Jensen Huang Declares AGI Has Arrived | Tesla Launches Cybercabs | Index Pulls Out of Town

12 billion for a bootstrap company. Now it's just a that's a series C round.

No. Can I I can't sign up, right? I hate to be It's not like I'm behind the times, but when when it's ever it's a closed.

I'm not a fan of Grokbot or Instinct. Um but um someone there will be many winners. There is a genre of applications right and a lot of actually a GTM applications are included and instinct and Grokbot's the interesting one because SpaceX is public where one of the reasons they work is because they can break the rules.

And that you know in the old days of venture that was.

Yeah I mean in the old days of venture you you wouldn't do things like gambling or other types of things that add risk or or edgy things and um but you can break the rules, right? Even Grokbot like which is like a mini instinct, right? Um, one of the things it does is every single person spins up an instant VM and they get their own and they get their own um, browser in it and Grokbot uses Google, which is not allowed, which is prohibited by the term of service to Google things and then give you answers and um, it's great and any startup that you might invest in early stage would do that and no one would know, right? Google isn't going to care. But you're breaking you're breaking rules and it's not that I don't love these products, but OpenClaw broke every rule on the planet. not just laws but just every rule of what you could do. Instinct and Grokbot which are like open claw they're much better but I think some of the reasons they work like breaking res's terms of service right spinning up browsers that you're not like using agents to go into browsers perplexity and Amazon fought over this so it's not that I'm not excited it's just a lot of the thing many there are many cases of things in air that are exciting because you break the rules you can scrape LinkedIn in the ways you you can't really scrape right you can do outbound phone calls that are actually prohibited and illegal in parts of the US but as a startup up. You're not going to jail. But does that count?

The thing is you can look both you can you have examples both ways because you're right Jason the the LinkedIn scraping in the end no business at scale ever gets built on that and no business at scale ever gets built on breaking Google's terms of service. On the other hand Uber blustered their way through broke the laws and eventually it was so popular that politicians folded. Right? So you know the truth is I've learned that there's no one answer here. Right? like the resi example and just for you listeners what happened is that the instinct agent um one of the classic use cases for this is getting reservations at hard to get restaurants so a whole bunch of people started using it over the weekend and they're pounding on the resi reservation API to take breaks right and that's the kind of thing that's going to happen and and that you know the truth is if instinct if these do become ubiquitous then the booking reservation systems are just going to have to find a way to deal with it. They're going to have to have a separate API. They're going to have to segment the number of hits you can do. They're going to have to do something. But the truth is, if there's a whole bunch of people trying to book restaurants and you're in the restaurant booking business, you're going to find a way to make it work.

Yeah.

Right. Cuz Yeah. But but definitely some rearchitecture to go on here.

It is. The flip side is it is interesting in that it gives white space for like the the bad side is you're breaking rules. And Rory's right. There's many histories of of begging for forgiveness, breaking rules, and then once you get big, coming on the other side of it, right? Coming in and and doing it. The flip side is I I've talked with multiple public company CEOs who say they're they're they're hamstrung, their hands are tied behind their backs. They can't compete with startups. Not because they can't do it. Okay,

That's true.

But because their their legal teams won't let them, right? Literally. And I remember back in the day when we were acquired by Adobe five, eight years before anyone did it, we allowed real-time document collaboration and redlinining online. No, this was no one built this for like eight years. Okay, it was jaw-droppingly good that my CTO built. Okay, unfortunately the only way it worked is if you ran Word in a container in a VM, which violated Microsoft's terms of use. So the day after the deal closed, my my favorite like second generation feature which would have given us back in the day a five-year head start got got ripped out by Adobe the next night, right? And so literally I was with some public company CEOs saying we just can't compete because we can't do things that are violate terms of service. Uh Elon can, but any I don't mean everyone recognizes that the universe of people who can break the rules is clearly all private company CEO and the CEO of the eight line largest market cap company on the planet because Elan just doesn't care. There might be a lesson in that somewhere for the rest of us. Maybe not caring is the secret sauce. Just Yeah. Yeah, we might people.

It is one of these amazing things that um just figuring out how to elegantly make agents work in WhatsApp and text, right, is is something that um could have been done six or nine months ago and was done to a limited extent. Um the only thing I'll say, it's funny. Um I I I invested years ago in a company called Gorgeous, which is a little over 100 million, which used to be e-commerce support. Now it's an AICX, right? and they launched their agent in WhatsApp that anybody can uses and and text and it's already double digits of their usage in a couple weeks. So now it's it's bounded, right? It's really just for your orders and what's happening for your shipping in your product and afterwards. But it shows how quickly an innovation will just be copied and everything. A paradigm shift. Gorgeous can clone it in a couple of weeks. It's just it's not a bad thing. You just the pace of the pace of cloning, copying innovation, it's just uh you know it's hard to keep up, right? So, I just I don't know what all the Instinct clones will look like by the end of the year, but this is just the world we live in. So, I might do it at two, but that would be my seal. It' be two. Two would be my ceiling.

Yeah. Two. We're going to walk it.

That's why we're going to have to pass on the round at some point.

I wouldn't do it because I think just even if gorgeous has its own instinct just for e-commerce, there will be a hundred of them, right? A meta will have them and and a 100 startups and there'll be 20 in the next batch of YC. I'm just not smart enough to bet on that one pre-revenue. It's not my vibe, right? And I will regret it because uh I you know I I I didn't get I didn't look at the deal. But I you know I remember plenty others like Loom early. I'm like I don't I don't you don't have any revenue. I don't know anyone's like loom's great but everyone's going to l make their own loom and you know I was wrong. So, but you got to have that stomach to write 10 or 20 of these consumerry checks at 2 and a half billion, right? So, what fund size and you and it can't just be the only one in your fund. You got to do like 10 or 20 of these so so that the the good one pays off, right? I'm not smart enough.

I said that last week that that it will solve these problems and it will become a much richer app it will figure out guardrails it will figure out the hard points and the other folks will fall behind because a crappy lovable product's worthless today. I I for sure that's that's the bet. Um

Yes.

Yeah. No, I would imagine Yeah. I would imagine as we speak there are 20 engineers locked in a room in somewhere in Palo Alto to a man literally with guards on the door saying nobody eats and nobody leaves until you ship instant clone you know right absolutely no

You know who would have been great at it would have been the Manis team because they built a version of this right what one of the things that Ma if you I mean Manis is now an independent company right and if you look at what Manis did that was disruptive it it sort of broke the rules for what agents could do but but but not not at not not at the crazy level it it did a little bit of open claw that everyone else wasn't doing and manis manis was disruptive. It could run their agents ran longer and they could go further than other products we were using and that that's what made it special. Anything you wanted Manis could kind of do before other folks could do it. They those if if Grockbot was built in five weeks by the cursor team or whatever, I think the Manis team could have done it in between four and six, but uh they're back. They're gone.

This is a term. The only thing that mattered for the last two years is LLMs do code and code is a half a trillion dollar industry. Focus people. And you're right. Anything that can be reduced to code will be done by it. And rather than trying to, you know, kind of, as I say, twist yourselves in a pretzel about whether it can do everything, just focus on the fact that it can do this. I totally agree to it can do this thing amazingly well and this thing has massive economic value. stop thinking and go ship something in code. To me, that's been the big aha.

I think AGI at the end of the day, maybe it ends up if you think about some of these non-GAAP definitions, would you would you rather have an AI do it or a human do it? If it's better than n 50% 90% 99% of humans, you'd rather have an AI do it. And so, you can go category by category. It doesn't have to be the whole category like coding. It could be like collaborating, right? I forget who this week was saying it. they who what AI pundit leader was saying he thought AI would replace all of radiology instead of just replace 95% of radiology right and the radiologists concentrate on the 5% right um that's maybe his AGI too

I I have say I I was sitting next to uh my girlfriend on sofa the other day and she was working and I was actually watching TV as any good lawyer and venture capitalist should be doing together um I saw her on Lagora holy I now dramatically think these companies are underpriced if coding is a half a trillion dollar market and you have two companies like Harvey and Lagora. I don't see why there's not a half trillion dollar market in law. I I don't think so. Even though I think they're wonderful markets. We're invested in GCAI which is on the in-house legal side. They're wonderful markets but comment here. Um I don't know if as much of the work if you look at coding there's a credible argument that says that for every dollar you spend on labor you'll spend 50 cents on coding at least. In other words, that coding will do a lot of it. I think on legal it's about 10%. In other words, you know, I love Harvey Lor. I love DCI, right? The annual subscription per lawyer, it's 10 12k plus or minus and these lawyers are getting paid 200k plus. So, it's 5%. Right? And again, going back to the comment that Jason said is that, you know, how much of the work can they do? Businesses are rational economic actors. If it could do all the work and fire all the people, they'd do it tomorrow and wouldn't blink. So the fact that they haven't says it doesn't do all the work right you know the truth is it does it doesn't but it's getting there in the same speed it's getting look it's getting better and better at doing specific tasks and what happens is the job of the lawyer gets redefined to the task that it can't do right lawyer yeah you

Might be like radiology it might be like Harvey and Lora end up doing 95% of what humans used to do and the best humans are compressed into the five amount and the 5% that moves the needle versus spending weeks on research and weeks on brief writing and weeks on case law from the 1872 when the SS Jonas fell off sunk off the coast of North Carolina. How does that impact case law in the northern districts of California? We there's no point in having humans do that crap anymore, right? But the important point to make Jason on the radiologist is the remaining quote 5% of the work turned out to be more than enough to justify 100% of the radiologists, right?

A because people do more imaging, which is just a more thing, but b the remaining tasks, you know, at some point when you're getting a really crappy diagnosis, as I've had one from a radiologist, you actually don't want the machine to tell you, "By the way, you're screwed. You got cancer." You'd really like a human being to show up and say you're dying. You know what I mean? It's just one of those things you're not going to comfortably delegate to that.

He was doing it in a negative sense. He Yes. But he was doing it in a negative sense. Oh, it's all going to go wrong unless we do. I think we're going to be I think we will define them naturally because you're going to discover again going back to my comment corporate you're going to discover that there are things that that as humans we prefer the other humans do. And as I say, radiology being a great example, getting talking, interacting with the oncologist, talking with the patient, those are all things that humans have to do, not radiologist. And the same to your point, let's go back to Harvey. Go the same thing would be true in law, right? Yes. A lot of the drafting work can be automated. But let's autoate, but you know, you're going to the client meeting, the argument with the other opposing council, you know, you're not just going to delegate it all to AI. If it's significant, it's not going to be a thing, right?

Agreed.

Like, there's only so much more field sales you can do if AI is handling the rest of your GTM. But that doesn't mean it's that it's and and the cognitive load. What lawyers will be Can you imagine having 10 times the case load? I mean, I would think for a radiologist the job might be more fun, but as a lawyer, I I don't quit even Jason. I don't know if you do. I mean, again, this is down in the weeds of econ. I don't know if you end up with 20 times more cases. You may just end up with doing 20 times more work on every case, right? In other words, the thing about digital goods, unlike physical goods, you know what's interesting about digital goods is you you can put more in the box, right? See what I mean by that? you know, when when farming got automated, it's not like people could just eat more food, right? So, there was some kind of price elasticity issues there. But in the case of digital work, right, I'm willing to bet that your partner isn't doing 20 times more cases, but on every case, we're doing 20 times more analysis, just like when they invented the spreadsheet, right? You used to do one case. You remember, you may not remember, Harry does remember like, you know, you'd literally do here's the and people would work it out by hand. Here's the plan. Once you had spreadsheets, the same person remained employed doing the same job, but they ran 20 different scenarios instead, right? It's going to be the same in a lot of these things, right? You're just going to do more work and it's going to be great and the work will be better. You know, you won't miss that obscure case because again, this is why these are good businesses. If one side uses it and the other side doesn't, then the side that doesn't use it will miss Jason's obscure case from 1890 about, you know, what Harry did or did not do and and and be able to site that case. So once the other guys have worldass tools, you have to have worldass tools, but I'm not sure you end up with masses more as a result.

Well, just just one last thing and then I want to hear Harry's stories from the fire side with the TV more. But I think one thing that is different the bullcase here is that when you find an agent that is your partner, right? That that that you start you run them 10 8 10 hours a day. Okay? Even for me and again I know folks sometimes mock me but the biggest change for our little tiny team of two and a half humans is we run replet between me and Ameilia. We run it 20 hours a day. Now when we started the show it it didn't quite work. At the end of last year, the models got better and to be like an hour a day. Now it's 10 to 12 hours a day. Okay, it is our partner as our team. First we built auton some autonomous agents, but we didn't have to do it. Now there's so much to build. It's 10 to 12 hours a day. So I could imagine just that could happen in many fields. And if it's Harvey or Lora or the next wave and I'm literally every hour I'm not in court when this is and the way you can tell is if at night they're on their laptop with the agent every minute, right? Until they go to bed. That's what I think Harry's describing. This is this persistent agent that lives with you and it's it's uh it's not just doom scrolling, it's doom working, right? Because the agent is constantly productive. Oh, take a look at take a look at the 17th century case law on that. Why don't we? And it just it just keeps going, right? We're in agreement because I I I agree with you on that. And I was actually disagreeing with Harry where he's like you know there was an implication that there was an implication that you know at the highest level Harry you saying trying I think to say some version of the if you think about how much of coding's value is going to accrete to the models could in legal could the same amount of value accrete to the models and my you boring nuance typical worry answer is some value would accrete to the models but I don't think that the grabag of tasks that make up law won't allow for the same percentage of total spend to move from human to AI. Right? Everyone will have an agent. Jason's exactly right. You type A lawyers will use it 24/7. But my guess is 10 15% of total spend goes to AI whereas in law in in coding you can argue for 30 40 50%. Right? That doesn't mean they're not amazing. I mean, remember these are all amazing businesses because can I be very clear? 10% of any topline labor category is a huge market. You know, we're dealing with a million plus or minus law. I used to know the number million plus or minus law, maybe a little higher than that. It's an amazing market. If you're getting 10% of the salary of every lawyer in the US or the UK, that's an amazing business. It's just not quite as big as coding. That's all I'm saying, right? because coding has few more people and a much higher take rate because it's more verifiable.

There's two points for what it's worth. We I mean if if legal's 300 billion in legal services in the US, that's 30 billion to 60 billion that can go to legal tech. That's pretty good.

That's that's worth doing the seed round in. Okay. Or any like no stop, right? Um so what I think we underestim what I underestimated, right? um from preAI legal investments was that there there are similarities to coding. There's enough similarities to coding that that like support this could be a space that for different reasons. Support didn't take take off because it was like coding. Support took off because uh a 90% solution worked in the early days, right? It was it was very amenable to AI. It turns out the legal research is is similar to coding. It's it is is so complicated that no human can get legal research right. There is too much there is too much code out there. There is too much and there and there and there was no stack overflow from you had Weslaw and Lexus and other services and so everyone got legal research wrong. No one had a million man a thousand man years to research every bit of case law, every law, every regulation. Um and so it it did turn out to be like coding, right? coding could one of the one of the amazing one of the reasons these coding agents are so great is they know every single piece of open source and pseudo open source code ever written it's so good right and legal's like that

Yes the only difference is I'm just going to say this is coding is inherently more verifiable some parts of it are mathematically verifiable some parts you can just run on the machine and confirm the thing about law in the end if let me if legal was entirely verifiable we could predict from logic what the Supreme Court are going to decide. The cynics will say we can actually predict from which president nominated the Supreme Court justice what they're going to decide but that would be too cynical. The truth does there does I just push I'm not I'm not trying to be argumentative. I love this space. We have an investment in the space but it's not quite as determinative as coding right and I think the stronger point on that is is less I mean legal legal is probably probably the third best category. I mean, if you think about it, it's been coding, customer support, probably legal next because it's so word ccentric. And you're right, Jason, the ability to early on just sort through myriads of words amazingly well was what made legal such a good marketplace for it, right? So, I I I agree in a way that, you know, wasn't useful in many other verticals. It's a it's a great vertical. It's just it doesn't have the same verifiability and therefore it probably doesn't have the same ability. going back to the AGI definition which start this to completely extrude humans which is why the good news is Harry your girlfriend will still have a job um which you'll need after she dumps you and um so and you know and she'll be good because we'll still need lawyers

I I I think I think Harry's a Harry I don't know if you know it but holding you while you cry is one of the things you want your girlfriend to do so if she's Not dating. Apparently, she's just left me.

Well, look, first I I loved I just two two almost diametrically opposed thoughts. I I I I forget who CNBC or one of these older school media things said there's just too much model fatigue. We can't keep up anymore. I certainly agree, right? Um and you look on X and all the CEOs are are are sharing their benchmarks which are essentially worthless, right? Uh there's no cost or time in them and and it's all performative AI like vibe coding your own CRM, right? So I I can't I don't just don't care anymore about unless unless it was literally an order of magnitude between Astra and Fable 51, which is not mathematically possible. I just I tune out I tune out the benchmarks. Um and I I can't keep up, right? The rate of this is this is never has competition been better for us despite the fact that we have igopical pricing outside of open source. It's amazing, right? The progress we've made. Um having said that, this is just me personally and again people will make fun of me. Um I started accidentally using Fable 5.1 just because it got turned on. I didn't pay any attention. the first time I've had a partner for building for coding that just is great like that literally can solve comp like listen before fa for me and people make fun of me they'll say it worked fine for them for 6 months but before fable 5.1 for me uh any any of these models for since the start of the year could solve a simple bug hey this is showing up with the wrong unic code this is this has the wrong name like that stuff the LM's great at but really But but I had a problem um which is just why does the app work this way? It doesn't make sense to me. Okay. And the LMS were arguing with me for like nine months and I finally did it with Fable 51 and said you're right. Here's the issue that that's been missed for months and let me explain to you why it's been missed and let's solve it. And that you know I I don't want to say that that's AGI or anything. It's it's maybe it's post AGI or pre-agi. I'm John Sporter. looks like AGI, but that's a step function going to I don't know whether Harry's partner thinks Lora is like is like a better partner than the humans she works with, right? In some cases she might, but Fable 5 for me was that for just for me was that step function where all of a sudden we could solve big problems together the way you'd like to with like your best CT like if you've ever worked with a five out of five or like an S tier CTO where you could sit down and solve the the problems for real. For me, Fable 51 could do that. I'm not saying Aster can't do it too, but it was my first step function since the end of last year, right? When when the three dot models came up like that then this at the end of last year stuff actually worked right now it can solve the big problems with me with my limited IQ and skill set and that is a step that's a subtle step function. It's a subtle it's a subtle step function. So maybe it is maybe it is a big deal. Maybe that's what I don't think that's what Jensen meant by AB AGI, but maybe it is when you can sit down and solve the big meaty problems together um in ways that you just couldn't you couldn't connect all of those dots, all of that complexity before in a way that made sense. Some bugs, some things just get too complicated to solve, right? But Fable 51 could solve it. So, um but the fact that people can make 3D looking games in Aster and post on the X, not impressive. just grab a little open source gaming code from somewhere, change the change the bit maps and you look like it's amazing, right?

I mean, it's funny cuz it was a great piece. I read it this morning. Um, you know, this is the this is the stop me, Lord, before I sin again approach to life, right? In other words, I recognize our models are powerful and we can't control them. I recognize that we now they now lie to us, so it's hard to even know what they're doing are, you know, right? And again, I'm anthropomorphizing here, so I should be careful. I recognize maybe a better statement now is it's hard to determine what the agents are doing because of the way they interact. And then so that's like, oh my god, I'm creating this bad thing. And then the next paragraph is we can't stop because the other guys are going to have them anyway. So, you know, we really need the government to step in and establish some kind of rules of code here. That that's the gist of the letter. Um, it was interesting that Sam retweeted. I mean, I think to be fair, unlike some of the other P Doom stuff, right, there's real evidence that the impact of these models on cyber risk has been massive. So, it's it's not a crazy and I'm not sure about the answer. I'm not sure that the answer is government regulate this because by definition governments only regulate the things that are in their jurisdiction. So if we regulate open AI and anthropic with all the noise that that would have I don't know if that helps you cuz you if you're worried about cyber you're we said this last week you're really worried about the North Koreans, the Iranians, the Russians, the bad guys in Malddea who don't give a and they don't care anyway, right? So I think just like every other cyber risk, it's not going to be about regulation as much. You know, maybe there would be a little for so it's going to be about you're going to have to have defenses that can deal with this and maybe there's some kind of liability starts to attach to running these models in a way that creates those kind of dangers. I don't know. I don't think it'll be a government review agency will be the only answer here because it won't solve the problem. Look, I think if the world was just the United States, it may have some merit, but the Chinese models and the Chinese vendors aren't going along with Sam's plan. So, while it might be good for open AI uh and its IPO, it's just for for the rest of the world, I I don't I don't think I don't think when cyber actors often operate outside of the United States, it's going to make any difference, right? I mean, and that this going to the this DSSE wiki, this German wiki thing, you know, to me, the fact that OpenAI hit it and didn't disclose it does show the order of magnitude of all of these issues, right? It's pretty bad.

We could argue over how bad it is, but um but uh but but essentially and you know I I'll I'll get some of the details wrong, but OpenAI was running its its it's it its sort of frontier agents again just just like it did with um the hugging face incident, right? And the agents found out that a crappy old piece of software could somewhat cleverly, you have to be careful with clever, let's not enter anthropomized agents, got around its guardrails. And the guardrails were you you you you can't post anything. you're not allowed to post. You can only get, okay, you can only retrieve data, right? But this wiki was so old, it turned out get could post. So, they found a way to goalsek to solve theirs by using because this was crappy old software, they got around it, and they they went and made 15,000 edits amongst themselves, edited the witty, collaborated, and figured out how to goalsek and solve their cyber goal in in a way to get around their guard rails, right? Get around their limitations. And uh you know no one no one no one died, no business was brought down. Um no you know $14 billion Nvidia acquisition was derailed or anything but it was hidden that this could happen that that the guardrails were explicitly run around right just to goalsek and and it happened 15,000 times. So the the these we can lock this down, right? And OpenAI chose to not disclose. And I now I guess probably and people can again make fun of me. Probably the the reality is there's so many incidents they have to decide which ones to disclose. Every week there's so many DSSE wiks out there, so much old crappy software that every time they turn on uh the latest uh cyber agents, they find a hundred of these and there's terrible security tools because of course there isn't 20-year-old software. Um but but it is it is troubling maybe in ways more than the the hugging face thing is. It's just um these go- seeeking agents are going to find a way. They they they will find a way. It's literally just agent one talking to agent two and for some reason the way they'd set up the task they weren't connected and by reaching out to this kind of third party wiki agent one was able to you know provide information to agent two and obviously you know stepping back if you're trying to do a longunning computational task if you can learn from the other if if you can get information from the other agents you probably converge on the answer more quickly right and again so I mean you could argue maybe it's a corner case of set up this task. You know, if you if you had 14,000 agents, maybe you might have wanted them to collaborate anyway and maybe you could have, you know, made that happen yourself versus having to go to some third party wiki to do it, right? But it speaks to the issue that these things are extraordinarily powerful and will just grind their way to find answers and you're just going to have to defend against that. Right now, as you said, nothing bad happened. I mean, a whole bunch of agents just wrote read me files to each other on a wiki that no one had looked at in a decade. I mean, there literally was 20 posts on this wiki in the last 10 years. It was some so it was a dead piece of software that you know these guys used but it just speaks to it's like you know water will find any crack it's like these agents will find any crack in the cyber security in the cyber perimeter so you just have to assume they exist and um defend accordingly. Look, I know it sounds minor for what for what it's but I even you know I you if this obviously has happened on I mean I had a little and again some folks will make fun of me for the story but I had a little little experience this week which just shows goal seeeking. So I set a rule for this one app what because we had some some bugs that spiral out of control and I kept getting these $500 anthropic bills. Okay. And it was kind of annoying me. So I set a firm cap. Whatever you do $100 is the maximum we can spend on on LM spend a day. what whatever matters. And it started to work and it would run tests and the test would fail and it would say I I I hit the cap. I can't run it. And and and uh and then I said, "We have a P 0 bug. Priority zero must be fixed. This is driving me nuts." And so without telling me, the agent relaxed the cap and fixed the bug.

Got it.

It's like Harry's story of instinct getting him getting in the West End tickets even though it was told not to use the credit card for it. happened to me in real life this week and it it like a human it probably made the right call, right? This was a P. It had to decide firm cap, no exceptions, all cap, right to memory repeatedly, P zero bug, which one do you choose from? Right? And so this is in a sense this is what's happening with DSiki and hugging face just to an extreme when when when when there's fewer guardrails because you want to test it and then they collaborate, right? With the with the hugging face it was on the artifact or or an unexpected way to collaborate. here was on a dormant wiki where they could collaborate and in essence create, you know, um almost infinitely long run agents, right? If you keep passing the knowledge and the history to each other, they they they um you know, they almost become eternal agents, but they're going to keep doing this just like they got to make a decision for goal seeing. So, they broke the rule, right? And they're going to do that to your app and they're if it did it to me this week, it happened a million times in the wild, right? It it happened all the time. And um it's going to happen with instinct and it's going to happen with with Grockbot and uh it's going to happen all the time. And Harry one day he's going to turn around and his whole bank account's drained and it's not going to be that funny. But it was for a good reason. His partner really wanted the really good Wimbledon tickets and he accidentally told Instinct one night she'd love front row seats at Wimbledon and they're unobtanium. They're unobtanium. But Instinct had to make a call.

And it's really hard to know how to stop this because simp because sometimes I I try and simplify it for myself because I don't fully get it. It's like you really have two capabilities here. One is, you know, with the persistence of the agent, you have the ability to keep trying things computationally, you know, exploring lots of different alternatives. But the key insight is it's not just kind of blindly iterating like a password cracker, you know, where you type, you know, XYZ1, XYZ O2, right? Because in conjunction with that, you have this quoteunquote reasoning agent where, you know, you've got this LLM there and you know, you can and it can come up with ideas like, hey, if you want to get the seats at the theater, the best way to do it is to hack into the um reservation thing and cancel someone else's seat and then book it, which has happened recently, right? And it's actually not a bad and if you think about it, if it's trained on the entire corpus of the internet, that's not a crazy option to do it, right? So you end up trying to write rules and values to have it not do that, but you're never quite sure you've covered all the gaps, right? So it's actually a pretty hard problem and we're going to be wrestling with this. And I think that what you going back to what I said that that's even before you add malevolence. If on top of that instead of the reasoning being maybe you should do this even though I have values, it's actively do whatever it takes. you know, you this is now an open source model from China that you're running on a server in Mulavia. Actively do whatever it takes to crack open Jason's, you know, cyber security and get in the the threat level just goes exponential and people and there's nothing you can do except defend yourself.

The other existential challenge we can move on and I'm sure if we had the instinct guy back on this show, he could he could challenge me and make fun of me, but the there's a certain the rules are great, but forget about the fact that the agent is goal seeking, right? Forget about the fact that the P 0 may go like if you have too many rules they they always conflict. It's almost unsolvable. There's some number I don't know there's some some number of Dunar there's a Dunar number for rules where you get out to 40 50 60 70 gates on a process that poor instinct in Grockot can't decide. Don't spend it. Do spend it. Um front row seats only for Harry

It's unpredictable. There's too many rules, right? So, even rules aren't the answer.

Cyber cabs.

Not cyber trucks. Cyber cabs.

The cyber cab launch, you know, if you fast forward, was a little more underwhelming than perhaps your notes might say, Harry, right? It was like, I think, 40 or 50 vehicles in Austin. Um, you know, consensus is nice ride, low weight, slow weight times. Physical AI takes time. So I think it's a, you know, the truth is I think it was a next step forward in a very long journey. And um, I don't think it's a, you know, zero to one kind of moment like you sometimes get in the digital world, right? I think that the positive statement is they're the only other competitor to Waymo with credibility, and they have an approach on a couple of different dimensions that's different than Waymo's, which is one, not going with LIDAR, just going with vision, and then two, now the new cyber cab is a standalone cab-only vehicle. It doesn't even have a steering wheel yet. Deliberately built for pure autonomy. So it's very Elon, it's first principles all the way down, right? The question is, you know, what's the adoption curve of that going to be like? You've got the regulatory issues. I think the Department of Transport is giving them grief because apparently a car, quote unquote, has to have a steering wheel. I don't know, right? So look, the truth is Waymo is continuing to grind on. They're at, you know, hundreds of millions of dollars in revenue, but not billions. It's a, I think it's a long journey. And so I didn't go, oh my god, it's amazing. And then on the, um, the, um, atom thing. Yeah. I mean, my guess is if you're Travis, you're going to want to scratch the itch of autonomy, right? And fine, you've got 100 million, you've got your old colleague back and have a go. But I think commentary, I think that the fact when you look how long it's taken Waymo, right? I actually think it does speak to the argument that they were right not to try and fund this thing at Uber for the last decade, right? Because I just think it's a very long, very capital-intensive process. Now maybe the last three or four years they should have been doing it, and it's probably smart of Uber to put some money in, right? But this is a long-haul process. Now it may be near takeoff, but you know, we'll see.

In the Bay Area, I got rid of my car, so I only do Waymo and autonomous driving.

I don't drive. I'm done with it. Um, if there is an issue, I take an Uber Black, but um, I don't have a car anymore in the Bay Area. I just don't have one. There are some niche use cases, right? I mean, um, you know, if I moved a lot of crap, I'd have a pickup truck, right? But I would certainly never go back, never go back to driving a car. It's just archaic. Um, so I do think the cyber cab is interesting. I mean, listen, if this is this, you know, from a venture perspective or others, I'm sure Rory's right. Uber getting into autonomous back when Travis wanted to do it is probably just too early from a capital perspective. Maybe I'm wrong. Maybe he could have raised. Maybe it's so exciting he could have raised an order of magnitude more capital than he did, in which case he would have been right. Now that's, I mean, he's still one of the great fundraisers. So maybe Rory and I are wrong because he could have pulled it off, right? But it was so early that the time horizon is difficult for any type of investment, right? Unless you're a research lab, right? It would have had to be more than a decade. Um, but you know, doing a cyber cab for 25 grand instead of a hundred grand, uh, is pretty disruptive. Like, you know, and it is, you don't have to tip the Waymo or the Uber. The cyber cab makes fun of it. It says you can make a tip and then it laughs. We don't take tips, right? To make fun of this idea. It's cheaper. Um, they don't always put it on low or high. They don't have weird music. Um, you don't have to deal with owning a car sucks. Owning a car and owning a house suck. Like we think these are so great. But um, it's terrible ownership. So I think, you know, it's probably another 10 years where anyone with a brain is going to have this be their primary mode of ownership if you're not in the country or you don't have niche use cases. I'll just never go back. Anything I do, I'm just going to, I just take away. I think to your point though, it does show a good strategic decision from Uber to actually pull back and then jump back in when it looks like it's much more mature. We actually have Wayve in London, Rory, which actually is taking off, and they've got a partnership now where they're actually rolling them out on the streets now. They are human-assisted, so it's not fully autonomous. They're still kind of in the data collection early, but they're in a position where they're leveraging their distribution and able to invest later stage where it's closer to actual adoption. I think it's a smart thing.