in AI.com.

Check out the Twist ticker, our daily newsletter at this week in startups.com slash ticker. Thanks again to our sponsors for making. making today's show possible. Follow the show on Instagram, follow the show on X.com. This Week in Startups publishes three days a week, Monday, Wednesday, and Friday at 5pm Central Time. You can submit an audio or video file question by emailing it to this weekin.com.