THIS WEEK IN AI: Nvidia Acquires HuggingFace, Leopold vs SEC, New Waymo Chip

We cover NVIDIA’s record earnings, strong GPU demand, and its position in the AI chip market, along with reported moves involving Hugging Face and open source AI. We also discuss Z.ai’s new model, Waymo’s autonomous vehicle hardware updates, SEC scrutiny of Leopold Aschenbrenner, and recent developments in robotics and Apple’s upcoming event.

------
🔒 Check Out Our Sponsor: LEDGER AGENT STACK 🔒
https://developers.ledger.com/docs/ai-tools/overview/?utm_source=Audio&utm_medium=Podcasts&utm_campaign=Limitless

------
🌌 LIMITLESS HQ ⬇️

NEWSLETTER:    https://limitlessft.substack.com/
FOLLOW ON X:   https://x.com/LimitlessFT
SPOTIFY:             https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQ
APPLE:                 https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890
RSS FEED:           https://limitlessft.substack.com/

------
TIMESTAMPS

0:00 NVIDIA’s Record Quarter
5:05 Hugging Face Joins NVIDIA
9:26 Cloud Margins and the Chip Play
11:36 GLM 5.3 Flash Emerges
15:40 Waymo’s New Hardware Leap
19:48 Tesla’s CyberCab Challenge
23:24 Leopold Under SEC Scrutiny
25:44 Anthropic Usage Looks Flat
29:27 Agents Need Better Security
30:20 Robotics Learns From Video
34:07 Apple’s Big September Event
36:36 Sam Altman’s Rare Watches

------
RESOURCES

Josh: https://x.com/JoshKale

Ejaaz: https://x.com/cryptopunk7213

------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures⁠

Josh works with Anthropic as a contractor. All views expressed are his own and do not represent Anthropic, its leadership, or its affiliates. Nothing in this episode is investment advice.

Episode Transcript

Ejaaz: The very same company that just got hacked by a rogue open AI model just got (00:00):
undefined

Ejaaz: acquired for $13 billion by none other than NVIDIA. (00:04):
undefined

Ejaaz: Jensen Huang believes so strongly in owning open source models in America and (00:09):
undefined

Ejaaz: he's made his biggest purchase yet. (00:14):
undefined

Ejaaz: But that's not without cause. NVIDIA just reported their earnings and they had a record quarter. (00:16):
undefined

Ejaaz: They earned almost $100 billion in just three months and the company is growing (00:21):
undefined

Ejaaz: 100% year over year. This is unprecedented for the most valuable company in (00:25):
undefined

Ejaaz: the world worth over $5 trillion. And every single bank has raised their price (00:29):
undefined

Ejaaz: targets for the NVIDIA stock. (00:34):
undefined

Ejaaz: But that's not all in this week's roundup. We had a new secret model that revealed (00:35):
undefined

Ejaaz: to be GPU's new GLM 5.3 flash, which competes at the level of Fable and OpenAI (00:39):
undefined

Ejaaz: GPT 5.6, but at a fraction of the cost. (00:44):
undefined

Ejaaz: We've got Leopold updates, unfortunately, of the bad kind, where he's being (00:47):
undefined

Ejaaz: probed by the SEC, where they've subpoenaed a bunch of banks for accusations of insider trading. (00:52):
undefined

Ejaaz: And we finally have an update on robots from the worlds of Waymo, (00:58):
undefined

Ejaaz: autonomous driving, and Tesla. All that and more on this week's Roundup. (01:02):
undefined

Josh: To start with the NVIDIA quarterly revenue, I want to give you a sense of how (01:06):
undefined

Josh: outrageous this company is. (01:10):
undefined

Josh: NVIDIA's quarterly revenue now exceeds the annual revenue of roughly 480 of the S&P 500. (01:11):
undefined

Josh: That is a crazy statistic, how big it is. NVIDIA is printing about just over (01:20):
undefined

Josh: a billion dollars in cash every single day that it operates. (01:25):
undefined

Josh: So the day you're listening to this, NVIDIA is printing a billion dollars of (01:28):
undefined

Josh: revenue. And that number is going to go up. They're projecting huge amounts (01:32):
undefined

Josh: of growth over the next quarter. (01:35):
undefined

Josh: And they're guiding now for 70% year-over-year growth in 2028 relative to the (01:37):
undefined

Josh: expectations, which were 44%. (01:43):
undefined

Josh: So this earnings report was about as good as you can guess. (01:45):
undefined

Josh: And wall street's kind of rewarding them now i mean the stock is up it was down (01:49):
undefined

Josh: after trading but it is now up seven percent (01:52):
undefined

Josh: early market we're looking on screen now at all of the banks that have changed (01:55):
undefined

Josh: their price target after reading this earnings report pretty much every single (01:58):
undefined

Josh: one of them has raised it significantly higher with the highest being four hundred and twenty dollars (02:01):
undefined

Josh: per share and for like to grow that much as a five trillion dollar company is (02:06):
undefined

Josh: a meaningful amount of revenue and i think this is very high signal that (02:11):
undefined

Josh: The market's still in a really strong spot. NVIDIA is basically saying like, (02:15):
undefined

Josh: hey, guys, we know we have a lot of competition. We know everyone's trying to (02:19):
undefined

Josh: build their own thing. In fact, Jensen was on CNBC yesterday and he was asked (02:21):
undefined

Josh: about OpenAI's Jalapeno chip. (02:25):
undefined

Josh: I don't know if you saw the CJ's. It was very cool because Jalapeno is a really (02:26):
undefined

Josh: great model and they are a really great chip. And they came out and said, (02:30):
undefined

Josh: hey, this is actually more efficient than the NVIDIA chips. (02:32):
undefined

Josh: And Jensen was like, dude, we're the kings here. We've been doing this for 33 (02:35):
undefined

Josh: years. You've been doing this for three months. (02:38):
undefined

Josh: Will talk to you when you've actually produced a chip, when you've put it into (02:41):
undefined

Josh: data centers and when you've scaled this thing. (02:44):
undefined

Josh: He does not seem concerned. The market really liked that. And now it seems like, (02:46):
undefined

Josh: I mean, NVIDIA is really off to the races and there's no end in sight. (02:50):
undefined

Josh: Everyone wants the chips. (02:53):
undefined

Josh: They are sold out for as much as they can make them. They're basically operating (02:55):
undefined

Josh: fully at capacity and they have a very clear path to dominating more and more of the industry. (02:58):
undefined

Josh: They're growing bigger by the day. They now are working with five of the largest (03:02):
undefined

Josh: banks in the world to help fund the build out of more data centers using NVIDIA GPUs. (03:05):
undefined

Josh: And it's this like unbelievable business. The entire world of AI is centered (03:10):
undefined

Josh: around this one company. (03:13):
undefined

Josh: And the earnings report is saying it's going to be like this for a while. (03:14):
undefined

Josh: So buckle up because we are up only for the foreseeable future. (03:17):
undefined

Ejaaz: I just want to show everyone a chart which basically shows NVIDIA's revenue (03:21):
undefined

Ejaaz: breakdown. This is unbelievable. (03:26):
undefined

Ejaaz: Over the last, you know, a year or so. And it's just absolutely exponential and crazy. (03:28):
undefined

Ejaaz: Now, I want to give a bit of background as to why this is happening with NVIDIA, (03:32):
undefined

Ejaaz: especially recently. Why do they continue to earn more money even though everyone's (03:37):
undefined

Ejaaz: accusing them of, you know, building up this massive bubble? (03:40):
undefined

Ejaaz: Well, there's a few things that are happening. (03:42):
undefined

Ejaaz: The most obvious one being is they've just raised the prices of all their GPUs by 15%. (03:44):
undefined

Ejaaz: Now, granted, a lot of that is because the cost of memory has absolutely soared. (03:49):
undefined

Ejaaz: So, you know, those memory stocks have also been very volatile recently. (03:53):
undefined

Ejaaz: NVIDIA has to pay the most for memory to build their own GPU. (03:56):
undefined

Ejaaz: So that's a main reason why they're projecting more revenue in the future. (03:59):
undefined

Ejaaz: That may not necessarily correlate with the profit side of things. (04:02):
undefined

Ejaaz: The second thing is, you know, your point around jalapeno, which is OpenAI's (04:06):
undefined

Ejaaz: new custom AI chip, which they're building themselves. Typically, (04:09):
undefined

Ejaaz: they buy from NVIDIA, but now they're building their own selves. (04:13):
undefined

Ejaaz: Some people see this as a potential threat, you know, like Google's TPUs are (04:16):
undefined

Ejaaz: to NVIDIA's kind of moat. (04:20):
undefined

Ejaaz: And Genset's response is simply, we own the moat for generalized GPUs, (04:21):
undefined

Ejaaz: for generalized AI chips. (04:27):
undefined

Ejaaz: That means any AI lab, whether you are a Neolab, so a smaller AI lab, (04:29):
undefined

Ejaaz: or a large hyperscaler or whether you are anthropic or open AI can use NVIDIA (04:33):
undefined

Ejaaz: GPUs for whatever type of model. And by the way, it's not just LL apps. (04:38):
undefined

Ejaaz: It's financial models. It's fine-tuned models. It's models that can help make (04:42):
undefined

Ejaaz: scientific breakthroughs. (04:45):
undefined

Ejaaz: Jensen owns that entire mode. When you compare it to what OpenAI has built, (04:46):
undefined

Ejaaz: Jalapeno, that's known as an ASIC. It's a very specialized thing. (04:50):
undefined

Ejaaz: They have to prove that out at scale. We haven't even seen these things in operations. (04:53):
undefined

Ejaaz: We're going to see it by the end of the year. So NVIDIA's response to this is (04:56):
undefined

Ejaaz: we still very much control 70% to 85% of the entire market, and we aren't going anywhere. (05:00):
undefined

Ejaaz: Now, the second bit of news on NVIDIA that I'm most excited about is this acquisition, (05:06):
undefined

Ejaaz: Josh, the $13 billion acquisition of this open source platform called HuggingFace. (05:11):
undefined

Ejaaz: Now, for those of you who don't know, So HuggingFace is basically this platform (05:16):
undefined

Ejaaz: which hosts open source AI models. (05:20):
undefined

Ejaaz: So typically when you are an AI lab, and you can be an AI lab, (05:23):
undefined

Ejaaz: if you launch an open model, you can deposit it or upload it onto the HuggingFace platform. (05:28):
undefined

Ejaaz: It's where everyone goes to check out the latest open source model. (05:33):
undefined

Ejaaz: And you can do a few things when you're on this platform. You can upload your (05:36):
undefined

Ejaaz: own data sets to fine tune different open models to kind of create your own version. (05:39):
undefined

Ejaaz: A very good example I like to use is Fujipu. (05:43):
undefined

Ejaaz: You know all their GLM models? They have about, I think it's like 10 official (05:46):
undefined

Ejaaz: models, but they're collectively around 155,000 GLM models that exist on Hugging Face. (05:50):
undefined

Ejaaz: And some are financially tuned for financial trading. Some are financially tuned (05:57):
undefined

Ejaaz: for research and all these other different niche use cases. (06:01):
undefined

Ejaaz: And you can go up, sign up on an account and get access to it. (06:03):
undefined

Ejaaz: Now, I ask myself, why on earth is Jensen paying 50X the amount of money that (06:06):
undefined

Ejaaz: Hugging Face earns to acquire this open source model? Like, doesn't he just (06:13):
undefined

Ejaaz: care about Anthropik and OpenAI? Well, the answer is simple. (06:17):
undefined

Ejaaz: Jensen cares deeply about open source because he believes in a world where it's (06:19):
undefined

Ejaaz: not just going to be Claude or GPT running the world. (06:24):
undefined

Ejaaz: It'll be hundreds and hundreds of very niche and also very broad models that (06:26):
undefined

Ejaaz: run the world. And people can collaboratively use different models at different (06:31):
undefined

Ejaaz: times. Agents run the entire world. (06:34):
undefined

Ejaaz: Why this is a bull case for NVIDIA and Jensen in particular is. (06:36):
undefined

Ejaaz: Guess what? They all need to run on his GPUs. So he's been investing heavily. (06:40):
undefined

Ejaaz: He invested in Poolside, I believe, last week for $9 billion to acquire 100 (06:44):
undefined

Ejaaz: of their employees so that he can build new open models. (06:49):
undefined

Ejaaz: He's investing tens of billions of dollars in Nemotron, which is NVIDIA's own (06:52):
undefined

Ejaaz: open source model, over the next couple of years. (06:55):
undefined

Ejaaz: And now he's made the third and biggest purchase of his entire kind of like (06:58):
undefined

Ejaaz: focus for open source. So this is a very specific strategy that I think the (07:01):
undefined

Ejaaz: banks are rewarding. And that's why they're raising their price targets as well. (07:05):
undefined

Josh: It's amazing to see all these acquisitions happening we have open router now (07:08):
undefined

Josh: we have hugging face going hugging face is essentially the github for open source (07:11):
undefined

Josh: ai it's where everyone goes to build these things it seems like there's (07:15):
undefined

Josh: three reasons they would do this and there's three reasons why nvidia has kind (07:18):
undefined

Josh: of been very adamant about moving to open source the first being what we just discussed (07:22):
undefined

Josh: open ai is building their own custom chips google is building their own custom (07:27):
undefined

Josh: chips amazon's building their own custom chips everyone is trying to build (07:30):
undefined

Josh: alternatives to the nvidia gpus it's very difficult to do but they're working (07:34):
undefined

Josh: on it so open source is a hedge in a way against its own best customers where (07:37):
undefined

Josh: it allows them to find more use cases more (07:41):
undefined

Josh: need and demand for tokens that can be generated in the case that they lose (07:45):
undefined

Josh: some of that market share to these other large companies the second is like (07:49):
undefined

Josh: open weights seems to very much be where the volume is going we're starting to see the trend (07:52):
undefined

Josh: Somewhat ossify in the sense that open weights models and open source models (07:57):
undefined

Josh: are producing a lot more tokens because they are cheaper to use they're mostly (08:01):
undefined

Josh: effective enough to get the job done for most people and (08:05):
undefined

Josh: a lot less tokens are going towards frontier intelligence to some extent that (08:09):
undefined

Josh: frontier intelligence is much higher profit margins it's going to be worth a (08:13):
undefined

Josh: tremendous amount of money for the people that use it (08:16):
undefined

Josh: but the amount of use cases that require frontier intelligence are going down (08:18):
undefined

Josh: because the quality of the open source models are going up so quickly so therefore (08:22):
undefined

Josh: um why would you not lean into that and if you are a gpu provider who gets paid (08:26):
undefined

Josh: per token generated and your performance for what to generate tokens and (08:31):
undefined

Josh: how many tokens you can generate on a like a per second basis open weights is (08:35):
undefined

Josh: definitely where you want to be you want to be able to cater towards the open (08:39):
undefined

Josh: source community that's going to be generating a tremendous amount of this inference (08:42):
undefined

Josh: and then the final part is just like the moat is continuing to move up the stack so like (08:45):
undefined

Josh: we have the gpus then nvidia is going to (08:49):
undefined

Josh: kind of optimize these models built for cuda which works by default with their GPUs, and... (08:52):
undefined

Josh: There will be a one-click deployment onto the nvidia cloud and there's just (08:57):
undefined

Josh: this huge vertical integration that they get by being supportive of open source (09:01):
undefined

Josh: and building tools for the open source community to very easily deploy on their gpus (09:04):
undefined

Josh: and that seems like the strategy and that is a strategy that is going to perform (09:08):
undefined

Josh: very well because as we know the cloud based businesses in these companies have (09:11):
undefined

Josh: insane profit margins they print cash and nvidia is like right in the middle (09:16):
undefined

Josh: of that so really exciting news from nvidia this week can (09:21):
undefined

Ejaaz: I can i put my tinfoilhound for. (09:23):
undefined

Josh: A second. Yeah, well, you guys, please. (09:25):
undefined

Ejaaz: You reminded me of something just now. So NVIDIA actually doesn't have a really (09:26):
undefined

Ejaaz: good cloud business at all, which seems remarkable for the guys that actually (09:30):
undefined

Ejaaz: create the GPUs that they then sell to Amazon and Google. They're on it. (09:34):
undefined

Ejaaz: Don't count them out. They make a bank of money. (09:37):
undefined

Ejaaz: Well, do you know about the margins for these cloud providers? (09:38):
undefined

Ejaaz: So like, okay, typically when AWS has their cloud system, like forget about AI, right? (09:42):
undefined

Ejaaz: And Google has, what is it? Google Cloud GCP. (09:47):
undefined

Ejaaz: They typically make around 80% margins. But with AI cloud, they make only 40 (09:50):
undefined

Ejaaz: to 50% because guess who's taking the bulk of the profit? It's Jensen, right? (09:55):
undefined

Ejaaz: And so they kind of want to be independent. They want to build their own chips. (10:00):
undefined

Ejaaz: But Jensen hasn't been focused on building his own cloud provider. (10:02):
undefined

Ejaaz: They had a service called DGX and it kind of didn't make much money at all. (10:06):
undefined

Ejaaz: My conspiracy theory is this. He's purchasing Hugging Face, (10:11):
undefined

Ejaaz: which one of their mainly used products is this thing called the inference service, (10:16):
undefined

Ejaaz: where basically not only can you go on and download these open models, (10:21):
undefined

Ejaaz: then there are hundreds of thousands, there's actually three million of them, (10:24):
undefined

Ejaaz: but you can inference them there on the platform (10:27):
undefined

Ejaaz: by just kind of clicking a button and that gets routed to different like neoclouds. (10:31):
undefined

Ejaaz: So my whole thing is Jensen is not only providing the chips, (10:36):
undefined

Ejaaz: but he's buying a company where he can effectively churn into his own cloud (10:40):
undefined

Ejaaz: business that can compete with AWS and GCP. (10:44):
undefined

Ejaaz: It's a really genius thing because he just focuses on open source models, (10:47):
undefined

Ejaaz: but he can build that out into private models in the future if he so chooses to do it. (10:50):
undefined

Ejaaz: The second thing is, remember last week when we spoke about NVIDIA raising $500 (10:54):
undefined

Ejaaz: billion to basically back a lot of the purchases that end up going back to himself? (10:58):
undefined

Ejaaz: Well, a question that we had on that show was, what happens if he can't sell the chips? (11:04):
undefined

Ejaaz: Well, guess what? If he's purchased Hugging Face, he could just sell the chips (11:09):
undefined

Ejaaz: to Hugging Face to serve that same cloud business. So there's like this kind (11:13):
undefined

Ejaaz: of circular virtuous loop, but it makes sense from a business sense. (11:17):
undefined

Ejaaz: And I think that that's how he might be trying to play it. It is a very smart (11:21):
undefined

Ejaaz: business model. I like seeing NVIDIA expand beyond just making bleeding edge (11:24):
undefined

Ejaaz: GPUs. He's like a very smart businessman, obviously. (11:29):
undefined

Josh: And Jensen, we trust, man. The dude, he's got to figure it out. (11:31):
undefined

Josh: He's been projecting these crazy revenue numbers for years now. (11:34):
undefined

Josh: And he just like happens to be right every time so at some point you got to (11:37):
undefined

Josh: trust the guy that's been doing this for 33 years and say like all right maybe (11:40):
undefined

Josh: he knows he's talking about here we have some other stuff yeah this is very important (11:43):
undefined

Josh: We were right first of all we recorded an episode got published yesterday if (11:47):
undefined

Josh: you're listening to this we promise you a follow-up here's the follow-up there's (11:50):
undefined

Josh: a mystery model that gave 100 trillion tokens of availability out to the public (11:53):
undefined

Josh: totally for free they said here come and use it do whatever you want with it (11:58):
undefined

Josh: it's all on us for a full week (12:01):
undefined

Josh: We're not going to tell you who it is, though. We discovered yesterday who it (12:03):
undefined

Josh: was, and it is Z.ai. It's the people that brought you GLM. (12:05):
undefined

Josh: And it's their new model, 5.3 Flash. This looks like a normal model. (12:09):
undefined

Josh: This is all fine. Until I was reading through this, I realized that all of those (12:13):
undefined

Josh: tens to hundreds of trillions of tokens that were generated were not only generated (12:17):
undefined

Josh: for free, but they were generated entirely on a chip cluster that runs on Chinese silicon, (12:21):
undefined

Josh: which is a first of its kind. (12:28):
undefined

Josh: I don't think we've ever had an instance in which Chinese silicon has produced (12:30):
undefined

Josh: this amount of tokens and this amount of intelligence. Exactly. (12:34):
undefined

Josh: As the benchmarks come out, we've started to discover that it's about opus tier (12:38):
undefined

Josh: agentic level, which is very good. This isn't Frontier, but it's 15 cents per million tokens. (12:41):
undefined

Josh: And it understands video, audio, and text. So this is a really impressive model out of China. (12:48):
undefined

Josh: And on the back of our Robot Olympics episode, EJs, that we just recorded talking (12:54):
undefined

Josh: about how dominant the robots are, now I'm seeing that they are actually getting (12:58):
undefined

Josh: silicon to work at scale with pretty quick inference and a lot of it. (13:02):
undefined

Josh: Like, I'm like, oh, wow. OK, China's like they're making moves this week. (13:05):
undefined

Ejaaz: Yeah, it's concerning a little bit. And I'll tell you why. (13:09):
undefined

Ejaaz: Recently, America, at least for the last like four months, has imposed export (13:12):
undefined

Ejaaz: bans on China, meaning that they (13:16):
undefined

Ejaaz: don't allow NVIDIA to sell chips to China, which, by the way, (13:18):
undefined

Ejaaz: just to kind of reference the quarterly earnings from NVIDIA, (13:22):
undefined

Ejaaz: all that money was made by not even selling a single chip to China. (13:24):
undefined

Ejaaz: Very impressive because China was like, you know, a major profit revenue holder. (13:27):
undefined

Ejaaz: But anyway, for the Chinese Silicon, right? (13:31):
undefined

Ejaaz: They're like looking at this, they're looking at this export band. (13:34):
undefined

Ejaaz: They're like, damn, we can't train Frontier models without NVIDIA chips. (13:36):
undefined

Ejaaz: I guess we need to build our own. And whilst they haven't built a good enough (13:40):
undefined

Ejaaz: chip that can be used to train their own AI models, they still use various different (13:45):
undefined

Ejaaz: techniques to kind of circumvent that, such as distillation. (13:49):
undefined

Ejaaz: They can build an amazing chip to inference the model, which by the way, (13:52):
undefined

Ejaaz: is where all the money is going to be made in AI by a vast order of magnitude versus pre-training. (13:57):
undefined

Ejaaz: When you look at the Chinese chip companies about six months ago, (14:04):
undefined

Ejaaz: they were two and a half years behind. (14:07):
undefined

Ejaaz: So can someone please explain to me how six months later, they're now here running (14:09):
undefined

Ejaaz: the top stack of one of the top open source models, which happens to compete (14:15):
undefined

Ejaaz: very well with US frontier models purely on Chinese silicon. (14:19):
undefined

Ejaaz: Now, if you're wondering, who's the company that built these AI chips? (14:22):
undefined

Ejaaz: Look no further than Huawei. They released a bunch of different things. (14:25):
undefined

Ejaaz: I think it's called, actually i'm not going to try and make this up because i can't, (14:29):
undefined

Ejaaz: uh remember the name of the exact chip but the cluster became live or got announced about a month ago (14:33):
undefined

Ejaaz: and now it's at full scale production now partly they're being forced to use (14:39):
undefined

Ejaaz: these chips because the chinese state government has mandated them to just not (14:44):
undefined

Ejaaz: rely on invidia chips for anything and they replied and said we can't do it (14:47):
undefined

Ejaaz: for training but we can do it for inference so let us do it (14:50):
undefined

Ejaaz: and this is the first real example and i don't know if you (14:53):
undefined

Ejaaz: used uh the model when it was in the aux alpha (14:56):
undefined

Ejaaz: stage um but okay so you you saw like it was pretty good right (14:59):
undefined

Ejaaz: but you would get a response really quickly it was super cheap didn't cost you (15:02):
undefined

Ejaaz: anything and so it functions very similarly to an nvidia inference stack and (15:05):
undefined

Ejaaz: yeah it is scary man like what happens when they (15:09):
undefined

Ejaaz: own the chip infrastructure layer and the robotics hardware manufacturing layer (15:12):
undefined

Ejaaz: like what else is left it's just they're. (15:16):
undefined

Josh: Making a lot of progress in a way that's like definitely a little unnerving (15:18):
undefined

Josh: i'm like okay this is this is new (15:21):
undefined

Josh: and i mean this is kind of the expectation right is you like i know huawei is (15:23):
undefined

Josh: banned in the united states a lot of the devices you can't really get them they're (15:26):
undefined

Josh: not sold anywhere the inverse is happening and what does a country do that doesn't (15:29):
undefined

Josh: have access to chips well they have to go and build their own (15:33):
undefined

Josh: and now they're going to continue to build their own and scale them and (15:35):
undefined

Josh: It's going to be really interesting to see how this plays out. (15:38):
undefined

Josh: They are not the only ones building chips, though, because we have a new company (15:41):
undefined

Josh: that's building chips. And this one I'm far more excited about. (15:44):
undefined

Ejaaz: Something good about Google. Look at this beauty. Is that what you're telling me? Is this Waymo? (15:46):
undefined

Josh: Yes. Waymo is looking good, man. So Waymo has this new vehicle, first of all. Oh, hi. (15:50):
undefined

Josh: And then even more importantly, they have a new chip architecture that they're (15:54):
undefined

Josh: introducing into this new vehicle design. (15:57):
undefined

Josh: And it seems like it's a fairly large improvement. I know you were going through (16:00):
undefined

Josh: the blog post, so maybe you could walk us through. But my understanding is that (16:03):
undefined

Josh: Waymo is about to get a serious upgrade. (16:06):
undefined

Josh: And not in the sense that it's going to cost less because these cars i think (16:08):
undefined

Josh: the number was like 260 000 per jaguar waymo like it's still very expensive (16:11):
undefined

Josh: but the chip that handles it (16:15):
undefined

Josh: is now 20 times more powerful than it was over the last eight years and that (16:17):
undefined

Josh: seems to make a meaningful difference in terms of the quality of your ride that (16:21):
undefined

Josh: you can actually go out and get today (16:24):
undefined

Josh: if you live in one of the areas you can go and ride away more fully autonomously (16:25):
undefined

Josh: i've done it you just i think you've done it um incredible company that has (16:28):
undefined

Josh: a really meaningful upgrade to the hardware today (16:31):
undefined

Ejaaz: So a few things with this new chip that they just created. (16:34):
undefined

Ejaaz: In prior Waymo generations, so in prior vehicles, because they have like so (16:39):
undefined

Ejaaz: many sensors, how many are there? Like 15 to 20 sensors? It's like cameras, proximity sensors. (16:43):
undefined

Josh: A lot of bulky ones, man. It looks crazy. (16:49):
undefined

Ejaaz: Sorry, I have to say it. I respect Google a lot. One of the ugliest autonomous (16:51):
undefined

Ejaaz: cars out there. The new ones are cool. (16:55):
undefined

Ejaaz: The new ones are cool, but mainly they've been horrific. The point is (16:56):
undefined

Ejaaz: the car has to ingest a lot of data translate that data into some kind of an (17:00):
undefined

Ejaaz: action, which is steer right, steer left, break, you know, get this person safely (17:06):
undefined

Ejaaz: from A to B as quickly as you can. (17:11):
undefined

Ejaaz: Now, in order to process all that information, you need a really powerful AI chip. (17:13):
undefined

Ejaaz: Waymo didn't have that. So they kind of used a hybrid of like a chip and a cloud (17:18):
undefined

Ejaaz: service. Now, that's kind of dangerous because what if the cloud service kind of breaks connection? (17:24):
undefined

Ejaaz: You end up being in a whole pot of mess. (17:29):
undefined

Ejaaz: Now with this new chip it brings all the compute on board the actual device (17:32):
undefined

Ejaaz: the device in this case being a car which i think is really cool so the entire (17:38):
undefined

Ejaaz: system josh is being run locally on device (17:41):
undefined

Ejaaz: in the single car so when in the trunk. (17:45):
undefined

Josh: Of the car i love that visual that they show exactly yeah (17:47):
undefined

Ejaaz: Wait let me let me get the the visual up yeah that's. (17:49):
undefined

Josh: Why they are showing the trunk if it wasn't clear for the last one it's like (17:52):
undefined

Josh: yeah all of that compute lives sitting underneath the trunk there. (17:54):
undefined

Josh: It's so, so cool. Pretty cool. Yeah. Yeah. (17:58):
undefined

Ejaaz: And it's a single chip or a cluster of these different chips. (18:00):
undefined

Ejaaz: And so it takes the input data from the sensors. (18:03):
undefined

Ejaaz: It processes everything through this one chip, which by the way, (18:06):
undefined

Ejaaz: is 20X more powerful and is 5X cheaper. (18:10):
undefined

Ejaaz: They lessen the cost. So do you remember in our last Waymo episode, (18:14):
undefined

Ejaaz: Josh, we were like, why do each of these cars cost like 100K? (18:17):
undefined

Ejaaz: It's because like more 260 or 280. Oh my God. (18:21):
undefined

Ejaaz: 260K. Yes. Now it costs cost $25,000 a car to pop out because they were able (18:25):
undefined

Ejaaz: to create the silicon chip and reduce the cost down and run everything lower. (18:31):
undefined

Josh: Is that real? This new car costs $25,000? (18:34):
undefined

Ejaaz: Dude, look at this. Wait one second. Where's the breaking news? Yeah, over here. (18:35):
undefined

Ejaaz: So it used to cost $100,000 to $125,000. Come to the street at Wall Street. (18:39):
undefined

Josh: Okay, this is an important distinction we (18:44):
undefined

Ejaaz: Need to make. Talk to me. Talk to me. Is this Miss Info? Tell me. Tell me. No, no, no. (18:46):
undefined

Josh: So this is for the hardware. This is not for the vehicle. Okay, (18:49):
undefined

Josh: that's where I was getting confused. (18:54):
undefined

Josh: So the hardware part is like the sensor suite that sits on top of the vehicle. (18:55):
undefined

Josh: So I guess it was like $125,000 for the last one on top of a $100,000 vehicle. (19:01):
undefined

Josh: This one's $25,000 on top of an assumed lower cost vehicle. So that's an important distinction. (19:05):
undefined

Ejaaz: Yeah, yeah, sorry, sorry. Just to be clear, we're talking about the chip hardware (19:10):
undefined

Ejaaz: here, like all of that, which kind of like, you know, makes the car sexy and (19:14):
undefined

Ejaaz: do the cool thing, right? (19:17):
undefined

Ejaaz: And so the point is, with this new chip, (19:19):
undefined

Ejaaz: been able to achieve a few things. Number one, it runs locally on the car, (19:22):
undefined

Ejaaz: which means it's cheaper to run. (19:26):
undefined

Ejaaz: It's way more effective. So it's quicker. It's way more reliable. (19:28):
undefined

Ejaaz: And it's incredibly efficient. So it costs Google less. It costs Waymo less, (19:31):
undefined

Ejaaz: which is why they've been able to kind of reduce the cost by like three to four (19:36):
undefined

Ejaaz: X, which is very impressive for any successive generation. (19:41):
undefined

Ejaaz: I just think it's a really cool bit of hardware device. You know, (19:43):
undefined

Ejaaz: we like hardware on this show. And I'm happy Waymo's taking this path forward. (19:45):
undefined

Josh: Very good. It looks cool it looks like a fun science experiment like the car (19:49):
undefined

Josh: itself looks interesting the new ohi it's (19:51):
undefined

Josh: fun it's different it looks like disney world this is a really cool promo video (19:54):
undefined

Josh: showing it oh my god wow yeah it's it's like a unique design i mean granted (19:57):
undefined

Josh: if you were to design a fully autonomous car i'm not sure why there's two seats (20:02):
undefined

Josh: still facing forward in the way they are with the steering wheel (20:05):
undefined

Ejaaz: Hang on wait did he turn around in this let's see josh. (20:07):
undefined

Josh: I'm not sure oh no no he's he's turning his body around so we mentioned the (20:10):
undefined

Josh: cost of these sensors is it's about 25 000 for the suite, the hardware, the sensors. (20:15):
undefined

Josh: You know what else is $25,000 each as? (20:20):
undefined

Josh: The entire CyberCab from Tesla, the whole thing, including the vehicle. It will be $25,000. (20:22):
undefined

Ejaaz: $25,000 per vehicle? (20:28):
undefined

Josh: Per vehicle, with the whole sensor stack. So what you'll notice with the Waymos (20:30):
undefined

Josh: is they have this huge amount of spinning LiDAR sensors and very complicated sensor stack. (20:33):
undefined

Josh: The Robotaxi that we're seeing on screen here, the CyberCab from Tesla, (20:39):
undefined

Josh: has, I think it's eight cameras, very similar to the ones that are sitting in (20:42):
undefined

Josh: your smartphone right now. (20:46):
undefined

Josh: So basically, if you imagine the camera's on your smartphone, (20:47):
undefined

Josh: they are placed into a vehicle and put in the right place. (20:50):
undefined

Josh: That combined with the cost of the actual vehicle is... (20:53):
undefined

Josh: Probably going to fall around $25,000 in terms of cost of goods sold. (20:57):
undefined

Josh: And it works. And it is being deployed starting September 3rd at scale. (21:00):
undefined

Josh: So what we see here is actual production ready level cybercabs that are going (21:04):
undefined

Josh: to be going around a select few cities. (21:09):
undefined

Josh: In the announcement post, could we pull that up so I can read it off? (21:10):
undefined

Josh: I'm not sure exactly which ones is going. I know there is one in Austin. (21:13):
undefined

Josh: I know there's one in Dallas. So yeah, Austin, Dallas, Houston, Miami, Orlando, Tampa. (21:17):
undefined

Josh: If you live in any of those places, you can go and get in a cybercab right now. (21:22):
undefined

Josh: And I think this is going to be a really unique and interesting inflection point (21:26):
undefined

Josh: for the autonomous vehicle world (21:30):
undefined

Josh: in general, because we have Waymo that's making really good progress. (21:33):
undefined

Josh: And Waymo is right now much more accessible. They're doing many more miles per day. (21:36):
undefined

Josh: And people have gotten comfortable with it. If you live in Los Angeles, (21:40):
undefined

Josh: if you live in San Francisco, having Waymos go around, it's just a part of your (21:42):
undefined

Josh: life. And you don't even think twice about it now. (21:46):
undefined

Josh: The CyberCab has an opportunity to really change that in the sense that the (21:49):
undefined

Josh: cost per mile on the cyber cab will be significantly less than that of the waymo (21:52):
undefined

Josh: when you look at the waymo vehicle it has like these big bumpers that's really (21:55):
undefined

Josh: big and it's not a very efficient car (21:58):
undefined

Josh: the cyber cab has much more efficiency which allows it to decrease the cost (22:00):
undefined

Josh: per mile and it's going to be really interesting to see how they're able to scale that because (22:04):
undefined

Josh: Waymos notoriously are geofenced they work in very specific locations i've been (22:08):
undefined

Josh: using a tesla fully self-driving for years now and (22:12):
undefined

Josh: with no oversight they work anywhere so as soon as regulation comes online We (22:15):
undefined

Josh: may see a world in which like 12 months from now, cybercabs are just roaming around everywhere. (22:18):
undefined

Josh: And we have the opportunity to get in them, drive them around. (22:23):
undefined

Josh: And the idea is that you'll never need to own a car because the cost per mile (22:25):
undefined

Josh: will be so low that it will offset any expenses incurred by actually buying (22:29):
undefined

Josh: the physical vehicle. So really cool time. (22:33):
undefined

Josh: For autonomous vehicles, autonomous cars in general. Keep an eye out for Wemo. (22:36):
undefined

Josh: Keep an eye out for the Tesla event that's happening next week. (22:39):
undefined

Ejaaz: You know where I'm not going to be keeping my eye out? In New York City. (22:41):
undefined

Josh: Because they like to just ban all innovation. (22:44):
undefined

Ejaaz: New York City sucks, man. (22:47):
undefined

Josh: It sucks, man. (22:48):
undefined

Ejaaz: It's so brutal. It's so cool, but it sucks. Like, bring us the new tech early (22:49):
undefined

Ejaaz: on, for goodness sake, please. (22:53):
undefined

Ejaaz: Now, I want to have a call out just very quickly on this particular topic, Josh. (22:55):
undefined

Ejaaz: Producer Luke, we love him. He is in one of these places. He's in Austin. (22:59):
undefined

Ejaaz: Guess how many times he's rode in a robo-taxi? Go on, guess. (23:03):
undefined

Josh: Well, I think once, maybe. (23:06):
undefined

Ejaaz: I think it's zero, dude. I think it's zero. (23:08):
undefined

Josh: He may have done one, but it wasn't a cyber cab. (23:10):
undefined

Ejaaz: I don't trust it. That image was AI generated. I need to see you, Luke. (23:12):
undefined

Ejaaz: I know you'll listen to this because you're editing this right now. (23:17):
undefined

Ejaaz: Like, I need to see you in a cyber cab, dude. It would be awesome. (23:19):
undefined

Josh: Anyway, moving on. We'll get a fact checker in here. (23:22):
undefined

Ejaaz: Who is our favorite investor in the world, Josh? (23:24):
undefined

Josh: Oh, you mean the 25-year-old who was fired by OpenAI, raised a billion dollars (23:26):
undefined

Josh: and then had 45 and then lost it all and then Citadel bought it all out and (23:30):
undefined

Josh: now he's being investigated by the SEC? (23:33):
undefined

Ejaaz: Wow, actually a very good, succinct summary. That's a very ChatGPT-esque, (23:35):
undefined

Ejaaz: Josh. Very impressive. Yes, that is actually it. That guy. (23:39):
undefined

Josh: Wait, but what's his name again? (23:42):
undefined

Ejaaz: Yeah, Leopold Ashenbrenner. (23:44):
undefined

Josh: Oh, yeah, that one, that one, Leopold. Yeah, yeah, that's right. (23:45):
undefined

Ejaaz: Okay, listen, like my hands are up for those of you who are about to like write nasty comments, right? (23:47):
undefined

Ejaaz: We love Leopold Ashenbrenner. I still think his thesis isn't intact. (23:53):
undefined

Ejaaz: He just, you know, don't trade with leverage, kids. (23:56):
undefined

Ejaaz: But he is now being investigated by the SEC or rather specifically, (23:59):
undefined

Ejaaz: the SEC has subpoenaed a bunch of banks which facilitated situational awareness (24:04):
undefined

Ejaaz: funds, which is Liverpool's fund, trades out of suspicion that potentially the, (24:09):
undefined

Ejaaz: there was insider trading. Now, this is something that we have kind of briefly (24:15):
undefined

Ejaaz: referenced and spoken about on the show. (24:18):
undefined

Ejaaz: Ajit Balwit, who is now officially Leopold Ashenbrenner's wife, (24:21):
undefined

Ejaaz: is the chief of staff at Anthropik. (24:25):
undefined

Ejaaz: And there's a few other types of things going on there. But effectively, (24:26):
undefined

Ejaaz: the question is, has Leopold been kind of getting information and trading on (24:29):
undefined

Ejaaz: that information when he should have kind of like publicly disclosed it? We don't know. (24:33):
undefined

Ejaaz: But this is just another kind of pitfall into this whole, (24:37):
undefined

Ejaaz: probe where, you know, Leopold's been taking hits, hit after hit recently (24:41):
undefined

Ejaaz: um he recently got back into the game actually he's like gotten a new open book (24:45):
undefined

Ejaaz: he made a 400 million dollar private investment in some other company goodness (24:48):
undefined

Ejaaz: knows where he got 400 million dollars from but he's doing the thing and technically (24:51):
undefined

Ejaaz: the fund is still up but yeah just curious to see why this all ends up i don't (24:55):
undefined

Ejaaz: know if you have any thoughts on this josh. (24:58):
undefined

Josh: Yeah it's mostly just like people don't like leopold and what does it take to (25:00):
undefined

Josh: open an investigation you file a complaint someone says okay we've acknowledged (25:03):
undefined

Josh: your complaint now there's an open investigation and then you can report the investigation (25:06):
undefined

Josh: it could be nothing it could be something chances are it's probably nothing (25:10):
undefined

Josh: and i think leopold's just down people want him to be down even more and chances (25:14):
undefined

Josh: are he's going to be just fine this will kind of blow boat over and there won't (25:19):
undefined

Josh: really be much that comes from it (25:22):
undefined

Josh: um in leopold we trust man i'm still rooting for the guy i'm excited to see (25:24):
undefined

Josh: the new 13f to see what on earth is going to be (25:27):
undefined

Josh: on the comeback portfolio after the entire public one got wiped out so (25:30):
undefined

Josh: we'll keep our eyes peeled as we always do on leopold I'm also seeing some charts (25:34):
undefined

Josh: here, EJS, on Fable 5 and Anthropic and like enterprise sales and businesses. (25:38):
undefined

Josh: What is what are we showing here? (25:44):
undefined

Ejaaz: I just think there's some chartsmithing going on here. So like a few things (25:46):
undefined

Ejaaz: going on here. So the Financial Times released a chart and it shows the usage (25:51):
undefined

Ejaaz: of Anthropic's different models over time. (25:55):
undefined

Ejaaz: And I don't know if I can open this up, Josh. I don't know if this makes it (25:59):
undefined

Ejaaz: makes it a little tricky for you. (26:02):
undefined

Josh: We'll move it. We'll move it. There we go. (26:03):
undefined

Ejaaz: But if you can have a look over here, you'll notice that Fable growth kind of (26:06):
undefined

Ejaaz: like accelerated in mid-June to the kind of the end of July and has kind of been flat. (26:10):
undefined

Ejaaz: Now, number one, I don't know where they've got these numbers from because Anthropik (26:15):
undefined

Ejaaz: is not a private company, so I don't think these are necessarily reliable. (26:18):
undefined

Ejaaz: But the point they're making is it has stagnated. (26:22):
undefined

Ejaaz: And instead, people or companies rather are choosing to use some of their other (26:25):
undefined

Ejaaz: clawed models. Now, there are many reasons why companies would do this, (26:29):
undefined

Ejaaz: whether it's cost, whether it's efficiency, whether it's just kind of like preference (26:32):
undefined

Ejaaz: or taste of like the model's personality. (26:35):
undefined

Ejaaz: But it's interesting to see that a frontier model isn't being taken up as much as they can. (26:37):
undefined

Ejaaz: And the point that they compared this to is that OpenAI's models have seen increased usage. (26:41):
undefined

Ejaaz: Now, my comments on this is simple, which is OpenAI is just like has a smaller (26:46):
undefined

Ejaaz: user base than Anthropik. (26:51):
undefined

Ejaaz: Now, it's still granted very large, but obviously any hike in usage, (26:52):
undefined

Ejaaz: which they're particularly seen with Codex specifically, which, (26:56):
undefined

Ejaaz: by the way, is a fantastic product, is going to result in a larger uptake or (26:59):
undefined

Ejaaz: percentage increase, right? (27:03):
undefined

Ejaaz: Whereas with Fable, maybe if the stagnation is the thing, maybe people are just, (27:04):
undefined

Ejaaz: you know, factoring in the cost. (27:08):
undefined

Ejaaz: Fable is notably, you know, more expensive than a lot of other models, (27:09):
undefined

Ejaaz: but it is also operating at the frontier better than any other model. (27:12):
undefined

Ejaaz: So I can see why maybe there's a hesitancy, you know, you've got all these new (27:15):
undefined

Ejaaz: open source models, GLM 5.3 Flash, and you're thinking, okay, (27:19):
undefined

Ejaaz: maybe if I can get 80% of my work done using a cheap model, I'll use Fable for (27:22):
undefined

Ejaaz: the other 20%. But the point is, the pie is growing bigger and bigger. And I think this is, (27:26):
undefined

Ejaaz: None other than a bit of a hit piece. (27:30):
undefined

Josh: Well, it looks like the chart is reflecting the pie growing bigger, (27:32):
undefined

Josh: right? It's like the numbers are going up into the right. (27:34):
undefined

Josh: The business spending in total based on this chart looks like it's going up. (27:36):
undefined

Josh: And maybe this is kind of the phenomenon that we described earlier where there's (27:41):
undefined

Josh: just a lot of demand for lower cost tokens, but the overall demand is continuing (27:44):
undefined

Josh: to increase. And like perhaps that is a conclusion we can draw from this. (27:48):
undefined

Josh: I'm confused what the other actually is. Like, are they talking about like Haiku (27:52):
undefined

Josh: or something? Because they have Sonnet, Opus, Fable. What else is there? (27:56):
undefined

Ejaaz: I'm not quite. (27:59):
undefined

Josh: Sure what's going on here i don't know but hey everyone's having i mean the (28:00):
undefined

Josh: models are great on a day-to-day basis i'm still using it i'm getting a lot (28:04):
undefined

Josh: of usage out of it and i'm kind of enjoying the time there (28:07):
undefined

Josh: so it's um yeah we'll see i think is kind of like where where we can leave this (28:09):
undefined

Josh: one is like okay well i'm i'm (28:14):
undefined

Josh: sorry for enterprises but like hey we're cooking like for twenty dollars a month (28:17):
undefined

Josh: a hundred dollars a month two hundred dollars a month like the tokens are flowing (28:21):
undefined

Ejaaz: I think there's a there's another reason though right which is like did you (28:24):
undefined

Ejaaz: see this whole debate around zero data retention. (28:27):
undefined

Ejaaz: I started yawning when I read that phrase, Josh, because I was like, (28:30):
undefined

Ejaaz: oh my God, what do you mean by this? (28:32):
undefined

Josh: But like, yeah, oh, sounds fun. (28:33):
undefined

Ejaaz: And another big thing for these enterprises is they want to use the latest frontier (28:35):
undefined

Ejaaz: models. And to be honest, when you're an enterprise, you don't care about spending (28:40):
undefined

Ejaaz: more money if it means you're going to make even more money. (28:43):
undefined

Ejaaz: So like, I'll spend money on the most expensive model. (28:45):
undefined

Ejaaz: But a big compliance thing that they face is this thing called zero data retention, where (28:48):
undefined

Ejaaz: if you have a company, a software company, which you kind of lease their tools (28:52):
undefined

Ejaaz: or whatever, and they retain proprietary information for their own usage, (28:57):
undefined

Ejaaz: which they could technically use to build better models, which is what Anthropik has to do right now. (29:01):
undefined

Ejaaz: But I believe they're in the process of evolving that potentially. (29:05):
undefined

Ejaaz: Who knows? This is just rumors at this point. (29:08):
undefined

Ejaaz: I believe this is something that will just kind of flip that switch and suddenly (29:10):
undefined

Ejaaz: we'll look back on that chart and suddenly it's spiked up growth. (29:14):
undefined

Ejaaz: So I just think this is a nothing burger. ZDR is very common. (29:16):
undefined

Ejaaz: I don't think OpenAI is enforcing it right now, which might, (29:20):
undefined

Ejaaz: be the reason why they're like accelerating faster over the last couple of months (29:23):
undefined

Ejaaz: but just something to be aware of. (29:26):
undefined

Josh: Okay so maybe the chart is nothing burger but the thing that is real is the (29:28):
undefined

Josh: use of tokens and the use of tokens are generated mostly by agents and if you (29:31):
undefined

Josh: are working with agents we have just the people for you their name is ledger (29:34):
undefined

Josh: ledger is offering a way to build with your ai agents because if you're worried (29:38):
undefined

Josh: about security you should be because agents as we know are (29:42):
undefined

Josh: not very secure at the moment they're getting very smart very quickly ledger (29:45):
undefined

Josh: has an agent stack that fixes this it's a series of open source tools that you (29:49):
undefined

Josh: can go and use in order to make sure that the agent (29:52):
undefined

Josh: is giving you the expected outcome that you want. (29:55):
undefined

Josh: It does this through a three-step process where the agents propose, (29:57):
undefined

Josh: the humans approve, and the ledger signers enforce. And this is all open source. (30:00):
undefined

Josh: It works everywhere that you work with your AI agents. It works in codex, (30:04):
undefined

Josh: it works in cloud code, it works in cursor. (30:07):
undefined

Josh: So if you are interested, the link is in the description down below. (30:09):
undefined

Josh: Thank you very much to Ledger for sponsoring this episode. (30:12):
undefined

Josh: Two nothing burgers in a row. We got to turn this around, man. (30:14):
undefined

Josh: Leopold, Anthropic, what do we got? This is not a nothing burger. (30:16):
undefined

Josh: This is a something burger. (30:19):
undefined

Ejaaz: What if I gave you continual learning robotics model, huh? (30:20):
undefined

Josh: Dude, it's a big week for the robots. First, we had the Robot Olympics. (30:23):
undefined

Josh: Then we have full self-driving vehicles. Now we have robotic learning. (30:26):
undefined

Josh: And this is one of many announcements that I saw this week. There was another (30:30):
undefined

Josh: one that I thought was super cool where you can actually go and get paid like (30:32):
undefined

Josh: $15 an hour to perform tasks in your house. (30:35):
undefined

Josh: And as long as you do that with the camera, that allows you to train the robots. (30:38):
undefined

Josh: You will be paid as if you are a full-time employee. (30:41):
undefined

Josh: It's like kind of cool, very fun. This is another version of that, (30:44):
undefined

Josh: it seems like, where they're just... (30:47):
undefined

Josh: Basically throwing a ton of data at these models and emerging from that data are (30:50):
undefined

Josh: physical capabilities that robots are learning from scratch and they're doing (30:56):
undefined

Josh: so in a way that's fairly efficient what i like about this example that we're (30:59):
undefined

Josh: seeing on screen right now too is it's not this isn't purely humanoid this is (31:02):
undefined

Josh: more of a narrow purpose uh narrow (31:06):
undefined

Josh: very specific type of robot that they're working on so it seems like this is (31:09):
undefined

Josh: a kind of new paradigm where robots are getting enough data to start meaningfully (31:12):
undefined

Josh: progress and get good at these tasks. (31:17):
undefined

Josh: And we're seeing so many unique and creative ways of collecting this data. I remember we did one (31:19):
undefined

Josh: doordash or something where they were paying dashers to strap cameras to their heads (31:24):
undefined

Ejaaz: It was doordash. (31:28):
undefined

Josh: It was doordash that way they can like go and collect data for what it looks (31:29):
undefined

Josh: like to deliver an item to a person's house and then taking all that data you get the antik (31:31):
undefined

Ejaaz: The pokemon go company. (31:36):
undefined

Josh: As well that's right that's right yeah so there's all these really creative ways of of (31:37):
undefined

Josh: paying humans to collect data to then automate the tasks that they're doing (31:42):
undefined

Josh: and i think it's really cool to see the actual progress come from this in terms (31:46):
undefined

Josh: of the ability of these robots to do more and more interesting tasks that we (31:49):
undefined

Josh: as humans do and maybe we don't want to (31:53):
undefined

Josh: and i think a lot of people probably default to like oh this is bad we're paying human beings to (31:56):
undefined

Josh: offload the responsibilities to robots but the reality is is like how many people actually (32:00):
undefined

Josh: love doing this stuff like how many people love to clean or love to cook or (32:04):
undefined

Josh: love to take the trash out or love to like deliver food (32:07):
undefined

Josh: there's perhaps much better options on the other side of automating that and (32:10):
undefined

Josh: it results in much lower costs for everyone who is using those services so this (32:14):
undefined

Josh: seems like a win, even though on the surface, I could see why it's scary. (32:18):
undefined

Josh: And this is a really cool example of that in practice. I mean, (32:21):
undefined

Josh: this is a pretty sweet video. (32:24):
undefined

Ejaaz: Yeah, I mean, just to kind of explain what's going on here, they made a breakthrough (32:25):
undefined

Ejaaz: in robotics models. Now, typically, the problem that you have with robotics (32:29):
undefined

Ejaaz: models is they're really hard to train in order to get the data to teach it (32:33):
undefined

Ejaaz: how to do really complex stuff. (32:36):
undefined

Ejaaz: You need to, you know, get those videos like you mentioned, Josh, (32:38):
undefined

Ejaaz: and you need to kind of like teach it a million different things that hard code it into a model. (32:41):
undefined

Ejaaz: It is incredibly complex and hard to do, which is why we haven't seen a cool (32:44):
undefined

Ejaaz: robot in your house as of recently. (32:49):
undefined

Ejaaz: The breakthrough these guys made is something called in-context learning, (32:51):
undefined

Ejaaz: which means you can feed this robot one video of you making pancakes and flipping the pancake, (32:54):
undefined

Ejaaz: and it'll be able to figure it out from that 30-second video and do it itself. (33:02):
undefined

Ejaaz: But the craziest part is, Josh, let's say. (33:07):
undefined

Ejaaz: I filmed you making a pancake and you flipped it and it fell on the floor, (33:10):
undefined

Ejaaz: right? The robot's seen this, right? So presumably you would think it would do the same. (33:14):
undefined

Ejaaz: It doesn't do that. It knows when you've made a mistake and it can take it further (33:19):
undefined

Ejaaz: and start making multiple pancakes at the same time and then stacking them on (33:23):
undefined

Ejaaz: top of each other. It has this intuitive feel. (33:27):
undefined

Ejaaz: So an example that they show in this video here, I think they already passed (33:30):
undefined

Ejaaz: it, is it uses a kind of dustpan and brush to kind of sweep a block onto the pan, right? (33:33):
undefined

Ejaaz: And then they replaced the handle with a banana. (33:38):
undefined

Ejaaz: And it was like, what would it do in this case? It realized that it was a banana. (33:42):
undefined

Ejaaz: It realized it wasn't a brush, but it still did the same thing and brushed the block onto the pad. (33:45):
undefined

Ejaaz: So it has a intuitive understanding of what it needs to do and is willing to (33:50):
undefined

Ejaaz: use any and every kind of tool to do the thing, which is the breakthrough in this robotics model. (33:54):
undefined

Ejaaz: And I think it's going to get us much quicker to having a robot in our home (34:00):
undefined

Ejaaz: that can intuitively just do things without us needing to hard code it or tell it what to do. (34:03):
undefined

Josh: Very very so close on the edge of the robot revolution is a good way of describing (34:08):
undefined

Josh: it we had the olympic games that we had a full episode about earlier this week (34:13):
undefined

Josh: so if you haven't watched that go check that out um (34:17):
undefined

Josh: the opening ceremony is a little bit militant for china which is scary so (34:19):
undefined

Josh: perhaps like you know take that one with a grain of salt um we have a few rapid (34:24):
undefined

Josh: fire ones at the end of this episode my personal favorite here we have a date (34:28):
undefined

Josh: for the apple event It's coming. It is September 9th at 1 p.m. (34:32):
undefined

Ejaaz: Eastern time. Is this Siri AI, Josh? (34:34):
undefined

Josh: Dude, we are finally... No. (34:36):
undefined

Josh: It's not going to happen just yet, but basically what this one is, (34:40):
undefined

Josh: is the hardware. So there's two types of events. There's the developer conference, (34:42):
undefined

Josh: then there's the actual hardware event. (34:45):
undefined

Josh: Developer conference released or unveiled Siri. The release of Siri will come (34:46):
undefined

Josh: shortly after this hardware event, and it will come paired with the new hardware devices. (34:50):
undefined

Josh: The new hardware devices are expected to be a new iPhone 18 Pro, (34:54):
undefined

Josh: a foldable device, which I'm incredibly excited to get my hands on. (34:58):
undefined

Josh: And more importantly, this is the first event with Apple's new CEO in charge. (35:02):
undefined

Josh: Tim Cook isn't going to be on stage here. it's uh john turness who's running (35:07):
undefined

Josh: it there's a new guy in charge he's a hardware guy this is all the new hardware (35:10):
undefined

Josh: we're going to be getting specifically built for (35:13):
undefined

Josh: apple intelligence and man we've been waiting for this for what three years (35:16):
undefined

Josh: to get actual intelligence inside these phones so very excited to see how that's (35:19):
undefined

Josh: going to play out what the hardware is going to look like how the software will integrate with it (35:23):
undefined

Josh: and probably a month from now we will all have brand new iphones with siri intelligence (35:27):
undefined

Josh: inside of them and that is a very exciting thing so Oh, please. Fingers crossed. (35:33):
undefined

Ejaaz: I mean, the chip thing is cool, right? I saw the M5 Pro and the M6 kind of get (35:36):
undefined

Ejaaz: announced and the price points. (35:41):
undefined

Josh: Of those are- New Mac Mini, New Mac Studio. (35:42):
undefined

Ejaaz: Kind of up there, right? So it's like, I think it's like 900 bucks for the M6 (35:44):
undefined

Ejaaz: or, and then 1,500 for the M5. (35:47):
undefined

Josh: Have you heard the high end of the studio? (35:50):
undefined

Ejaaz: Gone. Wait, wait, wait, let me guess, let me guess, let me guess. (35:52):
undefined

Josh: You're gonna be wrong. (35:55):
undefined

Ejaaz: Three thousand dollars no. (35:56):
undefined

Josh: I think it's like eighteen thousand five hundred dollars for them if you want (35:58):
undefined

Josh: to maxed out mac studio with like the m6 ultra and all the ram that you can get it's okay (36:01):
undefined

Ejaaz: Crazy i'm going into. (36:06):
undefined

Josh: That things have gotten but hey if you got the money and you want to run some (36:07):
undefined

Josh: local inference oh my god there's your cost of entry all these people talking about free models buy (36:10):
undefined

Ejaaz: An nvidia gpu at this point dude like what are we doing. (36:13):
undefined

Josh: Here that's crazy model that like just how about you just pay for like a subscription (36:16):
undefined

Josh: or like an api key i'm like let the big guys serve your inference because it's (36:19):
undefined

Josh: getting real expensive to do it locally by the Wait, (36:23):
undefined

Ejaaz: Blame the memory manufacturers for this. That's why the price is hiking across everything right now. (36:25):
undefined

Josh: And they got great margins too. I was looking at the videos and I'm like, (36:30):
undefined

Josh: man, these people are charging way too much money for this. They're driving up all the prices. (36:33):
undefined

Josh: Last topic. I thought this was really funny, EJ, as you put it in. (36:37):
undefined

Ejaaz: It's like, I'm a watch guy. Yeah, so basically, breaking news, (36:39):
undefined

Ejaaz: guys. Arguably the biggest topic of this entire episode. (36:44):
undefined

Ejaaz: Sam Altman just ordered seven ultra rare custom Swiss watches from this company (36:47):
undefined

Ejaaz: called Vanguard. are, I'm definitely butchering that name, with a branded OpenAI logo on the dial. (36:52):
undefined

Ejaaz: No commentary, right? First, let's observe. (36:57):
undefined

Ejaaz: Look at this masterpiece. For those who wear Apple Watches, I need you to just (37:01):
undefined

Ejaaz: take that thing off your wrist for a second and just appreciate the mastery (37:05):
undefined

Ejaaz: that is on your screen right now. This looks so cool. (37:09):
undefined

Josh: Can we take a second? Because I just so happen to have an Apple Watch in my (37:12):
undefined

Josh: hand. Please. It appears like the band on my Apple Watch looks oddly similar to the band on that (37:14):
undefined

Ejaaz: Watch absolutely and that is because the memory component the equivalent of (37:19):
undefined

Ejaaz: this is that complex cognitive function that is happening inside the thing who (37:24):
undefined

Ejaaz: cares about the plastic strap you can literally you see this little attachment over how many. (37:30):
undefined

Josh: Tokens per second does that generate (37:33):
undefined

Ejaaz: Oh god not enough not enough but um inscribed on the back of each of these watches (37:35):
undefined

Ejaaz: is if agi dot aligned deploy which was written by a programming language python dot Vanguard. (37:40):
undefined

Ejaaz: So the point is, I think this is like, you know, Sam kind of like giving a peace (37:48):
undefined

Ejaaz: offering before potential rumored. (37:52):
undefined

Ejaaz: IPO that I think is coming very soon and he wants to give it to his ranking (37:54):
undefined

Ejaaz: order of people that haven't left open air over the last two weeks. (37:58):
undefined

Ejaaz: Sorry for the shade. I've just seen a lot of executives leave recently, (38:03):
undefined

Ejaaz: but it's a beautiful watch. (38:05):
undefined

Ejaaz: People have been discussing what the secondary value of this might be. (38:08):
undefined

Ejaaz: And people are guesstimating this would be in the order of 500 K to a million (38:12):
undefined

Ejaaz: dollars. I personally think it would be higher because there's only seven of them. (38:16):
undefined

Ejaaz: And one of them will be worn by Sam's on Sam's wrist. So that's definitely not going to solve. (38:20):
undefined

Josh: This is kind of like um what it like the push presence like when when your like (38:24):
undefined

Josh: wife has a child it's like when your executives produce agi please (38:29):
undefined

Ejaaz: Please turn off now thank you well. (38:34):
undefined

Josh: This like i think this makes sense right it's like when your executives produce (38:36):
undefined

Josh: agi like hey guys thanks like here's a million dollar watch type thing um because (38:39):
undefined

Josh: i was listening to sam altman or i was reading the time publication of sam altman (38:42):
undefined

Josh: and what he was saying and they fully expect to have agi internally by the end of this year (38:46):
undefined

Josh: and for those that are not familiar we are about to enter september which means (38:49):
undefined

Josh: we are close to the end of this year so maybe perhaps he's getting these custom (38:53):
undefined

Josh: made in anticipation of their internal rollout of agi whatever they believe that to be (38:56):
undefined

Josh: um so that's just an interesting thing worth noting worth following lovely watches (39:01):
undefined

Josh: like hey the limitless guys i mean i'd take a half million dollar watches (39:04):
undefined

Josh: i think i'd make shows a million whatever it's going to cost that'd be cool (39:07):
undefined

Josh: but that's the round of this is a long one today we're like 40 minutes into (39:10):
undefined

Josh: this thing um so if you made it this far man (39:12):
undefined

Josh: Thanks for coming along with us on this journey. It's been a good one. (39:15):
undefined

Josh: A lot of stuff happening this week. (39:18):
undefined

Josh: The things are picking themselves up. The roundup's getting longer. (39:20):
undefined

Josh: There's more stuff to talk about. (39:23):
undefined

Josh: I'd say it's a big robotics week this week. A lot of robots. (39:24):
undefined

Josh: A lot of China talk this week. Good week for the Chinese. (39:27):
undefined

Josh: Whether it's good for us, we don't know TBD, but like good week for the Chinese. (39:31):
undefined

Josh: Good week for NVIDIA and Jensen. (39:34):
undefined

Ejaaz: Hey, it's good for Limitless as well. We're putting out so much content, so much to talk about. (39:36):
undefined

Ejaaz: What was the other thing? That was like the open rail acquisition that might (39:41):
undefined

Ejaaz: have been last week's news. Who knows? (39:44):
undefined

Josh: Who's on track, man? (39:45):
undefined

Ejaaz: We are covering anything and everything that is happening in AI and frontier (39:47):
undefined

Ejaaz: tech or whatever just makes, you know, Josh and I pretty excited. (39:51):
undefined

Ejaaz: And we're doing that four times a week. So if you haven't listened to any of (39:55):
undefined

Ejaaz: our episodes this week, please check it out. We also have a newsletter that (39:57):
undefined

Ejaaz: goes out to, I think it's like 110,000 of you that are reading our stuff every (40:01):
undefined

Ejaaz: single week, twice a week. (40:05):
undefined

Ejaaz: We do an essay and we also do the weekly roundup, which if you don't want to (40:07):
undefined

Ejaaz: listen to our beautiful, soothing voices telling you about AI, (40:11):
undefined

Ejaaz: you can just go read the quick summary digest we are covering you everywhere on spotify apple (40:13):
undefined

Ejaaz: as well as youtube if you aren't subscribed if you haven't left us a comment (40:19):
undefined

Ejaaz: if you haven't given us a rating what are you doing please do all of that, (40:21):
undefined

Ejaaz: send it share it with a friend it helps us out massively josh have i forgotten anything. (40:25):
undefined

Josh: I mean you can and i know this is kind of rough for me to say because i haven't (40:29):
undefined

Josh: been posting much but on screen our our handles on x are showing and you can (40:32):
undefined

Josh: go and follow us there for more information on the day-to-day you just i know (40:36):
undefined

Josh: you're always firing off stuff about updates so that's just like a fun little (40:38):
undefined

Josh: call out if you've made it this far. (40:42):
undefined

Josh: And, you know, you're interested in hearing a little bit more. (40:44):
undefined

Josh: I'll work to post more on that platform. (40:46):
undefined

Josh: But that's it. Like you can go feel free to like you're dismissed. (40:48):
undefined

Josh: Go touch grass. Go enjoy the weekend. (40:51):
undefined

Josh: Get off your phone for a little bit. You're now fully up to date and all of (40:53):
undefined

Josh: the things. AI frontier technology. (40:55):
undefined

Josh: You're in the loop. You can go out to your like weekend barbecue. (40:57):
undefined

Josh: I mean, close to the end of the weekend. We're almost we're almost there. (41:00):
undefined

Josh: I don't want to think about that. (41:03):
undefined

Josh: But yeah, that's pretty much everything. Thank you so much. I mean, (41:04):
undefined

Josh: share with your friends if you if you got some that would be interested. And yeah, enjoy. (41:06):
undefined