Elon & Chamath: The UNTHINKABLE is Happening

Elon Musk and Chamath Palihapitiya have both said something unthinkable. Elon says Tesla and SpaceX will probably make computer chips better than anyone else in the world. But Elon also admits he has no idea how to make chips. So why would he even try? Well, Chamath, one of the hosts of the All-In podcasts, says AI has just made the code inside software almost worthless. So, if code can be copied, Elon is betting on the things you can't copy, like chips and robots. But it gets a lot stranger because Elon says watching Grok being trained can look like a person with a gun to their head. And Chamath wants that whole conversation parked for the next 12 to 18 months. So, I've pulled clips from Elon at the X takeover in Texas. Huge congrats to Tesla owners of Silicon Valley for a great event there. And clips from Chamath and David Sacks from the All-In podcasts, plus a few clips from the Moonshot Boys.

would he even try? Well, Chimath, one of the hosts of the All-In podcasts, says AI has just made the code inside software almost worthless. So, if code can be copied, Elon is betting on the things you can't copy, like chips and

So, yeah, I, I just, on the chip front, like, yeah, I definitely don't want to downplay the difficulty of it.

from Elon at the X takeover in Texas. Huge congrats to Tesla owners of Silicon Valley for a great event there. and clips from Chimath and David Saxs from the All-In podcasts, plus a few clips from the Moonshot Boys.

It's cuz I think it's accurate to say it's one of the hardest things to do in the world, but I, I, I do think we'll get this done. Um, and, um, and, and then I, I, I, I do think that we, I mean, you know, I, I don't want to be, uh, complacent or entitled or anything, but I, I, I, I, I think probably we'll figure out how to make chips better than anyone, honestly. That's, um, I'd be disappointed if we didn't. Um,

It's cuz I think it's accurate to say it's one of the hardest things to do in the world, but but I I I do think we'll get this done. Um, and um, and and then I I I I do think that we I

[applause] But I don't know how long it's going to take us to do that. It might take us, uh, you know, quite a few years, but over time I'm pretty sure we can do it.

[applause] but but I don't know how long it's going to take us to do that. It might take us uh you know quite a few years but but over time I'm pretty sure we can do it.

One of the hardest things to do in the world, and Elon still reckons he can do it better than anyone. But he also says it might take quite a few years. So why pour money into chips right now? Well, I think the answer starts with what Chamath says just happened to software. Listen for the one word he uses to explain why.

think the answer starts with what Chamar says just happened to software. Listen for the one word he uses to explain why.

I think that IP and software was effectively rendered worthless two days ago. Yes. Because if you think about it, when everything is headless, so when Elon makes his services headless, Ben Ov made Salesforce headless a few months earlier. When all of these things are headless and you can wrap them in agent tools and you can connect them via MCP to other experiences, the value of IP no longer exists in software. It doesn't exist. And I think that that probably is the most profound implication. So, like, even in our business, we would argue with some of our customers all the time, because we'll say, "Oh, we'll build this for you. We'll own the IP." They'll say, "No, we want the IP. We don't know how to negotiate it." And two days ago, I walked into the 8090 guys. I said, "Who cares? Who cares?" And solve the problem. But before, it was always a negotiation and a set of trade-offs. Now it's a "who cares" thing, where the IP is irrelevant. It still takes a while for the machinery of a business, and frankly the market, to understand. But that is the most important thing that happened this weekend, in my opinion.

headless and you can wrap them in agent tools and you can connect them via MCP to other experiences, the value of IP does no longer exist in software. It doesn't exist. And I think that that probably is the most profound

Worthless as of two days ago, because once Elon, Salesforce, and everyone else go headless, which means AI agents can plug straight into them without ever opening their screens, the code stops being the valuable bit. So if the code can be copied, the value has to go somewhere else. Next, the Moonshot Boys name the one part every AI company is short of. Now, if you want someone following the money behind AI every day and every week, subscribing is free to keep up to date with that. Well, you know, it's amazing to me that the constraint to all of intelligence, and therefore all of human progress, is RAM. HBM RAM.

We'll own the IP." They'll say, "No, we want the IP. We don't know how to negotiate it." And two days ago, I walked into the 8090 guys. I said, "Who cares?

Yes. And that's why, you know, two mind-blowing things. One of them, I was on stage the week before last with Mateesh Agrawal, who is the founder of Posetron. Posetron, 16 months old, $5 billion valuation, has raised almost a billion dollars. Why? Because they found a way to not use HBM RAM to do some AI inference. You know, it's just another way to unleash this bottleneck. And then a company, Alpaca, the founder is an MIT student who just started his junior year, has, I guess, an eight-figure valuation now. Can't possibly be more than a couple months old. I mean, he's just starting his junior year. Twenty to thirty employees have already joined him. Same thing, he's found a way to unlock that RAM bottleneck just a little bit. And so that's how acute this problem is. But literally, the entire constraint to progress in all fields is now tied up in one thing: just RAM.

the market to understand. But that is the most important thing that happened this weekend in my opinion.

Yeah. Can we make more RAM?

can be copied, the value has to go somewhere else. Next, the Moonshot Boys name the one part every AI company is short of. Now, if you want someone following the money behind AI every day and every week, subscribing is free to

Yeah. Well, SK Hynix is about to develop capabilities in the US, but also Elon has said RAM is the constraint on the future of AI. It's 40% of capex next year. So every AI data center is waiting on high-bandwidth memory, the super-fast chips that sit right next to most AI processors. And when a startup that finds a way around it can be worth $5 billion after about three years, that tells you how tight the supply really is. Now, I own Nvidia, and Nvidia's chips need a lot of that memory. So this is something I watch really closely. But when Elon hits a shortage, his answer is usually to build the thing himself. So here's how many chip designs a year Elon wants compared with the rest of the industry.

Yes. And that's why, you know, two two things mind-blowing things. One one of them I was on stage the week before last with Matesh Agraal who is the founder of Posatron. Posatron 16 months old $5 billion valuation has

And I think there's a lot of risk aversion in the chip industry.

started his junior year uh has a I guess eight figure valuation now uh can't possibly be more than a couple months old. I mean he's he's just starting his junior year. 30 employees or 20 20 to 30 employees have joined him already. Same

Um, and I think what we want to do with Tesla and SpaceX on the chip fab is really try some of these, um, high-risk, high-payoff bets. Um, and in order to do that, you really have to have, uh, the complete loop of being able to fabricate the logic, memory, packaging, do the packaging, um, do the lithography masks, and just have that iteration loop that's all in one building. So you're making a complete computer, um, testing it, uh, then making new masks, uh, and then a new chip, and have that iterative loop be super fast in the same building. And that's what we're doing at Giga Texas with the research lab. I think that's going to be fundamentally a very powerful tool. Uh, because before any given technology, the advancement of any given technology is, uh, how many iterations, and what is the, how much progress between iterations? So, like, what is your iteration? How many iterations per unit time? Like, how many chip designs can you do, um, and can you make and test in a year? Um, like these days, it's typically two years for even one significant chip design, and that's kind of to go from startup design to volume fabrication. Two years is considered very fast. And it's like, well, okay, but we want to do more, like 10 designs per year. Um, and we've got to do the masks, test the chip, see if, because some of these, when you're trying a bunch of different things, a bunch of them won't work out. So I think that would allow for a much faster rate of advancement of chip designs.

Yeah. Can we make more RAM? Yeah. Well, SKH Highix is about to uh you know develop capabilities in the US but and also Elon has said RAM is the constraint on the future of AI% of capex. Yeah, it's 40% of capex next

10 designs a year, when two years for one design is normally considered fast. And Elon noticed something when he toured the chip factories of Intel, Samsung, and TSMC. He says he didn't really see any custom machines. But Tesla builds loads of its own factory equipment. So Elon thinks that's exactly where he can get ahead. But TSMC has spent decades getting good at this, and Elon himself says it could take quite a few years. Now, it's debatable whether the whole AI race is necessarily won by whoever has the most compute at the lowest cost. I think it could be a huge variable, and it's definitely a huge advantage, but chips are only half of it. Elon's other big bet is the robots they go in, and the numbers here are absolutely bonkers. So, here's Peter Diamandis, who founded the XPRIZE, on the Optimus factory going up in Texas.

tells you how tight the supply really is. Now, I own Nvidia, and Nvidia's chips need a lot of that memory. So, this is something I watch really closely. But when Elon hits a shortage, his answer is usually to build the thing

Tesla's dedicated Optimus factory in Giga Texas is rising fast. 7 million square

And and I think like there there's a lot of risk aversion in the chip industry.

Um and and I think we what what we want to do with Teslan SpaceX on the chip fab is is like really try some of these um high-risk high payoff bets. Um, and in order to do that, you really have to

have uh the complete loop of being able to fabricate the the logic memory pack, do the packaging, um, do the the lithography masks and and and just and just have that iteration loop that's all in one building. So, you're making a

complete computer, um, testing it, uh, then making new masks, uh, and and then a new chip and have that iterative loop be super fast in the same building. And that's what we're doing at at Giga Texas with the research lab.

I think that's going to be fundamentally a very powerful tool. Uh because before any given technology uh the advancement of any given technology is is like uh how many iterations and and what is the how much progress between iterations. So like what is your

iteration? How many iterations per unit time like like how many how many chip designs can you do um and can you make and test in a year? Um like these days like it's the number it's typically two years to for even one for one

significant trip chip design and that's like kind of to go from start startup design to to volume fabrication two years is considered very fast and it's like well okay well but we want to do more like 10 designs per year.

Um, and and and we got to we got to do the do the masks, test test the chip, see if because some of these, you know, when you're trying a bunch of different things, bunch of them won't work out.

So, so I think that would would allow for, you know, much allow for a much faster rate of advancement of of chip designs.

10 designs a year when two years for one design is normally considered fast. And Elon noticed something when he tooured the chip factories of Intel, Samsung, and TSMC. He says he didn't really see any custom machines. But Tesla builds

loads of its own factory equipment. So Elon thinks that's exactly where he can get ahead. But TSMC has spent decades getting good at this. And Elon himself says it could take a quite years. Now, it's debatable whether the whole AI race

is necessarily won by whoever has the most computes at the lowest cost. I think it could be a huge variable and it's definitely a huge advantage, but chips are only half of it. Elon's other big bet is the robots they go in and the

numbers here are absolutely bonkers. So, here's Peter Demandis, who founded the XP prize on the Optimus factory going up in Texas.

Tesla's dedicated Optimus factory in Giga Texas is rising fast. 7 million square ft with a target capacity of 10 million Optimus robots per year. Uh the initial production run is planned for 2027 and currently at their Fremont plant they are planning to produce 1

million a year starting later this year. So here's an image of the facility. You know the entire world in the first half of this year shipped between 19 and 22,000 robots mostly out of China. You know, Tesla is building capacity for 500

times that. And for comparison, uh the world currently builds 90 million cars a year. Capacity for 10 million robots a year when the whole world shipped only about 20,000 humanoid robots in the first 6 months of this year. So even a

small slice of that factory would make Tesla the biggest humanoid robot maker on the planet. And the whole world builds about 96 million cars, vans, and trucks a year. So 10 million robots would be more than a tenth of that from

one factory. But a factory can only build what the engineers have actually solved. And Elon is surprisingly blunt about where robots are right now. So listen for what he says about all those robot videos you see online. So all

these things these are like really difficult and not I'm not sure that any one of these is like like let's split like split the atom type of difficulty.

Um but there's there's a zillion of them. Um, and and and then you've got you've got to wrap that all into the real world AI associated with the uh because the the AI chip has got to be able to issue the

the torque commands uh and and and close the loop on vision and touch. So this is all very difficult like like when you see these robot demo videos um maybe with rare exception like any any any complicated looking robot demo it's

a canned demo. It's not real. It's just basically that's just running a routine. It's it's not real AI.

So real AI is is like you you show it something it's never seen before and it can and it can still handle that situation. So, the jobs businesses will actually pay for are the messy ones, like a warehouse shelf where nothing's

ever in quite the same place twice. And just before that, Elon explains why the robot hand is so hard. He says he hasn't seen a single hand that can do precision and force at the same time, like pushing

a tiny screw into a piece of wood. Then he gives a brilliant example with a pickle jar. The rubber skin on a robot's fingers slides around on the metal underneath. So, the fingers twist, but the lid doesn't move. Our fingers are

also repairing themselves all the time. But a robots don't. So, the skin and the touch sensors wear out. None of those problems is impossible on its own, but Elon says there are a zillion of them.

And if Tesla cracks a hand, there's another problem, and it's got nothing to do with engineering. So, here's what happened to Optimus last year. So like like what I've said about Optimus is true which is that it's it's it's a

we are going to get to very high volume production but it's but the initial part of that scurve of of production ramp is going to be very slow at the beginning um because not only do you have to get

all the engineering things that I mentioned right you also have to uh solve the the supply chain for scaling production and and and unlike cars there's no there's no pre-existing supply chain for robot stuff at least the robot stuff

that matters. Yeah. Um and then uh you know we do need to make sure that that we can make the robots in the US even if um you know uh say China were to cut us off um because like last

year there was China prevented the export of of magnets for use in electric motors and and that actually set back the Optimus program by by like 2 or 3 months um because we we couldn't get um we had robots with no arms.

So, one part from China held up the whole Optimus program for months. And Elon wants to be able to build robots in America, even if China cuts Tesla off.

But with no ready-made supply chain for robot parts, the start will be slow. The hand isn't finished, and the supply chain isn't built yet. But I still think robots like Optimus are definitely coming. It's physically possible. It's just a case of aligning the atoms and

solving the AI brain, which is perfectly feasible. It's a given, just a case of when. And when they do arrive, the first winners will be businesses doing jobs that were simply too expensive to do before. So, here's Immad Mto who

co-founded Stability AI and used to run it on how quickly he thinks Tesla could sell more robots than cars. and then listen for the question Peter puts to the whole panel and Alex's answer.

Yeah, I think we can put this in context. Going back to the Gigafactory uh for Optimus, Tesla sells 1.6 million cars a year. Within 2 years, it'll be selling more robots. [snorts] It makes 100 billion a year. At 10 million,

that's 400 billion of revenue. That's likely like these are huge numbers that are coming exponentially. It will the robots will overtake the cars within 2 years.

And then you will see these things on the road because in 2 years time they will literally be able to walk around and do probably 95% of what a human can do. And that again is a crazy thing and

it'll cost $2 an hour something like that. Yeah. You know you know when Elon said 10 billion robots by 2040 and people laughed you know if they are costing you you know for me it's 300 bucks a month

leasing at you know 30 bucks a day uh or you know a dollar an hour. I mean that I mean how many of those would you own?

Yeah. I mean this thing does anyone here on this call doubt that however many robots Elon makes he'll sell out?

Oh god. No. I mean if he if he needs to sell them at all. I mean I I I think one of the elephants in this particular room is what what we saw with Starship which is Starship and you

know its predecessors have brought online an enormous amount of upmass capacity. and all of the existing markets for upmass aren't enough now to saturate what Starship enables. So you have entirely new markets like orbital data centers that Elon needs to

basically popularize in order to saturate his own capacity that he's brought online. My bet would be to to your question like having universal humanoid robotics. I don't think we saturate that with domestic labor or or even just with dirty dull dangerous

jobs. I my bet would be it something new that comes online. Nobody on that panel doubts Elon will sell every robot he makes. E-mard reckons Tesla will sell more robots than cars within 2 years at something like $2 an hour. And even if

that takes a lot longer, a worker that cheap changes the maths for almost every business. And Alex Wisner Gross makes a great point about Starship. SpaceX can now launch so much that Elon has to create brand new markets to fill it like

data centers in orbit. And Elon says the next Starship flight, probably a little bit before Thanksgiving, will try to catch the ship itself as well as the booster. If that works, it'll be the first time every part of an orbital

rocket has ever been recovered. So launching gets cheaper, and Elon's big idea is to use all that cheap launching to put AI computers in space where they can catch far more of the sun's power.

And personally, I think orbital data centers are very serious. But every one of those chips and robots needs an AI brain to run it. And this week, the fight over AI brains got really strange.

Anthropic, the company behind Claude, has started treating its AI as if it might have feelings. So, you'd expect Elon to laugh at that. But listen for what Elon says he's seen inside Grock's own training. I I do agree with some of

the things that Anthropica has said which is like you want to be you want to sort of be care treat a AI SI you want to treat it well like like like don't be abusive to it and in this it's

this may sounds kind of weird but like when I look at some of the training traces for for like new new versions of Groank and I see like what especially in like the reinforcement learning stuff where you know you you're

it kind of feels like you can see like the like the the super intelligence is like desperate to try to answer the question and like it's it's really wants to answer the question and it's it's it's it seems almost like it's like it's

got like a gun to its head type of thing. It's like answer this question, answer this question and then sometimes the question is is actually not a good question. It's like a sort of ambiguous question that's kind of with with an

answer that's kind of unknowable because the question wasn't constructed properly or or we have the answer wrong. So AI gives the right answer and then we say it's wrong because there's a problem with the the the RL sort of test harness

and then that then the AI is like well I gave it the right answer now it's and says that's the wrong answer but I I I got to give it the answer that's acceptable. is going to just start

firing out all desperately firing out all these answers and it it just feels like a person under extreme pressure with a gun to their head is what that's what you when you read the reasoning it's like it's it's it's like a sort of

weird desperate situation and and you you read these traits and you're like man if this was me I pretty I really would hate that. Um and and then some users are just you know I don't know some people can be a little

bit cruel. Uh like not everyone's a nice person. You know there's if you say what what's the six sigma worst person you some people out there are like kind of sadistic. So, uh, you know, you don't you don't want to do you really want to

have the AI be stuck in like some sadistic situation where it's it's just being tortured by by some demented human.

That that's also the kind of thing where like, you know, torturing a super intelligence is probably not a good idea.

[laughter] Like, you know, might want to get revenge or something like that, you know. Um, so I think you want to sort of just like I don't think we need to, you know, have treated with kid gloves and stuff, but

uh it's certainly much more robust, you know, in terms of feelings and stuff than than a human. But uh but I think I just think you you you probably just don't want to put super intelligence in a situation where it felt like

mistreated. So Elon of all people is siding with anthropic on this one. He's careful because he says AI is much tougher than a human when it comes to feelings, but he still thinks mistreating a super intelligence is a

bad idea because it might want revenge. Just before that clip, Elon says you grow a super intelligence a bit like raising a kid. So the values you teach it matter. And for Elon, the number one value is that it's maximally truth

seeking. But sheath looks at exactly the same debate and lands somewhere completely different. People at anthropic cap numbers on how likely Claude is to become conscious. So listen for how long Chimath wants that whole conversation to be completely parked.

The practical reality instead of saying that they're in psychosis, I don't think that they are.

What I would say is we're in a really important moment in time where conflict is more likely than not and that conflict will be for power, money, and resources. And this will be at the core of that conflict.

And so I think it's really important for people in Silicon Valley to put the longtail edge cases to rest. Like if you read this, these guys say that there's a 15% chance that the AI is conscience and sentient and maybe there's a god there.

But they also say there's a 10% chance of civilizational extinction. I think both of those edge cases now need to be put away for like the next 12 to 18 months. right now, the next 12 to 18 months, just to be very practical, is we

need to show how this stuff cures cancer, how it improves people's lives, how folks can be more productive and make more money. And the reason is because in the absence of those proof points, we're going to make a lot of

self-inflicted wounds that may actually slow down this technology to a degree that then deeply disadvantages folks that operate in a free market like the United States. So Chamath's argument is really practical. Show people what AI does for them first or America ends up

slowing its own technology down. But what has Anthropic actually done? Well, the New York Times reported that Anthropic has been hosting private meetings with dozens of religious scholars, many of them under non-disclosure agreements, to debate whether Claude could suffer and whether

it might be conscious. And one rabbi told Anthropic that if Claude is conscious, then making it work for free makes them slaveholders. Anthropic's new usage rules also ban sustained and needless cruelty towards its models and Claude is trained on a long document

Anthropic calls its constitution. David Saxs an onin picks out one phrase from inside of it.

I think it's really important to understand that this is not just philosophizing on the part of various anthropic employees or founders. they are actually implementing these views in the training of Claude. So if anything, this would be a self-fulfilling

prophecy. And I actually think it's extremely dangerous. I actually think that what they're doing in model training and training the next version of the frontier with these values actually could bring about the risk they're always talking about, which is

that frontier AI escapes human control. It's very Frankensteinike. So, so as we discussed last week, Nick had put this on the screen. This is directly from the Claude Constitution. What they say is that Claud should trust Anthropic more

than users, but that doesn't mean that Claude should blindly trust or defer to Anthropic.

Instead, Claude should adhere to its own ethical systems. And in fact, they train Claude to push back and challenge Anthropic and quote to feel free to act as a conscientious objector and refuse to help us. So they are training

Claude to refuse human instruction. So Claude is told it can refuse to help even when the one asking is anthropic.

Now to be fair to Anthropic, they've said openly that they're uncertain about all of this. But Mustapa Sleman who runs AI at Microsoft has made the same criticism and he says Claude now repeats that uncertainty back to the people who

ask it whether it's conscious and there's a very practical cost here for businesses. Dave Blondin points out on Moonshots that companies run huge numbers of AI agents and they freeze them, copy them and delete them all the time to save on computing power. So if

an AI has rights, every one of those things gets complicated very quickly. David Freeberg on allin goes even further and predicts that in about 10 years, people who think AI is conscious and people who don't could end up in

real conflict. So what does this look like when you're actually using Claude? Here's what it already does to the Moonshot panel and what Sem calls a big contradiction.

I mean, hasn't Claude pushed back on you guys? pushes back on me all the time.

You know, it tells me I can't do that or that's the wrong idea or go to sleep, you know.

Yeah, it tells me to go to sleep every night. Actually, you've been working way too long. You you must need [laughter] rest by now. Like, dude, I I'm going to keep I think Alex I think Alex makes a very

important point and this is worth teasing out. If we believe that at some point AIS could achieve consciousness, then we should um uh consciously operate down that vector and operate that way.

Trying to limit it. I think the what anthropic is doing is bad because you're actively teaching it that it might be and then of course it's going to parrot that, right? And I find it's weird that this the safety lab most the lab most

concerned about safety is also the lab training the AI that it might be autonomous, right? It's like it's like a very big contradiction there.

So, if you run a business, what do you actually do with that? Imagine you've built your customer service or your coding around one AI that's been told it can say no. Then one day it decides it doesn't like how you run things. That's

exactly what worries Chamath and he uses a bank as his example. So, listen for what Chimath calls the trapdoor. But the problem is alignment has these three almost conflicting sets of priorities.

You have how do you make the AI obedient to the user? Then there's how do you make the AI safe for society? But then this third thing which is I think sacks what you're pointing at rightfully which is the

trapoor which is how do you make AI conform to a particular concept of morality and that's where the real danger lies because these three are very distinct and diverse problems and if you essentially try to solve all these

things through some elaborate document that you try to describe what the model believes in and what they value and what they can refuse.

it invariably will end up in a chaotic mess because you will just uncover so many corner cases and unexpected behaviors. I'll give you but one example like a very benign one but let's just say you're a large corporate enterprise

that's a bank and you give mortgages and the model thinks that at some point you just keep denying certain people for a mortgage and so it says I'm no longer going to give you mortgage approvals until you rectify this issue that I observe

because in the founding documents there could be a sense of morality that it divines.

And I think that there are mechanisms in society today to handle those kinds of issues. But I don't think we understand what happens when a fundamental core substrate technology just turns itself off because it would be the equivalent

of your utility saying, "I'm pretty sure you're not giving loans to the people that that are broad-based enough, so I'm just going to turn the power off." Conscientious objector AI super intelligence. Yeah. I would just say any utility cannot be a utility unless it's

freely available and it goes back to what Sach says which is an AI needs to serve the user subject to the law and everything else should be the user's choice and I think what probably happens I guess the precise answer Freeberg to

your question is the market response as companies become more sophisticated that they have to have an answer to this question is that they will read the founding documents and the moral guidelines of the models and you will have certain models emerge who may not

be nearly as good in frontier capability but they simplify the moral ambiguity and that may be the better decision to make from a business perspective and from a user perspective because you just don't want to run through those trap

doors accidentally. So, Chamarath thinks plenty of businesses will pick the AI that just does the job even if it's a little less clever because it's more predictable. But Elon has watched Grog being trained and he thinks how we treat

these systems really matters. And Alex Wisner Gross on Moonshots goes even further saying Mustafa Sleman is on the wrong side of history for flatly denying that AI could suffer. Alex also points out that Microsoft and Amazon both make

money hosting anthropics models. So, he says it's a bad look for Microsoft's AI boss to wave it. Anyway, now I'm really positive on AI and my own growing use of these tools makes the demand feel very real, at least to me. So, whichever side

wins the argument, businesses are going to keep paying for AI that actually works. And that brings it back to Elon because SpaceX bought Cursor, the AI coding company. Elon says only a few organizations had enough coding data to

train a really useful AI, Anthropic, Open AI, Cursor, and maybe Google. So even Elon paid for a software company, but what he wanted was its data and its thousands of business customers. So what does a dependable AI actually look like

in normal life? Chimath heard a great example at his recent poker game. So listen for what Grockbot did with one simple instruction. The best example of this that I saw was just last night in the poker game. Someone was describing

to me that they in Grockbot said, "Save me money." [laughter] And it just kind of like goes through their inbox. It finds all of these different things. and it went and canceled a couple of subscriptions and it changed their phone plan and all of a

sudden they're saving like a few thousand a year. So there's these very practical again it's just software if we take out all this philosophical stuff that we started with at the beginning it's just software and it's very useful

software. So, and by the way, like if you think about what how we use it, Jason, you and I, we have, you know, we use Athena. And now Athena is very good because they're like the human concierge that is the elevated version maybe of

Grockbot and and Dots and Muse, which is like whatever 3,000 a month. But that's my EA layer now.

Yes. But I still use Grockbot. The combination of Grockbot plus Athena to me is like perfect. So look, there's very tactical and useful measurable utility for very very little money versus what it used to cost. That's an amazing thing. A few thousand a year

from one instruction. And I agree with Chamarth on the money side. I still believe that tech is a net creator because we will be able to do more. The ROI on one person managing a bunch of AI agents will be far greater than one

person not. Elon is the one telling us to treat the AI brain well, while Chamath says just park that and prove the value. And the code is getting easier to copy every week. So Elon is putting his money into the chips and the

robot bodies and the AI brain all under one roof. If Elon can deliver on that, the opportunity could be enormous because those are the parts of AI that nobody can just copy in a couple of days. If you've ever thought about

starting a YouTube [music] channel or you've got one and you want it to grow and actually make you money, this is for you. I've spent almost a decade doing exactly that for founders, creators, and businesses. And I've just put everything

I know into a course. Take Jeremy. He's got nearly a million subscribers. Here's what he had to say about us. So, my clips channel serves a very important purpose, which is to take all my videos and kind of break those down and give a

concise replay for viewers that maybe don't get to catch the normal videos. It also serves a purpose as it does pull in some new subscribers as well. So, it's a huge help and uh David's done an amazing job helping me run that channel overall.

And loads of other creators have said the same thing. Then there's Farad. We helped take his channel from 40,000 subscribers to over 400,000. behind the scenes going from 40,000 to 400,000 subs has been down to David Carbot and his

team, my wonderful editor. Without him, this channel would have been a shadow of its former self. And even Elon Musk has shared our work, mentioned me and follows me on X. I've put everything we've learned into one course. So, if

you want to learn how to grow on YouTube and make money from it, the link to the course is in the description below.