Sam and Shaan dig into the rise of single-purpose robots — lawnmowers, restaurant automation, pizza bots — and why that wave may arrive before humanoid robots do. They then break down the Elon Musk vs. Sam Altman lawsuit, tracing OpenAI’s origin story from a gentleman’s agreement to fight Google’s AI dominance, and arguing that Elon is losing the legal battle but winning the PR war.
Speakers: Sam Parr (host), Shaan Puri (host)
Episode Intro [00:00:00]
Shaan: All right, everyone, we just finished recording this pod and we wanted to come back and give a little bit of intro, because I feel so many different emotions. Here’s why — the stakes are super high. We’re talking about what’s going on with robots in the real world: robots being used for lawn care, restaurants, all different types of things. But then we go over the Elon Musk versus Sam Altman lawsuit. What is this lawsuit between two of the power players of Silicon Valley? What’s our take on it — which might surprise you — as well as some of the brilliant nuance and marketing and PR going on here. We kind of call it out. This is actually really, really important in my opinion. Listen to the entire thing. You have to listen to the first half to build up the story and the stakes for the second half. Trust me, you’re going to understand exactly what I’m talking about when you get through the whole thing. Enjoy.
Timing Is Everything: The Robot Moment Is Now [00:00:45]
Shaan: All right, what up Sam?
Sam: Hey, what’s going on. I got to tell you something. When it comes to kissing, eating ice cream before bed, or business — timing is everything, my friend. And I think timing is important in all aspects of business and in technology. There’s a moment happening right now. I can feel it. It’s like when you hit your funny bone and your whole arm gets hot. I can feel it right now. It’s about to change. Robots are going wild. I got to talk to you about robots.
Shaan: They’ve been around for a minute, but now it actually feels like a reality.
Sam: Exactly. Right now a lot of people are talking about Tesla because Elon Musk came out and said the future of Tesla is this humanoid robot — Optimus or something like that. He’s like, “Hey, we got this robot that can walk around, pick up stuff, do tasks, it’s going to change the workforce.” And your boy Brad Adcock — he’s got a company called Figure, which you’re an investor in. They have their robot. I think they just raised like a billion dollars?
Shaan: Last week I think they raised $600 million at a two and a half to three billion dollar valuation. Nice little investment there for me. I was in on the first round, I doubled down on this round.
Sam: Is this going to be your Mona Lisa? Your big win?
Shaan: It could have been anybody’s because you had him come on the podcast and tell the world what he was doing. Good on you. Multi-billion dollar valuation right now. That’s a big deal and people are really excited about it.
Sam: There are stats that say labor is 50% of the GDP. The labor market is ten times bigger than transportation, which makes sense why Elon is like, “Forget cars, it’s all about robots. We’re an AI company, we’re doing robots.” He is very good at repositioning and reimagining the vision of the company over and over again. It went from affordable electric cars, to self-driving autonomous vehicles and robo-taxis, to “your car will become an appreciating asset instead of a depreciating asset,” and now it’s actually not even about cars — it’s about AI robots, and that’s the future.
Sam: But I want to tell you about something that’s less fancy than that, because I think it’s going to come first. And that is single-purpose robots — AKA roombas on steroids.
Single-Purpose Robots: The Dishwasher Standard [00:03:30]
Sam: You know the Roomba — it’s a single-purpose little vacuum robot that goes around your house and works. It’s actually useful. An even better example is the dishwasher, which is in pretty much every home in the United States right now. It saves you a ton of time. When your dishwasher breaks — terrible experience. The dishwasher is a single-purpose robot that’s actually out there in the wild. We should be striving to create something as useful as the dishwasher. That should be the North Star right now for a lot of companies.
Sam: So I went on Twitter today and I saw this great video about a company called Electric Sheep. Have you seen this?
Shaan: No. Electric Sheep — that’s a great name.
Sam: I think it’s a lawn mower. It is. Imagine if you took the Roomba and put it on a lawn. It’s a robot that mows your lawn, does all kinds of landscaping. What’s interesting about these guys — not only are they a company building a lawnmower robot with a camera on it that can go around your lawn, they’re actually buying up and rolling up landscaping services. So it’s a combination of a sweaty startup and a moonshot startup together.
Sam: These guys are going out, buying landscaping businesses in the Bay Area, and using them as a testing ground for their robots. In their office it’s just like astroturf — you can do a square of astroturf pretty easily. But in the real world your lawn has all kinds of lumps and dips and divots and trees and rocks, and the robot has to be able to work. Same way Tesla is putting cars on the road collecting data all the time, collecting way more data than the next biggest self-driving car company — these guys are doing that with their lawn care robots. They’re buying landscaping businesses and using them as a sandbox. They take their robots out, test them, gather data, and over time they’re like, well, we’re buying these landscaping businesses at one valuation, and then we’ll slowly start replacing the workers with robots and improving margins. It’s like a private equity play as well.
Sam: I thought that was a pretty fascinating approach I hadn’t heard before. I can’t decide if that’s overkill — maybe it’s not actually necessary to do two hard things at once. But I did find it very interesting. What do you think of that strategy?
Shaan: I’m looking at their website right now. It’s slick. This is cool. I agree with you the overkill thing bothers me — that could be true. The problem is the founders went to Stanford, and they’re smart enough to build a robotics company, which means they’re almost smart enough to get themselves into trouble. Anytime you use the words “vertically integrate,” be real careful. You might be mid-witting your plan right now.
Shaan: For this founder, there needs to be a Venn diagram — he’s like the dumbest smart person, and he needs to hang out with the smartest dumb person, and we need that overlap to make this the perfect business.
Sam: This is awesome. This is cool. Another thing I worry about — it’s badass, but is it a badass team executing on a small opportunity? Like, is lawn mowing a big enough prize?
Shaan: I think so, because any time private equity is rolling up something, there’s a lot of money to be made. They don’t roll up small lemonade stands. They go into cash-flowing industries. If you add up all the landscaping businesses — either they’re using them to prove the tech and then go lease or sell the robots to every landscaper in the country, or they’re going to try to out-compete on price. I don’t love that idea, so I think it has to be: use this to prove their tech, then go license out to every other landscaper.
Sam: On their about page they list their advisors — Dan Foley was CEO of a company that sold to Bright View. I Googled Bright View. Bright View is a commercial landscaping company that did roughly three billion dollars last year in revenue. It’s publicly traded.
Shaan: So the newsletter guy called the landscaping industry “small” and then looked it up.
Sam: I called it small, then I looked up Bright View, and I realized I’m an idiot. I don’t know anything. Every great man is willing to change his mind. I’m only one Google away from saying sorry.
Sweetgreen’s Infinite Kitchen: Trading Opex for Capex [00:10:00]
Sam: All right. So that’s the first one. Now, where else are robots being used? Again, these single-purpose robots — not general robots. The same way we got specific AI before AGI, I think we’re getting specific robots before general robots. It’s been like 20 years since DeepMind could beat Gary Kasparov at chess. We solved chess, then we solved Go, then we solved math, then we solved route planning — single-purpose AIs before general AI. Similarly, I think the timing of the opportunity right now is not the Tesla robot or even the Figure robot. Those are the biggest wins eventually. But I think before then, almost by definition, we have to see specific single-job robots.
Sam: So this podcast is listened to not only by people who are trying to make their first million, but by people trying to make their next million. And one of those people is the CEO of Sweetgreen. I was looking up where else robots are being used and I saw that Sweetgreen now has what they call their Infinite Kitchen — basically a robot-driven version of Sweetgreen. You normally go in and there are like four people working the line and four people in the back doing prep. What they did is they bought a robotics company for $50 million, then used that technology to build a concept where you basically have one person in the store and the rest is all done by robots.
Shaan: Is this Jonathan Newman?
Sam: Yeah, I know that he’s a listener. You want to know how I know?
Shaan: You’re getting free Sweetgreen and I’m not.
Sam: Even better — he emailed me and took a screenshot showing that he subscribed to My First Million on YouTube, Apple Podcasts, and Spotify. So if you want to be like Jon and start a multi-billion dollar company, you’ve got to subscribe.
Sam: So check out this video. It’s basically like a giant wall where all the ingredients are in these robotic dispenser arm things. Whatever you order, the bowl goes across a conveyor belt, the bowl is spinning so it can receive all the ingredients at the right spots. Dispensing almonds is pretty easy, but feta cheese — which is really crumbly — you’ve got to be really careful with it. They fine-tuned this thing and now it can do 500 bowls per hour.
Sam: I DM’d him and I was like, “How much can a human line do?” He said, “Maybe 170 to 200.”
Shaan: Wow. More than 2x the speed.
Sam: You have a robot doing work that can go 24/7, never calls in sick, always puts the right proportion in every bowl. You get perfect food cost control. And the strategy — you trade Opex for Capex. Can you explain that a little more?
Shaan: Today all restaurants run on human labor. Your human labor is your operating expense — Opex. For every hour you’re operating, every bowl you’re selling, you’re paying wages. Let’s say your labor costs come out to 33% of revenue. You can’t really change your food costs, but could you change the labor cost?
Sam: Right. What he did is every restaurant that puts in these robots costs more money upfront — more Capex. In their last earnings call they said it’s about $500,000 to put the robot in. That’s $500K of Capex, which you can depreciate — good for tax purposes — and it’s a one-time cost versus Opex, which is every single day of operation. And what they found was that for that extra $500K investment, they get at least an extra 7% of net margin. That creates a break-even point — maybe two years in you’ve paid off the cost of this robot. And he’s like, “The $500K capex? That’s today. That’s the most expensive it’ll ever be. Costs are only going to drop as we scale and simplify the design.”
Sam: I asked him: in 10 years, what are the odds that your restaurants are basically robot-driven, not human labor? Is he allowed to say what he said?
Shaan: I told him, is there anything you can’t share? He goes, “No, I’ve shared all this publicly.” And his answer was — it’s very high.
Sam: Just thinking about that out loud: in our lifetime, a lot of restaurants are going to go from lots of people who’d rather be checking their phones than standing there in a stupid uniform, to — you walk in, push a button, your bowl gets made by a robotic arm, and you walk out. There’s going to be one human whose job is robot maintenance — “Oh, the carrots machine jammed, let me go unjam that” or “I need to refill the container of apples.” The humans are all in the back doing fresh prep of ingredients, and the robots are serving the customers. That’s how they split it today.
The Pessimist’s Case: Will Robots Kill Jobs? [00:17:00]
Sam: You and I are fairly optimistic. We try to see the bright side of most things. But let’s try to be pessimistic for a minute. The argument you’re going to hear from a lot of people is: “This is going to ruin the economy, ruin jobs.” Let me ask you a question before you give your opinion — you own a business that employs warehouse workers. What’s your annual turnover?
Shaan: We use a 3PL now, but before when we had it in-house, I would say every year we’re replacing about 90% of the team.
Sam: I was reading — I think it was Amazon, some company that employs hundreds of thousands of warehouse workers — for them it was 100% turnover. I remember talking to Brett Adcock from Figure and I go, “Brett, are you preparing to be the most hated man in America?” He’s like, “Why?” I said, “Because you’re taking jobs.” He said, “Well, I’m not actually taking jobs. There are all these warehouses that literally nobody will work at. They cannot convince people to work there, whether they can’t afford to or just aren’t interested in that type of work. Plus the average turnover is 100%. They just can’t keep people in these jobs.”
Sam: So my question to you is — are you pessimistic at all about what happens in 15 or 20 years? My first job was a laborer and then a dishwasher.
Shaan: My first business was a restaurant. I was the guy chopping tuna and doing dishes every night. I think there are two versions of pessimism. The first version is: it’s not going to work. The second version is: it’s going to work and that’s a bad thing. You’re talking about the second version, right?
Sam: Yeah.
Shaan: I think it’s going to work. 100% it’s going to work. And by the way, I asked the Sweetgreen guy — I go, “What’s the non-obvious insight? The thing that’s only obvious when you’re in the weeds?” He goes, “Well, it’s like most things in technology. The first 90% is easy — which makes it really easy to have a sweet demo and everybody’s immediately like, oh man, this is going to change everything. And it will, eventually. Probably longer than we want.”
Shaan: He goes, “The first 90% is easy. The next 9% is ten times harder.” Simple example — dispensing almonds: easy. Dispensing the right amount of crumbly feta cheese is literally ten times harder than the almond problem. On a Tesla: my car self-drives, and the autopilot is mostly good unless there’s a flashing yellow light, in which case it freaks out and I almost die. Or if it’s a dewy morning, or foggy, or if there’s construction — edge cases everywhere. And then he said the last 1% is ten times harder than that.
Shaan: With self-driving cars, the problem is nobody wants a 99%-good self-driving car, even if that’s actually five times better than a human. It’s like — if I’m going to die, I’m going to die at my own hands. But I don’t think you have the same problem with restaurants. The size of the prize is smaller, but the barrier to deployment is much smaller. If you ask for light sauce and get medium sauce, it’s okay. It’s not a self-driving car crash.
Shaan: In software development we used to say this — I would walk around eating cashews going, “Hey guys, how’s it coming?” And they’d say, “Well, we’re about 90% of the way there.” I’d be like, “Great, 10 minutes left!” And they’d be like, “Wait — 90% done means we’re about halfway done. The last 10% always takes at least double the time you’ve already spent.”
UBI and the Coming Labor Reckoning [00:22:00]
Sam: Like ten years ago, Universal Basic Income was like a thing. UBI. I’m a capitalistic guy and I was like, “What? That’s the worst.” Then I learned two things. One — do you know in Alaska they give you like five grand? The oil dividend — you get a share of the money they’re making from natural resources. I think sometimes it’s a tax credit where you pay less in taxes or if you don’t owe anything, you literally get a check for like $4,800 or $5,500.
Sam: The second thing I learned was I started researching UBI and I was like, “Wait, why is Sam Altman advocating for UBI? Why are all these smart robot-makers advocating for it?” And I thought — when my wife comes home and the kitchen is clean and she asks what I did, it’s because I did something else I’m not telling her about. That’s what they’re doing by starting UBI. Sam Altman, along with a few other guys, were doing experiments in San Francisco where I think they were giving $800 — I don’t want to get the numbers wrong — to people for some number of years, then they stopped. The results weren’t entirely positive. It wasn’t like, “Oh my God, this is great, let’s roll it out nationwide.” But it’s still an interesting concept.
Sam: Because when I learn more about these robots, I actually think — I’m just old enough where I don’t think this is going to impact me. But for most Americans who are young now, it’s going to have a massive impact on their future.
Shaan: Yeah. And that second concern — it wasn’t the immigrants, it was the dorks building robots that took all your jobs — I do think that’s true. And it’s kind of a paralyzing thought, because it’s not just labor jobs. The knowledge jobs seem to be going first. Every two weeks it’s like, “Oh, video editors are out of a job. Photoshop is a useless skill now. Lawyers don’t need as many associates.” The breadcrumb trail is getting laid, and you’re like — what happens in V12 of this product? V12 is just going to do the job for doctors, for designers, for video editors, for everybody.
Shaan: It reminds me — they did this interview with Elon Musk and they said, “You’re really worried about AI, right?” He said, “Yeah, I’m extremely concerned about AI. It stresses me out, I lose sleep over it.” They go, “So why are you working on AI? Why are you developing it?” He goes, “Well, I think this is going to happen either way. It is inevitable. And then I just had to decide: do I want to see the AI apocalypse in my lifetime, or do I want to die and not get to witness it?” He goes, “I decided I’d like to witness it.”
Sam: Which is just hilarious because I thought he was going to say something noble like “I can use this for good.” Instead he’s like, “I’d just rather see it than not see it.” And then he says, “And then I felt good. I just continued on.”
Shaan: That’s kind of how I feel too. Like — am I really going to stop the progress of technology? No. Am I going to anticipate this now and change everything I’m doing? That doesn’t make sense either. Let me just make hay while the sun shines and see what happens.
Sam: I think when it comes to AI and robots, it’s like a squirrel preparing for winter. Between now and the next 10 years, the name of the game is just pile away as much as you can to get above a certain threshold, to make it through this winter that’s never going to end. Between now and whenever this moment comes — whether it’s 5 years or 20 years — the name of the game is just hurry up and board up your home for the storm. So you’re safe, you’re independent, you’re all right.
Shaan: I don’t know what the right mindset is, but that’s also mine — get it while it’s hot. I’m not blinders-on about this stuff. In fact the name of my original holding company was Inevitable Outcomes. I was like, well, why don’t I just figure out what is inevitably going to happen and invest in that? Once you know something is inevitably going to happen, you just have to figure out which company’s going to win. You’ve reduced the problem set down from “will this happen?” to “which company’s going to win, and when is the right time for a business like this to exist?”
More Robot Plays: Kernel, Pizza Bots, and the Figure Factory Visit [00:27:30]
Shaan: I started this episode talking about timing because I think single-purpose robots are going to be the big winners in the next cycle. You’ve got the knowledge work — LLMs, ChatGPT-type stuff — but then you’re going to see these specific robots. Both are going to be really successful. What you’re not going to see — I don’t think the humanoid robot is ready yet. That’s further out. Just my opinion.
Sam: There’s also the founder of Chipotle’s new restaurant concept that’s all robotic-first. What is it called?
Shaan: So the key insight that the Sweetgreen guy shared — he goes, “The key is not to retrofit the restaurant.” Everybody tries to take their existing restaurant and install a robotic arm in it, because it’s cheaper to not rebuild. His quote was: “Everybody’s trying to retrofit. I’m different. I’m willing to blow the whole thing up.” From first principles — if this was a restaurant that had no humans on the line and was going to have robots, how would you design every part of it? And then build that. They did two locations this year, they’re up and running, and now they’re doing 10 more to scale it up.
Sam: What’s it called, and what do they serve?
Shaan: It’s called Kernel. Steve Ells — the guy who created Chipotle — he’s got a new one called Kernel. It’s basically like some kind of vegan bowl concept. You can see videos of it where it’s dispensing stuff into a bowl and a robotic arm is just shaking the bowl to mix it for you. It’s pretty hilarious.
Sam: The crossover of vegans and apocalyptic AI people — I don’t think that Venn diagram is going to overlap nicely. Like, “Hey, do you want to go to this restaurant that has no humans working because, you know, screw people?” “Do they serve hummus?” That conversation is not going to go well.
Shaan: It’s plant-based and machine-operated. Who’s got that tattoo? Nobody. That’s not a huge crossover.
Sam: But it looks cool. So he has two or three locations. I don’t think this one’s as figured out or as promising, but there’s a bunch of people doing this. SoftBank poured hundreds of millions into a pizza bot that I’ve talked about a lot. Pizza is the big prize — it’s a simpler dish, single item, most popular fast-casual food. If you could build a pizza bot that makes a perfect pizza every time, perfectly circular, perfect ratio of ingredients, never calls in sick, works 24/7, consistent — even if that thing costs $500,000, you could sell it to every Domino’s, every Pizza Hut, every chain pizza shop. That’s the big opportunity.
Sam: And I like to think of myself as an independent thinker, but when I went to Brett Adcock’s factory and he showed me — he’s like, “Check this out, we’re learning how the Achilles tendon in a human works, and we just got our humanoid to move its foot in a dorsiflexion way. And over here our guy is making the knee” — I saw this stuff and I was like, “Can I just come here and join you?” I went home to my wife Sarah and I said, “I’m not smart enough, but will you quit your job and go work here?” I was just so into it.
Shaan: And the cool thing about this topic is it’s happening right now. It’s not “one day in the future” — it’s happening now, and over the next couple months and couple years we’re going to see it in play.
Sam: That’s also the magic of Silicon Valley. One of the reasons to be in the Bay Area is you will meet somebody, be at a dinner, hear something, drop by somebody’s office and see what they’re working on. And you walk out being like — my life is meaningless, everything I do is meaningless, I am a small pale dot on a giant rock. Most people find that disempowering. But for me, I wanted to be in situations like that. I wanted people who could blow the ceiling off of what I thought was possible, what I thought was cool. And be like, “No no no, you thought that was interesting? This is interesting.”
Sergey Brin Shows Up at the Hackathon [00:33:00]
Sam: That happened the other day. Did you see these videos that were going viral? Some people were at a hackathon in Sunnyvale or Mountain View, one of those places — the AI House or something like that. During the hackathon, this guy walks in with a cool jacket and disheveled hair like he’s been in a cryogenic freezing tube for 10 years. It’s Sergey Brin. The founder of Google just showing up at the hackathon. He gave an impromptu talk.
Shaan: What did he say?
Sam: He just got really into details about Gemini — bugs they’re fixing, how it works, how they’re thinking about it. Not some grand visionary talk. Just talking to a bunch of guys. And it went viral because there’s a guy in the audience whose shirt literally looks like someone’s body flesh with boobs. He asked a serious question — he said something like, “Gemini got in trouble for being too restrictive on generating images of certain groups, what’s your view on that policy?” And nobody called him out. Sergey didn’t point at him and say, “Can we get a different question? Not the guy with the boob shirt?” Nobody even acknowledged it. Just, yeah, carry on.
Shaan: Live and let live. Maybe that’s the San Francisco way.
Sam: Very strange place. But strange in a way where kind of amazing things happen — the founder of Google just shows up at your hackathon and answers questions. That’s pretty wild.
Shaan: I experienced this all the time when I lived there. I’ve gone to someone’s office and there’s just huge screens with something weird playing and they’re like, “Yeah we’re just working on this website” — and it’s normalized. Or I’ve come into the office and people are sleeping under desks. Just weird stuff that we’d make fun of or see on the Silicon Valley TV show.
Sam: I would go to these hacker houses where they’d have shared kitchens and half the time it’s just like protein powders and supplements — these guys just live off powders. That’s how Soylent started. That same guy — when a chip card came out that lets you tap your credit card to a machine — he inserted the chip surgically into his hand so he could just touch his hand to pay. Being around freaks like that is fun to make fun of because it is silly, but it is also awesome.
Shaan: I would see this stuff all the time and think — I’m so happy you freaks exist. This is the coolest thing I’ve ever seen.
Sam: Did you ever go to these hacker houses?
Shaan: Yeah. Just like triple bunk beds, they’re filthy. But you know what — damn, I can’t even look down on this, because one out of the 18 of you is about to become a billionaire in the next five years. I just don’t know which one. Whose neck beard is the longest? Who showered least in this room? I’ve got to guess who to write a check to.
Sam: I had a buddy at my company — Quinn — who had this crazy Sideshow Bob-style haircut. His girlfriend had beaten him at some word scramble game on iPhone. So he stayed up all night programming a robotic finger and a camera to solve the word scramble automatically so he would never lose again. It’s like the stories you hear about Kobe Bryant or Michael Jordan and their fierce competitiveness — but applied in the absolute nerdiest way. And it results in kind of amazing things.
Shaan: The first feeling is, “That’s stupid.” And then it’s, “No wait, I’m stupid. You’re better than me.” I feel that on a regular basis around these weirdos and I love them.
The Elon vs. Sam Altman Lawsuit: Framing the PR War [00:38:00]
Sam: We also need to talk about this lawsuit. So Elon Musk sued Sam Altman. There’s good analysis out there about it, but I want to point out a couple things. Here’s my frame: Elon is good at PR. He’s not just good at engineering, not just good at business — he’s also kind of a master of PR. Tesla spends basically nothing on TV commercials while every other car company spends billions, and yet Tesla’s brand is huge. He does this with Twitter, with Tesla, with SpaceX. And now seeing him do it with this lawsuit — he’s going to lose the legal battle but he’s going to win the PR war. I think.
Shaan: Have you ever been through a lawsuit? Do you know what the discovery phase is?
Sam: I haven’t been through one but I’m familiar with the idea of Discovery. So I was going to sue someone once — I thought they violated a contract and I’d lost money. I hired a lawyer and they said, “You’re probably going to win this, but you’re not going to win that much money.” I said, “It’s the principle.” They said, “Okay, but do you know what Discovery is?” I said, “No, tell me more.” They said, “We’re going to get access to your email, your phone records, your text messages, and they’re going to be able to see all of it. In many cases that becomes public record.”
Sam: And I thought about it — did I just text my wife the other day to let her know that I had a little digestive incident? I think I did. And I don’t want that in public. That’s like a three out of ten on the embarrassment meter. What’s my ten out of ten? Would I be okay with them seeing that? That’s when it finally hit me why people say “don’t write crazy stuff on text messages or email.” So I didn’t end up going through with it.
Shaan: In this lawsuit we’re going to see all types of private conversations. I think Elon doesn’t give a damn, but the average person would — including me. I don’t know exactly how Discovery works, what parts go public and what don’t. Because I think Elon has a lot to lose with that too. Nobody wants their stuff out there in public. But he does seem to operate like everything he says is something he wants the world to see. When he texted the former Twitter CEO and the guy replied asking to talk, Elon just replied: “What have you done this week?” Like in his mind, he’s like, I’m going to say stuff I want the world to see. It’s all part of the game.
The OpenAI Origin Story: Deep Mind, Larry Page, and the Gentleman’s Agreement [00:42:00]
Sam: So in this lawsuit there’s actually some stuff I found pretty interesting. The whole document starts off with something like — humans have transitioned from a labor economy to a knowledge-based economy, and now knowledge itself is getting disrupted by AI. First AI beat the best chess players, then it found the best routes between two points, then more complicated games like poker and Go. What it talks about is that Elon Musk saw this happening. He was on the board or an advisor to a company called DeepMind. He met the founder — Demis Hassabis, I think — and walked away with the feeling that this guy isn’t really concerned with AI safety, he’s just trying to build really powerful AI.
Sam: Then Elon found out Google was going to buy DeepMind. He’s like — Google has the most data because they have Gmail, Search, YouTube. DeepMind has a monopoly on AI talent. The result will be that AGI, when it eventually gets built, will be in the hands of the biggest private monopoly in the world. And so he goes to hang out with Larry Page, the founder of Google, and tries to warn him. He says, “Larry, I’m worried about AI. What are you guys doing about safety?” And basically Larry kind of pats him on the head. “Oh, don’t worry about that.” He says, “AI could take over, could eradicate the human race.” And Larry goes, “Well, that’ll just be evolution.” He goes, “Don’t be such a speciesist, Elon.”
Shaan: He’s like — yeah, I’m pro-human. Sue me.
Sam: So he gets really pissed about this. And he says that’s when he stopped talking to Larry Page. That’s the trigger, according to Elon. Again, I think a lot of this is Elon crafting a narrative. I’m not saying these things didn’t happen, but he is very good at creating an origin story, creating a bad guy, and positioning himself as the hero.
Sam: So now he teams up with Sam Altman. Sam Altman, who’s now CEO of OpenAI, in 2015 wrote the following: “The development of super machine intelligence is probably the greatest threat to the continued existence of humanity. There are other threats, but they are unlikely to destroy every human in the universe the way SMI could.” What an opening sentence, Sam.
Shaan: So Sam is worried about this, Elon is worried about this, they get together —
Sam: Right. And then Elon goes to PayPal co-founder Luke Nosek and tries to quickly emergency-raise $500 million to buy DeepMind instead of Google, just to keep it out of Google’s hands. He fails. He can’t pull the deal together in time. DeepMind gets bought by Google. Now Elon’s really worried.
Sam: So Sam Altman then goes to Elon and says, “Hey, why don’t we create a nonprofit AI lab to catch up with Google?” And actually — did you know the origin of OpenAI was supposed to be inside Y Combinator, where Sam was president? He called it — “What if we created the Manhattan Project for AI? What if YC did a call to arms, brought together the best people we can, started working on this as a research project?” He said, originally, “We’ll give them equity in YC as their upside, so they can work on this as a nonprofit.” And no single person would have control over whatever they made.
Shaan: How do you pay top AI researchers when Google is paying them a fortune? We’ll give them equity in YC — YC is our huge cash cow.
Sam: And he said: “We’ll have a board of directors — me, you, Elon, and then three others. If it’s not obvious what the right thing to do is, the five of us will decide.” That was the governance idea. That was the initial email. And Elon’s reply to that email was: “Agree on all.”
Shaan: That’s it? Just — “Agree on all.”
Sam: Just moved straight to the end. “Agree on all.” So they start, they’re like — cool, we’re making this nonprofit, it’ll be open source. Elon comes up with the name OpenAI. “Let’s put it in the name — this is open-source AI.”
Sam: Over the next five years, Elon contributes $44 million to this nonprofit. And they develop some pretty interesting tech. GPT-4 is in the 90th percentile at the bar exam, in the 99th percentile for the GRE verbal section, even 77% on the Advanced Sommelier Test. So in the lawsuit, Elon tries to argue it’s already AGI — which it isn’t — but his whole legal case rests on the idea that there’s been a breach of contract. They breached the founding agreement.
Shaan: But here’s the thing — there is no actual founding agreement. What he’s calling the founding agreement is a handful of conversations, some emails, and the initial Articles of Incorporation. He calls all of that together “the founding agreement,” but it’s not really a formal agreement. In those emails, sometimes they’re talking about giving equity in YC — that’s just brainstorming. In the Articles of Incorporation it specifically says things like “we will open source this whenever we deem appropriate” and “we will develop this for the benefit of humanity.” It’s vague. They basically left themselves complete latitude to do whatever they want. And that’s what they did.
Q* and the For-Profit Pivot [00:50:00]
Sam: Elon says a couple of other things. One — GPT-4 is already so good at those test scores. Two — now they’re developing this thing called Q*. Have you heard about Q*?
Shaan: No. What is it?
Sam: Q* is reportedly — rumored to be — the next thing OpenAI is working on, next level after GPT-4. Instead of GPT-5, they’re using some method called Q* which they think is a massive leap in intelligence. Even GPT-3 to GPT-4 were substantial improvements, but not a huge leap. Whereas Q* is reportedly the huge leap.
Sam: There were rumors where Sam Altman came out and said something like, “I was in a meeting the other day and I saw something that just — blew my mind.” And he hasn’t shared any details. This is also right when the big board coup happened — when they fired Sam Altman. One of the rumored reasons was that the board saw how powerful Q* was, got worried that Sam was moving way too fast, and felt this weapon was getting too powerful.
Shaan: So Elon talks about Q* in the lawsuit, which is kind of like confirmation that this thing exists. Although the legal language he uses throughout — he goes, “on knowledge and belief” before every claim. Which basically means: I don’t have hard evidence, but I know some things and I believe this to be true.
Sam: Right. A couple other big pieces. He also says — and this is really the strongest part of his case — how is it that you can create a nonprofit company that is supposed to be nonprofit and open source, raise millions of dollars from donors, take $50 million from me, hire the best people in the world, write down that this is your charter — and then end up a closed-source, maximum-profit company, 49% owned by Microsoft, the most valuable company in the world?
Shaan: And he goes — if it is possible to create a nonprofit, hire people, basically fund all the research with tax-deductible donations, and then flip to a for-profit when you make breakthrough technology — why is everybody not doing that? Why is that not just the default for every single technology company in the world? Either it’s illegal, or everybody should be doing it. It can only be one of those two things.
Sam: And I love this argument from Elon. This is a tremendous argument. There’s no answer to that question yet.
Shaan: We are — we’re fans of Elon’s work, but we’re lukewarm on him as a human being. And we’re only hearing one side of the story. But just from what I know in the last ten minutes — he’s right.
Sam: Then he gives this analogy — again, master of PR. He goes: “It’s like if I donated to a nonprofit charity that said we’re going to save the Amazon rainforest, and then that charity took all our money and created a for-profit logging company that was just cutting down trees in the rainforest and selling the lumber. That’s what’s happened. That is OpenAI.”
Shaan: And then he includes this quote — and I predict it’s going to become very problematic for Microsoft. This is the Microsoft CEO, during all the OpenAI drama when Microsoft stock was tanking, saying: “If OpenAI disappeared tomorrow, we have all the IP rights, all the capabilities, all the people, the compute, the data — everything. We are below them, above them, around them.”
Sam: The CEO of Microsoft said that publicly?
Shaan: Yes. And that also weakens the case that OpenAI is this independent, nonprofit, non-Microsoft-controlled entity.
Sam: That’s a phrase — “below them, above them, and around them.” That’s me when I go to Chick-fil-A and order nuggets. I am below them, above them, and around them. I have them in me. What a phrase.
Shaan: By the way — they called this the Manhattan Project. That is so good. All of this is just epic. Can you imagine these nerdy AI engineers who were probably bullied throughout school, and now they’re being fought over with $5 to $20 million annual packages? That’s insane.
Sam: Everything about this story is insane. You’re right — this needs to be a movie. If you’re making the movie or want to make it, get in touch with me. This is going to be like The Social Network all over again.
Recruiting Wars and Elon’s Paper Trail [00:56:30]
Sam: I want to read you some Elon emails that got surfaced in this. So one of the key things early on — they had this dinner with Sam Altman, Greg Brockman, Elon, and Ilya Sutskever. Ilya was one of the leading deep learning researchers at Google, and he became the chief scientist of OpenAI. The big turning point was when Elon recruited Ilya away from Google. That’s when Larry Page stopped talking to him — “you recruited my top guy, you declared war.” And recruiting Ilya was a herculean effort. Google was offering him a blank check.
Shaan: This is a movie. For sure this will be a movie.
Sam: Then Google started offering their other talent crazy lavish packages — trying to price everyone out of the market. Millions of dollars per employee minimum, some people $10-$15 million annual packages. So Sam and Greg email Elon saying, “Hey, it’s getting really hard to recruit. Google is doing this.” Elon replies — and I’ll read this directly: “We need to do what it takes to get top talent. Let’s go higher. If at some point we have to revisit what our current people are getting paid, that’s fine. Either we get the best people in the world or we will get whipped by DeepMind. Whatever it takes to bring on ace talent is fine by me. DeepMind is causing me extreme mental stress. If they win it will be really bad news, with their ‘one mind to rule the world’ philosophy. They are obviously making major progress, and they should, given the talent level over there.”
Sam: So he basically gives them the green light: I’ll keep funding, go get the best talent. In 2016 he gives them $15 million. In 2017, $20 million. In total, $44 million over five years.
The Strongest Argument and the Logging Analogy [01:00:00]
Sam: Now the problem: Elon, the biggest and original backer, says he doesn’t own anything in the for-profit OpenAI company. And here’s what he says — this is the strongest part of his case. How is it that you can create a nonprofit, raise millions of dollars from donors, take $50 million from me specifically, hire the best people in the world, write this charter as your founding document, and then end up as a closed-source, maximum-profit company that is essentially a subsidiary of Microsoft?
Shaan: And his answer — if that’s possible, if you can just start a nonprofit, fund research through tax-deductible donations, and then flip to a for-profit when you hit a breakthrough — why is every company not doing this? Either it’s illegal, or everyone should be doing it. He includes the Microsoft CEO quote saying “we are below them, above them, around them.” And then the logging analogy: “It’s like if I donated to a charity to save the Amazon rainforest, and they took my money and started a logging company cutting down trees and selling the lumber.” That is OpenAI.
Stepping Back: The Mission vs. the PR Battle [01:03:30]
Sam: I want to leave it with this. Dharmesh Shah — the founder of HubSpot, who we’ve talked to a bunch — told me something that I tell myself all the time when thinking about this. Dharmesh is super into AI, goes to a lot of insider conferences, has a good grasp of it. He said: “It’s probably not going to be as good or as bad as you think it’s going to be.” Him saying that in a calm, fatherly voice has been the one thing I keep coming back to.
Sam: Sam Altman and all these guys are brilliant marketers, brilliant business people, brilliant at building this technology. But they’re also brilliant at PR. I have to remind myself constantly: it could be good, it could be bad, it’s probably going to land right in the middle. And these guys are just excellent at telling their story.
Shaan: We should also say — Elon is creating a direct competitor to OpenAI. He has xAI, his AI company, and Grok, which is the competitor to ChatGPT. He obviously has interests in this. And I think one thing that’s clear is that it is mission-driven. The same way I used to think Jeff Bezos’s rocket company was just a billionaire measuring contest — like, I already won Earth, now let me win Mars — but then you read that even 25 years ago Bezos was obsessed with space, had been trying to donate to it and advance the cause, and now he’s in every engineering and planning meeting every week. It wasn’t what I thought it was.
Shaan: Similarly here — even though I think part of this lawsuit is Elon trying to bog OpenAI down and spin the PR battle while he has a direct competitor — it’s also clear that OpenAI would not have existed, would not have been successful, had he not genuinely had this fear and belief about what needs to happen with AI a long time ago. The paper trail shows it.
Sam: What a pod. I feel amped and scared and excited and nervous all at once.
Shaan: It’s like a great first date. You’re supposed to trigger adrenaline or fight-or-flight in people — that’s what bonds you. I feel like that just happened to us.
Sam: Yeah, except on our first date we went on a motorcycle, and now we’re talking about the world maybe ending in 20 years. Good job on this one.
Shaan: To folks listening — go to the YouTube and tell us in the comments: do you think this is going to be horrible, do you think it’s going to be great, or is it what Dharmesh said — somewhere in the middle, not as good or as bad as we think? Good pod. We’ll end it there.