Stripe’s Crypto Wallet, M&A Momentum, and Robots Replacing Soul-Crushing Jobs | E2138

30:57Go to NorthwestRegisteredAgent.com slash twist and get your business investor ready today. For just$39 plus state fees, you can set up your company the right way. Fast, private, and compliant. Go to NorthwestRegisteredAgent.com slash twist today. hey, here's what 99 out of 100 of these things prove. Things get cheaper for customers. Customers' lives get better. I'm trying to think of an example of any acquisition where things got worse. Even the one you were talking about, where there was consolidation in food delivery, and there has been some. Yeah, yeah. What was bought tended to be the unprofitable, dying companies or sideways companies that everybody saw the writing on the wall like lift's not gonna make it out postmates wasn't gonna make it out you know whatever grubhub this or whatever like they were just things that were yeah exactly they were they were kind of things that just weren't as good as the ones that consumers selected so in that case consumers picked doordash because doordash listen i'm gonna i'm team uber you know that but i think doordash had a better interface and had better restaurants on it they were about 20 % better on a product basis.

32:11That's why they were leading. Uber has since caught up. Uber has added some other things, but I would still give a 5 % edge, maybe a 10 % edge to DoorDish's UX. I think they've just, they've polished it a little bit better. And I think they had some exclusives and that's called competition and their fees are still going down. The fees are going down. The membership programs are going up. The robots are all over LA. There are Coco robots everywhere on melrose place i'm staying in the like melrose uh you know place you know kind of area of la this week and i have seen i don't know if it's the same five but i've seen no less than 50 deliveries going on and these things are the same five robots it might be the same five and i just they're they're on the same five blocks but these cocos are everywhere on melrose place and just to add to this everybody uh coco recently raised 80 million dollars according to my dear friend rebecca skutek over at tech run um i actually didn't have this little bit of data jason in the uh the docket i cut it to focus on some other rounds but because you brought it up i wanted to throw it out there can we spend them i think we should add them to the um robotics section of quiz 500 or at least consider them and we should check that we have those other robotic companies in it those robotic companies people are not taking people are not taking into account exactly how impactful robots are going to be i have just two examples okay those robots and then if you can pull up figure tweeted on and it's a number one on their uh this is the number one story i think on their x feed for figure robotics remember figure the company figure they make a humanoid robot they were raising at an extraordinary valuation i think it was close to $40 billion.

34:01People were complaining the back channel inside the industry was it was a bunch of SPV, special purpose vehicles with what venture capitalists consider gen pop, rich dentists, putting in money on a free revenue company that didn't even have a product in market. That was the claim. That was the back channel. That's what the hand wringing was about. You might've even seen Ruloff say, we're back into this market where SPVs are being done by people with no skin in the the game in other words the you know the people who are running these spvs aren't putting their money in they're just getting the value of it and so they don't care what the valuation is because they didn't pay for it you put all that together uh and then there was this famous bmw story where the bmw their lighthouse customer the customer that was supposed to convince you to pay 40 billion was uh said something like yeah we're not using that in production and then like two days later it was like well we are in production so obviously this got to the founder i think this is my dime store psychologist i think this got to the founder of figure who probably was like yeah you know they're accusing me of securities fraud here it's a pretty serious shit uh i'm going to prove that our product kicks butt so they've released not a six minute video a 60 minute video of a figure robot and i'm going to talk over it here alex there's a figure robot it's on a tether i believe i think it's hanging from a tether because these things don't stand very well um so they're not like the boston dynamics one but this is sorting packages putting the barcode down and then moving it and so pretty slick so yes in your tech you've been covering technology for two decades so you understand that if it can do this today next week it will be going 15 faster and then every six weeks it will be going twice as fast until physics like the physics of you know moving objects in the world but this will become lightning fast and if the error rate here i don't know if there was an error rate stated but it was just given the general um it was given a very general kind of like a chat gpt prompt hey just put the um just put the um barcode face down right and jason just to be clear this is not cgi this is not this is an actual figure robot doing this that's my understanding okay so so two things stand out about this one the dexterity of the robots tremendous just backing up your point that things are moving very quickly but also this i think details precisely and exactly why i'm pretty excited about having more robots in the market and why i'm very bullish on humanoid robots in particular this job sucks right like who wants to i don't want to stand there and do what is this a package a second all day long just yeah it looks like it's doing one two well we've stopped at a weird time here four five it did two and five seconds it looks like six seconds do one every three seconds right now so 20 a minute yeah yeah so you're correct um now how many people work in amazon factories doing something like this we'll assume this gets 10 percent better every month every seven months is twice as good so this will be doing odd size packages larger packages dealing with arrows it'll be doing a lot of other like slightly more complicated but you're right people are getting paid 15 to 20 an hour to do back breaking monotonous labor like this which is kind of akin to being in a torture chamber let's call it what it is to spend eight hours doing that is it's work it listen i am no judgments you got to pay for your kids, food, got to pay for shelter.

37:40I'm not making any judgments on people. I've been there. Literally, I've been there in life when I was working week to week to have food on my table and watch my parents do it. So no judgment on the people, but as a human, as you're pointing out, to do this kind of labor is soul crushing. You can do it in a fun way. You can make the best of it, but it's soul crushing for 15 bucks an hour, 10 bucks an hour. These will cost$1 an hour and they will work 23.9 hours a day yeah just gotta plug them in and squirt some wd-40 on them and off they go you're done you're done so the job description the the job destruction the job retirement that we're gonna see i'm writing a blog post on this i'm gonna release this weekend it's going to be greater than anybody anticipates uh and in our industry okay we don't want to talk about it anymore my thesis is five years ago we would talk about it alex you know why because it was hypothetical oh it's no longer hypothetical waymo was hypothetical nobody was riding in them nobody was getting coco or that other company you know delivering their burritos serve robotics delivering their burritos nobody had seen these robots doing things and cafe x you know i would get on the margins people saying like oh is this going to get rid of every barista job and i was like it's going to get rid of a lot it's going to get rid of the people who are not doing it for artistic artisanal purposes because you want your coffee perfect you go to cafe x it doesn't make a mistake if there's a mistake it's like you know educate whereas 30 40 50 percent of drinks at dunkin donuts and uh starbucks have a mistake literally like one in three or four yeah uh so anyway my thesis is job destruction in the hundreds of millions i'm going to put it at four or 500 million, according to my early estimates.

39:26Globally. Globally, four or 500 million jobs, three million, two to three million jobs retired a year. I'll say one to three jobs. I have to really refine this, but I think one to three million jobs in the United States are going to be retired per year. So the 60 million people, 62 % of people participating in labor, 150 million people working jobs, I could see 1.5 to 3 million of those jobs going away per year. And what jobs replace them? Is the question. There will be jobs to replace them. How quickly? Because there could be a period of time in which we delete some jobs but haven't made the new jobs yet.

40:03I'm a very long-term believer that we're going to stay at full human employment for forever. But I'm not convinced that it won't be a period in between the two. And that's what I'm concerned about for folks who currently make 15 bucks an hour doing this for Amazon in a warehouse. But I have driven night shift berry harvesters for a farm for 12-hour shifts for eight bucks an hour. And let me tell you, worst job of my life. Like just sounds pretty rough. Dude, you drive a one mile an hour in a straight line and you just want to die. So yeah, I'm totally in favor of this. Here's my question. How did we get from stable coins to here?

40:38So we were talking, I have no idea. The audience will tell us. We were talking about, oh yeah, acquisitions at the companies. Right. Lina Khan. Yes. Competition. Yeah. And then robots. Okay, I'm going to take us back to the beginning. and just answer the other half of your question. The other company that we have in the Twist 500, and I've lost the tab now, is OneMoney, and they're building their own network blockchain strictly for stablecoins. So the idea there, and this is a kind of a binary bet, it's either a billion dollars or it's zero, is that existing blockchains are a bit slow. And so if you want to have something that can handle the kind of transactions per second that a Visa or MasterCard handles, you're going to want something more purpose-built just for that.

41:18The anti-argument here is, is that base the l2 that uh coinbase runs is very good so we'll see who um who ends up being right but that's the latest on stable coins and competition yeah and robotics uh pretty amazing how quickly and you know there's another founder lesson here that i think everybody should just internalize um because it's important sometimes it takes the third swing at the bat fourth swing at the bat for these things to become reality we saw e-cash before the crypto movement then we saw crypto then we saw icos we saw all kinds of permutations over 20 years around digital currency even paypal right um and over that 30-year journey journey it ended with the the optimal product is a stable coin that's what consumers want they don't want to worry about their money is going to disappear.

42:15They want to know it's back. They want it to be lightning fast. So we came to a very basic conclusion that people knew in the PayPal days, the PayPal mafia knew it had to be safe and it had to be fast. So you had to trust it and it had to be fast. And it just turned out the manifestation of that was a stable coin. Robotics, we knew it had to be intelligent. We knew it had to be predictable and we knew it had to be cheap those boston dynamics ones were absurdly expensive you know those are half million million dollar yeah you know units they didn't have llms they didn't have you know an ability to say what's on the table in front of me yeah put the barcode face down and it's like i know what a barcode is and i know what face down is because i'm a language model that's pretty basic shit so the manifestation of that was a large language model in commodity hardware that was driven to that point by the smartphone and the eevee revolution which made small batteries and incredible sensors and incredible processors and video games from nvidia so the ultimate manifestation can be predicted if you just assume all the technological barriers become cheap and then it's just a matter of when it arrives this is getting so cheap that uh i couldn't find it in my notes i'm not going to spend five minutes looking for it but there's a company in their latest YC batch that made a$200 robotic arm that you can prompt using natural language and it can play chess for you.

43:41Robotic arms used to be industrial only that it cost you a bajillion dollars and you need to set up for one single use on a assembly line for cars and they would do one weld 20 billion times. Now they're just desktop toys you can program with English. I mean that just makes me so excited, Jason. The future is going to be freaking awesome and I'm excited about it. So yeah, these two or three joint um gantry robots i think they're called um is uh you know there's there's different types of these but i think that's a jantry um cartesian jantry robots uh three linear axes providing straight line movement are well suited for pick and place tasks and assembly and those have been around for a long time they are wildly expensive and they're mounted and you mount them and you pay i think they you know they got down in the the ones in the cafe x machines uh you know it's pretty public knowledge you can look at them i think when they started those were 30 40 thousand dollars and now they're ten thousand dollars if you want to have an industrial one that you can run for 10 years and not worry about now if you want one for home use and you're just going around yeah i could see conceivably making a 500 one so here is uh the latest entry this is vassar robotics i did find it and uh here is the demo i was watching earlier today that i when i was going through the uh latest cohort from yc it's doing like a language model uh commentary on its moves but it's yeah this is a toy janky but here's the thing 200 bucks and it's sold out yeah it's a toy yeah but toys become tools uh that is a another pattern in silicon valley where things that feel like toys quickly become, you know, uh, tools and that toy, like Canva was a toy for a lot of people.

45:27Um, like Microsoft paint was kind of a toy. Uh, Canva was kind of a toy. Then they become essential and that will start. And like, that looks like something like for educational use to have fun, like Legos, that's a kind of modern day Lego and smart of them to, you know, make them, uh, as Paul Graham always says, you know, founder of YC do things that don't scale. It looks like a 3d printed you know janky easy to criticize you know weird looking product they sold out how many did they do a hundred you come up with a million ways to diss it just like people did with the roadster oh he's making a hundred roadsters for 150k it's a lotus elise it doesn't have power steering it's it's loud it has no suspension when you hit a pothole you feel like your fillings are going to come out of your teeth like this was the criticism of that car and now you have model wise self-driving around austin you know i've seen the videos i've seen hopefully they're safe you know i i kind of feel like there should be a safety driver in them for the first year you know putting aside tesla or any company yeah yeah you know just to go off on another diversion here i think like having driven fsd since inception and you know i i just got the latest juniper it's unbelievable but i think it's probably an intervention every hour for me on the hardware four and two interventions an hour on hardware three and there's a website it's really cool there's a website tracking disengagements what they call critical disengagements and there's an amazing group on reddit if you want to really get into fsd and where it's at today the fsd group and then this other group of like tesla super fans they're recording through the computer and uploading all of the data and telling to tell it tell him tree what is that word telemetry telemetry thank you and they're publishing it so if you find this fsd tracker i think it's called fsd tracker tesla yeah fsd community tracker tesla fsd tracker.com there's a bunch of software uh One of them is called MatterKey.

47:30And this is like super interesting. And it shows the number of drives with one or no disengagements, city distance to critical disengagement in kilometers and miles. And so you can see over time what this different software is. And right now it's saying every 200 miles, there's a critical disengagement. Well, 200 miles is incredible achievement. It's going straight up. so this is um yeah i've been down the fsd rabbit hole uh and fsd is extraordinary i you know my gut tells me just to appease regulators and avoid you know what happened to cruise and uber and waymo in the early days which and zooks just happened you know these little fender benders or god forbid a human getting hit that's going to happen in all self-driving cars teslas waymos it's just a matter of time.

48:25Random stuff happens. So we should be judging these based on how much safer they are, as you've pointed out, than human drivers. But that's not how they'll be judged by regulators or the public or the press or just society writ large. The majority of people are looking for these. I've come to the conclusion to be superhuman, not human. Superhuman means, I think, perhaps LIDAR to see through fog and see through rain, which obviously Tesla's taken a different approach, although they're testing mules I saw on this FSD forum on Reddit. But here's the data, number of drives with no critical disengagements, 98 % of drives.

49:01So, you know, basically it means like you could do 98 drives out of 100 and not have a disengagement, but that also means one in 50 people who take, imagine if one in 50 people taking a Waymo got a critical disengagement, that would be often. That means if you used it, you know, daily, or if you used it, you know, When you go out to dinner twice a week, four a week, 200 a year, you would experience one a quarter, one critical disengagement quarter. That's too much for a human's brain to process. They will videotape it. It'll become a social media thing, et cetera, just like it did for Waymo when that Waymo went around in a circle.

49:35So I think they should, I think the states should have the right to make their decisions. I don't think this should be a federal mandate because I believe in states' rights. number two i think that they should have safety drivers be the standard for all autonomous driving for the first year or x number of miles hundred thousand million miles whatever it is because a safety driver costs nothing and they have the safe they have my understanding is they said there's two or three safety drivers per car for tesla during the test and that there's one uh or 1.5 for waymo you know these backup drivers who are monitoring everything so if that's the case like there's no shame in the safety driver game put one in there for the first year and just have them sit there and talk to the customers and put their hands down and say here watch it go i'm here in case there's a critical disengagement a kitten runs across there because we want to be better than humans and here you see um city distance to critical disengagement uh and this is only for people who have 2 000 miles reported this yellow chart here and as you can see uh you know back when it was 10 and 11 and 12 you know people were going version number version number of the software thank you uh you know people were going 61 miles then 146 miles back down to 110 then all of a sudden in 12 and 3 148 and 237 so that's really that slope is very notable it's getting better and better and better it will be getting probably exponentially better but still every 200 miles if The average ride is five.

51:04That means every 40, which was the math we just did. Every 40 rides, you get disengagement. That's a little too much for consumers. Now, of those, some number will be picked up by the remote drivers who are connecting to the car over LTE. Who knows? Maybe we'll put a satellite dish in those. I was about to say, who owns a satellite internet company? I mean, you put Starlink in there, it would be even... I don't know if the latency would be lower than 5G, actually. Oh, good point. The top line bandwidth would be higher, but maybe more consistent. or maybe you just have both and you just send the information on both and if one you know you have some sort of redundancy actually that's probably what they would do that's what i would do so uh because the cost is so de minimis uh so here you go uh keep an eye on that and then the fsd group in reddit you can see all the nuances for example they figured out uh it's amazing what this crowdsourcing can do they there's a thing where if there's tire tracks on the road you You know, somebody peeled out or slammed on the brakes and they left tire tracks.

52:02Yeah, yeah. Confuses the hell out of cameras. It thinks there's something on the road. Which makes total sense. You think about a camera, all of a sudden there's a black streak. And it's a random black streak. It's not a predictable black streak. The car could have slammed on the brakes for one second, for five seconds. It could have spun out and made a circle. It's just like some random blackness. And I think that's where LiDAR wouldn't get confused. And so I have a feeling Tesla's going to have a very successful launch. But I think it should be very slow. and steady with safety drivers. That's my best advice is safety drivers for the first six months or a year, let them talk to people.

52:37I don't know what you think about the sort of deployment of this technology, but Waymo took five years with safety drivers, right? Something like that. Yeah, they took forever. My question is, would you have this done on a per market basis or a per fleet basis? For example, Waymo is going to go to DC next year. Would you say that Waymo needs to have safety drivers for the DC market itself or because they've already done this in other other cities they don't need it anymore that waymo does safety drivers every time they go to a new city because each city is unique that's what i was shooting for here so essentially i think that your 100 000 or million mile mark i would do on a per market basis for all these companies 100 % per city per geo and i i used waymo two or three times while i was here the wait times are terrible uh the cost is more expensive than uber uh but it's kind of rock solid in the area of la that it's in it's on surface streets i thought it was actually too aggressive there were a couple of moves that waymo did that i was like that's a little too aggressive uh like it cut somebody off and so i was looking at it and i was like i'm kind of shocked that a waymo hasn't gotten into a serious accident in los angeles because people in la are frisky with the way they drive and i i can't believe you know i think it's just amazing they haven't gotten into a serious accident yet or that that date is not out there and the only caveat i'll give you to that date I was just sharing at Tesla FSD tracker.com is there's been a lot of back and forth of if you should trust that data because I've heard both arguments.

54:07These people are out to get Tesla, their shorts, and these people are super fans. They're long the stock. So when you, when you add a stock and a bet like Polymarket, shout out to our friends at Polymarket. I don't know if they have a robo taxi bet, but we should pull it up if they do. That would be pretty cool. uh the when there's a public market stock at stake here you could have people place a bet for a million dollars and then fake a crash in a tesla um and there was a very controversial tesla crash where the person claims fsd drove them off the road and it was actually there was data that the person jogged the steering wheel before it went off the road and disengaged the fsd oh so that's not so good but they but the person driving claims they were still in fsd but the data shows they move the wheel so this is kind of like shows one of those gray areas between the two oh here we go will tesla launch a driverless robo taxi service before july this is the most pertinent polymarket i could find 28 chance but let's look at how they settle the bet when you're doing a poly market, you want to know how this is reconciled to the rules.

55:22This market will resolve to yes if Tesla publicly launches a fully driverless taxi service by June 30th, 11.59pm Eastern Time. Otherwise, it will resolve to no. Any service that allows a member of the general public to summon and ride in a Tesla vehicle operating without any human, on board or remote. Oh, okay. Actively controlling the vehicle will count. A human may be present in the vehicle or monitor remotely for emergency intervention, but they must not be physically positioned to take control. For example, no safety driver in the driver's seat and must not actively steer, brake, accelerate, or otherwise drive the car under normal operation.

55:55That seems fair when they fully explain. I mean, I think that's a good way to resolve it. And I think that's probably, this is why the devil's in the details. Yes. There it's all likelihood that this will be still be in private beta and you would need to be accepted to the program. So a public member couldn't just download an app and do it. Like, you can just download Waymo now in Los Angeles. Yeah. You don't need permission, so. All right, great show. Anything else we need to get to in terms of housekeeping here? I think - On the housekeeping side of things, no, I have a really great Reddit rapid response for Monday for us to get through.

56:29Oh, let's do it. Oh, you want to do it today? Let's do it. Yeah. I love a rapid response. A Reddit rapid response. Here we go. All right, so I was over on our dear sister forum. Oh, by the way, Reddit - is suing somebody we didn't touch on that did you see that reddit is suing somebody for uh ingesting their forums and then there was the last week we missed it the anthropic so they're suing anthropic and then disney is suing another company so all this journey midjourn is suing by disney is suing the nudify people who did the that ai app so quite a lot of okay let me break it down for everybody.

57:09Reddit, the online social forum is suing Anthropic, the foundation AI model company, best known for its clawed family, if you will. The problem is Reddit's kept telling them, please stop scraping us and Anthropic's been scraping them. Now I'm sure both sides will have allegations, but there's now a lawsuit. The second thing Jason's talking about is Disney, the famous consumer media brand, suing Midjourney, which makes generative AI models that make videos and images. Not a huge shock to see that. And then what's going on with meta and nudify they're suing the parent company of that service to stop advertising on their platform because they've kept doing that even though meta said no because they didn't want to be associated with taking money from apps that will make uh jason help me out here what's the term for when you make a synthetic nude or somebody yeah that's not revenge yeah it's non-consensual artificial imagery yeah nudes i guess it's disgusting yeah so um the first two are the really important ones on an industry the the third one's obvious you shouldn't do this stuff.

58:09It's illegal. And they're going to pass laws for it. The first two are really worth meditating on. I've been saying for a year, you should not be able to take Disney's characters and make Jedis. That's their opportunity in the market. You can be certain Disney is evaluating what consumers will pay to make our birthday card with me as a Jedi. If Disney offered me to make jedi knights i was just at disneyland i did rise of the republic rise of the rebellion is incredible ride disneyland's an incredible experience my my daughters want to go to star wars section immediately and they sell the ability for you to buy a lightsaber and construct it they sell jedi robes they sell pictures of you on the rides it's like the whole it's their ip that they spent billions of dollars on tens of billions have been invested in it they bought it for billions of dollars the opportunity to create a jedi image with you as a customer is disney's and disney's alone and the fact that you can make a jedi version of somebody on one of these services is because they ingested disney's ip yes here's an example of that they didn't train it by the way they stole it they stole it they took it without permission yep that's stealing you can't steal people's ip so if somebody wants to create shrek minions spider-man darth vader it should say we don't have the rights to those characters you can make a green org you could make a samurai with a laser sword but you can't make these other ones go to disney which has that opportunity and by By the way, Claude or Grok or Chachipiti, what an amazing opportunity.

1:00:00Give Disney a billion dollar deal for a hundred million a year to be the official partner to make these things and let them promote it for you and share revenue. And now you've figured out a way for the writers and for the artists to get paid. And the companies that own their art get paid even more. To underscore Jason's point here, we're not going to get to it. We don't have time, but Tolbit, a TOS 500 company that does kind of the marketplace between IP and AI companies, Jason, also created by humans. And there's another one on the list as well that's escaping me. Now, they dropped a report all about essentially AI scraping and how prevalent it is and how much is happening.

1:00:39The latest change here is that now most AI scrapes are not coming from model training, but from RAG. Essentially, when you do a search and it goes out to find information for you, those are now the majority case. And the amount of traffic that is sent back remains effectively de minimis. So I think you're describing one facet of a larger problem, but we have not figured out how to repay IP holders for AI use of shit. Very simple, make the decision. If you're willing to pay$6 billion to Johnny Ive for his company, you should be willing to pay a billion dollars for Disney for 10 years. Literally, you could buy all the rights to the Disney characters for a billion dollars, I'm sure they would go for that, but you're giving Johnny Ive six.

1:01:21So it think people haven't thought this through license reddit like google did for 50 million 100 million a year it's not a lot of money compared to the value you're getting yes and then reddit should be looking at their top contributors and saying we've created a contributor fund people who have accounts over five years old and that have done over a thousand posts are now going to get a 20 rev share in the licensing data we have and then just start shipping them reddit gold and they can donated to charity or they can sign up in 1099 when they hit a thousand dollars or five hundred dollars and take that money out there could be a beautiful system here that shows the world how to do it correctly it's not too difficult what i just described could be built in under 30 days by the entire industry they're choosing not to do this they're choosing not to do it well because right now it's free if they win these lawsuits they're not gonna win they're gonna lose they're gonna to lose and they're going to lose hard.

1:02:20Like when I, and when I say lose, they're going to get injunctions against them. So what they should be doing, Sam Waltman's a great deal maker. It's like his best skill. Google's got a lot of history with doing deals. Go to folks and say, Hey, what would be reasonable for the next five years? Let's come up with a system for five years. If people want the output of your characters, they want to put in prompts around your IP, we can alert you when that is, you tell us your keywords. And then when the output happens, we can say officially licensed by. Just like if you want to use Getty images as your big, bold images on your blog, you know, that's not fair use.

1:02:56Now, if you want to crop, you know, 5 % of a low res or a third of a low res Getty image and you're making commentary on it, yeah, you could do fair use. But if you're doing that every day, all the time, you know, you're probably going to need to get permission. So there is like a balance here and just be thoughtful about it. This is my message to the industry is like, be fair. put the fair in fair use very simple folks well I couldn't say that better myself alright Jason do you want to do the red rap response or save it for Monday? you know what I've got a call with the founder right now so I'm going to demur we'll do it on Monday thisweekinstartups.com slash docket and we have a TikTok account that I want everybody to go find and an Instagram account go find those follow those and just say hi to us over there we're trying to build up those channels and we'll see you all next time on this week startups have a great weekend bye everybody