0:01
YOU'RE WATCHING TVN >> TVPN.
YOU'RE WATCHING TVN >> TVPN.
Today is Tuesday, November 3rd, 2026.
We are live from a combination of the TVP Ultra Dome and a hotel room hundreds of miles north in San Francisco, but we are still in the Temple of Technology, the fortress of finance, the capital of capital.
And of course, we're going to tell you about ramp. com. Time is money. Save both.
Easy to use as corporate cards, bill pay, accounting, and a whole lot more all in one place.
And let it let's let let's let this San Francisco Halloween goer tell you about ramp as well because we got a video from >> pull it up.
>> How Let's pull up this video of someone with their Halloween costume.
>> Time is money save both.
>> Oh, I like the SD temple of technology.
The fortress of finance, >> the capital of capital. >> The flash. >> Great.
Some sort of combination Halloween costume plus vibe reel.
Really got the full experience out of that costume.
Not just showing up, hoping somebody else takes a picture.
Clearly shooting some A-roll, adding some B-roll, getting a good good viral video. >> So good. So good.
He's getting He's getting a pack of real merch.
Uh, John, how do you think the delay is right now?
Can you think you think it's like 5 seconds, 10 seconds? Probably hard to say. >> Couple seconds.
I'm talking right after you end.
How does it feel on your side? >> It feels okay. >> Okay. Okay.
>> Would you look at that? >> Yeah. Uh, yeah.
How how was how was your trip home?
You got home late last night?
>> Uh, I I I slept uh five hours >> on the plane. >> On the plane. Yeah. >> Wow. Very nice.
I slept on the plane, too.
And then that kind of threw me off when I got home.
I couldn't sleep and then I woke up and slept a little bit. >> Skissue. Okay.
Well, >> um, >> let's let's move on to the news.
First, let's tell you about Railway.
Railway is the only one intelligent cloud provider.
Use your favorite a more.
Um, so, uh, the the the big news, Apple and Palunteer had earnings.
We are in the midst of tech earnings season.
Apple and Palunteer they both beat on earnings.
There's a lot of this happening where uh the market is in such in such turmoil that even though uh earnings are coming back positive uh analysts are reading into uh margin compression that may or may not happen in the future and that's driving stocks uh to be a little bit weaker sometimes sell off.
Kind of depends on the stock but Apple reported record September quarter revenue while Palanteer nearly doubled its revenue from a year ago.
Both stocks are down around 2% this morning.
The Financial Times reported today that even though Apple beat earnings, the company is warning that rising memory chip prices and the cost of running its AI tools could cut profit margins. >> You don't >> what? >> You don't say. >> Yeah.
I mean, stuff's going to get expensive and Apple's had a long history of holding the line on pricing as much as they can.
Like they love to put up a big clean number and then stick to it for a little bit.
And so it can be they're going to be in a jam as uh as memory prices go up midcycle when they don't want to change prices because it's never been a company where they change prices every month, every quarter.
It's been, you know, every every year or two or or even even every three years they'll they'll hold the price hold the line in the price for a really long time.
Uh uh anyway, um Palunteer uh Q3 revenue came in at 2.
3 billion annualized at 9. 25 billion.
It's a $400 billion company.
Uh and Karp had a had a uh he he he had a big moment on the earnings call.
Uh US commercial revenue surpassed its US government revenue for the first time reaching 984 million versus 953 million.
versus 953 million. This is sort of the the the the dual buckets of are you betting on on the the foundation labs the openthropic combination or the Microsoft Palunteer the diffusion the enablers the cognitions and uh Palunteer has has certainly been growing like they their
messaging is very similar to Microsoft now around uh you know own your data let's work with you to implement AI we'll help you transform your organization uh And so, uh, I mean, it's been been receptive and that's why they're, uh, that's why their commercial revenue is growing. >> What are you laughing at, Drew?
>> What are you laughing at, Drew?
>> Uh, just you just laughing at you. Um, no.
Uh, no, no, I was I was sort of smiling because it's like, you know, put putting putting Karp and Satia in the same bucket.
It's like, you know, uh, very different personalities, but ultimately have a very, uh, similar sort of, um, pos, you know, sort of pitch. Uh, which is cool. >> Yeah.
Chat is asking about merch timing.
Merch store wanted to jump in quickly because we do have a date.
Uh November 12th, the TVPN merch store is going to be going live.
Uh it will be released first into the YouTube chat.
Uh and it will sit there.
I'm I'm assuming we'll we'll we'll leave it there for like 10 minutes so the chat can uh can feast and then we'll post it uh publicly over on X and everywhere else.
Yeah, a lot of work, a lot of work went into that.
Um, let's go over to Buco Capital Bloke breaking down Reddit.
Uh, he says, "Redic call was absolutely fascinating.
Everyone should listen to it as it's a bellweather for where the internet is going.
AI mode is increasingly the default Google experience.
AI mode provides rich info and data in line.
Less insensitive to leave search or less incentive to leave search.
AI mode incentivizes dialogue.
Google wants to keep wants to keep you engaged to serve more ads at each turn of the conversation.
Google will milk that for as much revenue as they possibly can and transition from a search company to a cloud service and chip company as traffic from humans becomes more scarce.
This will put Google in an increasingly adversarial position to the rest of the internet. That's interesting.
This is a massive headwind to all media type companies who are not destinations.
This is a massive headwind to all content driven gotom market motions.
Especially in the SMB, you have seen companies like HubSpot and Monday.
com get crushed by this funnel collapse.
We are entering the endgame, a knife fight for the remaining and dwindling volume of human traffic on the internet. Interesting.
Um, they say people still want Reddit.
This is from Steve Huffman, the CEO on the call.
People still want Reddit.
It's just Google is not sending them to Reddit the way they used to.
What's interesting is like Google Zero has been, you know, coined by Neil Patel at the Burge like years ago.
I I don't know five years ago, 10 years ago, something.
He's been talking about that for a long time.
Uh when Roger from Kadas came on the show, he talked about how they had embraced it there and that was sort of a a bombshell to hear uh you know, someone in that position kind of admit it publicly that that's no longer a growth engine for them.
Uh it's uh it's it's interesting to hear that it's spreading to Reddit.
Um but uh it's certainly it's certainly uh uh crazy times for Google because they they it's the innovator's dilemma.
They have to meet the customer where they want to be with the new product.
But at the same time, a lot of people sort of hate AI mode or at least publicly hate on AI mode.
Like the the the the long history of just Google stuffing more and more ads in the search results never really broke through.
But now people are publicly saying, "Oh, I don't like the AI mode.
I want to go back to the the one that monetizes better," which is kind of this funny dichotomy of um >> for the first time just like just randomly anecdotally.
If you talk to people, they'll they'll be like, "Oh, I'm like frustrated."
And it can be kind of an annoying funnel because you you wind up in AI mode and then you're like kind of trying to click somewhere and it can be sort of jarring.
It just requires like a new a new flow to like get into. I don't know.
Um, what what what what's your overall thinking on Google these days, Jordy?
>> Uh, Dylan Patel, >> I just think I just think if if you think AI I I think if you think uh AI is real and going to be big, like it's it's very very hard to make a super coherent bare case, right?
bare case, right? even if search revenues even if search revenues >> decline >> um you know every year and and we're sort of at peak search they just have so many other assets and they can use that um you know search cash flow to finance so much other infrastructure and M&A and things like that so um I mean people are
nervous about it but it's still $200 billion of effectively I mean that's revenue but it's high margin revenue that they can deploy elsewhere and uh I mean the take him uh take is that in the race to win AI builders, Anthropic and Open AAI now account for 86% of enterprise LLM API spend up from 67%. So
So the duopoly is consolidating around the frontier leaders and Google's market share went down more than 50% in terms of not in terms of LLM usage for consumer but but for builders in the enterprise.
Uh that's >> that revenue is flowing through GCP.
It's it's it's paying for TPUs, right?
And so it's like Google's getting their Google's getting their cut, right? >> 100%.
>> Um and and yeah, it's like you just cannot >> it it's when you look at the sort of collective, you know, the collection of assets, the exposure to everything from, you know, SpaceX to to Anthropic to the TPU business to the cloud business.
It's like, you know, this is still one of the greatest businesses >> uh in history.
>> Where do you sit on the question of like, is Google in founder mode or not?
>> Because like we always get these headlines of like, oh, Larry and Sergey came back in.
they're they're digging in and then you'll see like they bought another estate in Florida and and it's like going back and forth on like who's actually and I know they have the voting control so they could at any point come in and be like we're doing some huge M&A deal or we're completely reorganing this organization.
We're going to completely we're going to go Mark Zuckerberg mode and completely rework the organization.
Um, but at the same time they have been out and had a real CEO in place unlike Meta.
So they're kind of this hybrid like they're not quite >> it doesn't feel like it doesn't it it certainly doesn't come across like the company is in founder mode in the way that you in in the way that you feel it with uh a meta, right? Yeah.
>> Um and so and I and I think that's what matters, right?
I think um the big obviously you know the headwinds for Google is like losing >> um you know people almost forget about the massive talent exodus from like what a few months ago right? >> Yeah.
>> Um >> and um but but again it's like as long as they can stay competitive around the cloud and and chip business you know there's a lot of upside.
>> What is going on with this New York Times headline?
Google will turn off search for schools that reject AI.
Andrew Curran is sort of correcting the record here, but uh it it sounds really aggressive.
It's like the carrot of the stick. Did you see this? >> Yeah. Yeah. Yeah.
Um I mean they're just saying they're they're effectively saying like if you >> we are every all of our products are AI.
So if you tell us not to use AI then none of you can't use any of our products, right?
But it sounds much more aggressive to be like you can no longer Google search.
But I think their point is that >> AI is >> everywhere across every product surface area. >> Yeah.
>> And so it's unavoidable.
>> So Andrew Curran says, "This article is getting posted everywhere this morning. I can see why. Very very controversial.
Does but this decision is being incorrectly framed as punitive.
There's no such thing as search without Gemini.
Even if you don't turn AI mode on, you're still talking to Gemini in the search box.
The old Google search is gone.
Yeah, that's it's a good point.
Like I I I'll give Google some points for, you know, this is just the way things are now.
But it does sort of yeah, necessitate like a Google search classic or something.
I mean, I guess schools can do duck.
go O or some search engine that hasn't embraced, you know, LLM search if they want to offer that that more classic uh old school experience. But I don't know.
Might be time to learn the Dewey Desimma system. Hit the library. What do you think? >> I think I think so. I think so.
>> Google is not going to save you at this point.
If you're trying to learn, you gotta you got to hit the books.
Uh I don't think you're gonna What's going on with Grockbot?
>> Let's tell everyone about Crowdstrike.
Your business is AI, their business is securing it.
Crowd Strike secures AI and stops breaches.
The Grotbot thing was very weird.
So Elon chimes in on November 1st, which was what, Sunday morning, I guess, or Sunday afternoon.
Says the cost of serving Grockbot was unsustainably high relative to revenue.
She says, Elon says Grockbot costs are unsustainable rel relative to revenue.
Computer use is expensive, but AI companies are racing to make it cheaper while giving it away for free. Thank capitalism.
we get free digital labor before they figured out how to profit for it.
>> Um I'm I'm I was so surprised by this.
I think everyone was surprised by this. What was your take?
>> Um so the I I think the current read is that they were using Opus 55 under the hood. >> Oh yeah. Yeah.
They're a big customer too.
And so um but but still like you would think that they would just eat the eat the losses for some amount of time given that uh given that they you know they're a$2 trillion dollar company.
They're like I think somewhere around a million you know uh weekly activives.
Like it's not a >> it's not a massive um uh product at this point.
And so this whole thing was just like very I I still don't know what the read is.
Like I'm just very confused.
>> I mean I'm just doing it back of the envelope like like >> like he's he's going and publicly saying that I believe that in the future Grockbot will consume the majority of our compute. >> Mhm.
>> And then now you're saying like the cost of serving our product is unsustainably high relative to revenue. Thanks.
>> I mean, I guess the steelman would be like the next version of Rockbot, people will be paying more for so even though it's consuming all the new data center compute capacity.
>> No, but but but like assume assume the sort of continuous decline >> in the actual cost >> to serve your users.
Why are you what like >> maybe maybe he was having a big night out? >> I don't know.
Also, it seems like I mean it's not like he turned Grockbot off.
He just This is just he just denied someone a reset, right?
Wasn't that what happened? >> I know.
I think people were complaining that um >> there's not their usage just like rocketed to 100% and then Sure. >> It doesn't.
>> So, they kind of rightsize the the the tokconomics internally.
Yeah, it is odd because even if they're like 10% of Anthropic's revenue, that's less than a billion dollars of cost.
And if they have, you know, a million people paying a hundred a month, that's 100 like they're they're lo they'd be losing a lot, but they have a lot of money.
But I guess uh they just, you know, Elon's definitely one to rip the band-aid off and kind of rightsize things and maybe just wait.
Also, I've just been surprised at the uh the level of adoption at least the level of uh maybe it's just on X, but people talking about actually using it.
Uh it it it precursor the enterprise the power user go to market motion was uh a little confusing.
It made a lot of sense to stuff Grock in X as like this helpful little answer bot for community notes and questions on facts and stuff like that when you're see when you're reading a news article having Grock summarize it.
that seemed like a good use case.
But um the idea of like bridging that bridging from that like the X the everything app that's like mostly for fun and mostly for like gossip to like it's in my Slack now uh was always a little bit of a stretch uh versus like a Palunteer or Microsoft or you know any of these other enterprise products that have more mature go to market motions.
Anyway um >> what's um how's Whiz Khalifa responding to the situation?
Whiz Khalifa had some insight, some he had some wisdom.
>> He sort of fired back.
>> The work is the work is the reward.
>> The work is the reward.
>> I think that's because he he he ran out of usage.
>> He hit his usage limit on Grockbot and so he's sort of reminding himself >> the work is >> I got to do the work now. Yeah.
My Grockbot can't handle it, >> but the work but it should feel rewarding.
>> I like I like a piffy quote every once in a while.
I I haven't seen him take over the timeline for a while, but >> Naval has been awfully silent since Whiz Khalifa posted this. >> True.
Uh here's an exercise for the viewer.
figure out where I am because apparently you can listen to my audio and you can run it through some analysis and and hear the various planes with their Doppler signatures, the Doppler curves of nearby airplanes and you can triangulate where I am.
I already gave you a clue.
Um but but I I do wonder how detailed this needs to be.
I've heard that it's like 15 minutes of of of clean audio.
If we're passing this through this call and there's noise on your side a little bit >> and I'm assuming you're on a Zoom call, so there's some noise cancelling happening naturally. >> Yeah. Yeah.
It's not like a clean like file that's continuous on my side coming through to you. >> Yeah. Yeah.
We're going to need a lot of soundboard to keep me anonymous. This is the new cyber.
>> I'll just keep spamming it.
Just spam that cat because we got to confuse the people that are trying to track me down using the latest.
No, this is this is a crazy example of like uh AI enabled surveillance. >> This is like Yeah.
Everyone is going to have access to super intelligence.
So, you get to do things like this. >> Yeah. Yeah.
Which is, you know, >> Yeah.
I mean they genuinely for for people that have recorded like long form vlogs outside at their house unedited there's no cuts in the audio like this definitely could dox your location but geogesser ge rainbolt has been able to guess locations for the the old way analog way just by looking at the photos and looking at Google maps for a long time.
So, uh, security was always a layered a layered problem, but, uh, this is this is a fascinating just mathematical, uh, breakthrough or or or situation.
Um, what what else is the time?
>> Let's check out this uh ad from Ool Capital.
>> It's just um just a picture of this guy. >> I love it. >> Then learn more. Learn more. >> I'm curious now. I'm curious.
Um, Thomas says, "Hey, Meta, waste my money."
Because they'll, you know, they'll of course offer you >> the ability to kind of just promote any any kind of content.
They make it really really easy.
You can do it with a credit card in the app. >> Not as easy as X. X has a finger.
>> They have the guy that Yeah.
>> and points to the promote button, which is going to be bigger than the post button soon.
the I I had to set up X again and the Nikita of the app is remarkable.
Like if you don't enable notifications, it it has this like this the slider bar that's like notifications on or off.
And if you say no, I don't want notifications, it slides into a state that looks like it's broken and looks like you have an error in the app.
Like it's it's genius to get you to like be like, "Oh, I got to correct this."
It really there's so many different patterns to get the uh the the the retention up for sure.
They're they're working on it.
Anyway, let me tell you about Codeex.
Codex is a powerful workspace we're going to work on with AI agents.
Whether you're writing code, analyzing data, creating content, or automating business workflows, Codeex helps you move projects forward from start to finish.
Uh we have a we have a great show lined up.
We have Harley from Shopify joining to share Shopify's Q3 results in just two minutes.
Then Bridget Mendler from Northward Space is coming on 11:45.
And then Dylan Patel and a semi-analysis crew are coming on 11:55 to break down GDPX.
Um there was a there was a viral picture of a cat that was born with a mullet.
It was unfortunately uh AI generated, but uh >> brutal.
>> I think with AI is both the problem and the solution here because AI is the problem that generated the cat with the mullet.
But with advanced breakthroughs in biotechnology driven by advanced AI models, it might be possible to breed a real cat like this in real life. >> Yeah.
A lot of biologists are going to be looking at this. >> Yeah.
>> And saying, actually, I was focused on, you know, extending human lifespan, but I'm actually shifting focus to, >> you know, developing the mullet cat.
>> I could see a new startup.
You know, we have that Palmer Lucky uh you know, uh Warhound for the for the for the dog for the war fighter.
>> Mullet Cat, new startup.
Send it to a billion >> for sure. >> Um >> Peter Rahal. >> Yes.
>> Back with a new brand called Rowdy.
>> It's a absolutely fantastic name.
He's coming after >> Doritos.
Uh it's his holding companies.
Uh, Andrea shares it's his holding company, Medici's first savory EPG snack, and it's pretty addictive.
Their tangy ranch is the closest thing to reverse engineering Doritos Cool Ranch.
It's going to be interesting to see how Fredo reacts.
Um, really, really good execution on this.
Um, you can just tell already that it's going to be a hit.
I mean, this is, >> you know, the the Doritos a Cool Ranch was literally created in a lab to be the most >> uh the most addictive uh product uh snack in the world.
And uh I'm I'm almost certain that that Peter would have replicated it pretty perfectly.
>> Yeah, you are you are you you're you are a Doritos guy.
This is this is things that you know something that people don't really know about you, but um you you you clearly will be in a situation where like the only food available is Doritos and you're like twist my arm, twist my armist, but you'll grab like three bags and then you'll just go >> I would go further.
I I would say I live sort of a Doritos coded lifestyle.
>> I I would say that I my entire life is Doritos coded in any way. >> Absolutely.
Well, uh I I I love this because I I think uh he's not overthinking these categories that he's going after.
Like clearly there's a market for Doritos, those particular flavors.
There's so many other things if you were trying to go into like the new chip category.
You could go way too protein forward, try and change up the flavors and just make it a little bit too confusing and not legible to the Doritos customer.
this is something that a Doritos customer could see and just pick this instead.
Um, and so I I think that there's this is a good like counterbalance to the war that's happening in the rest of the new healthy chip category.
This is just like one click away from Doritos.
And it's more like Doritos than like Masa, for example.
That that that was the that was chat. >> Yeah.
>> Um before we bring in Harley, let's pull up this video >> of uh LeBron's newest sponsor. >> Yeah, this is huge.
>> I just went to the 76ers in Philadelphia.
Did you see the new sponsors of the 76ers?
>> Wait, >> it's on LeBron left breast.
>> Don't you Is it you or Bloom Energy?
>> I am Bloom Energy, but that is a different Bloom Energy. There is an AI company.
I think they paid like $30 million to sponsor the 76ers.
No one other than you finance nerds know what energy is.
The amount of messages I got that thought I just signed the 76ers is the funniest thing that's ever happened to me. >> So good. >> So >> so lucky. Yeah.
It's better to be lucky than good. I don't know.
Greg is very good as well.
But it's just like >> mandate of heaven.
>> Bandaid of heaven moment. For sure. For sure.
Let's bring Harley in first.
Let me tell you about public invest across all asset classes.
Automate anything with AI agents.
Run trading strategies, risk management, money movement, and monitor markets with AI agents for investing on public.
And we have Harley from Shopify here to tell us about everything with with Shopify's Q3 earnings, the health of the e-commerce economy.
Let's bring in Harley and catch up with >> Boom. >> He's back.
>> J, great to be on the show.
Um, I think are are you are you virtual? Are you local? >> I am virtual. I'm in San Francisco. I'm traveling.
So, I got the fake background to cry and kind of >> I think I I think it actually plays really I was going to say if there's no gong, I'm out of here.
>> No, no, Jordy is We got We got you covered. We got you covered.
>> Uh, give us the first gong worthy number. I want Jordy to warm up. >> Okay. All right.
Well, let's let me give you three um because uh three these are by far the most important. GMV 119 billion. That's up 29. 5%. Revenue was 3. 7 billion. That's up 32%.
and free cash flow margin was 22% which >> you guys know this at our scale is is remarkable I held out some massive companies massive brands this quarter uh Mac Clinicique Balain Lululemon all move into Shopify we're seeing more and more of this incredible you know commerce
happen on on aentic services I think Q3 was really the era of like these new personal agents which we've embedded with all of them we're beginning to see real traction on the agentic Um, but this is this is, you know, this is the best of Shopify. >> Lululemon, that's a huge one. Is is AI
>> Lululemon, that's a huge one.
Is is AI accelerating the pitch for a big brand like that?
Because I imagine AI is incredibly useful at replplatforming, moving things around, uh, taking product detail pages, translating them, actually just doing all the grunt work that might be required to change e-commerce providers.
I think every retailer is looking is being asked by their board, hey, what is your AI commerce strategy?
What is your agentic shopping strategy?
And one of the things that I've said to when I meet these brands, you know, even some of the, you know, bigger brands or or older brands like, you know, Estee Lauder or Birkenstock or, I don't know, um, Hunter Douglas, is I will say, look,
ultimately what I think you've all had to do for the last 20, 30 years is you've had to not just be an incredible retailer and and a brand producer and a product producer, but you also had to effectively become a technology company. You had to understand like you know
You had to understand like you know online stores and you had to understand servers.
You understand data rooms.
You have to understand you know how this all kind of fits together and and you many cases you even had your own like onrem servers.
I was like I think that era is over.
I I think the era is now get back to building incredible cosmetics or great you know yoga uh apparel or whatever your product is and let us handle the infrastructure around all all of commerce.
I think the agentic everyone's sort of thinking about it.
I think the numbers are actually I mean it's still on a very small base because it's so new.
You know, we we did $400 billion last year of GMV total.
So, it's still a fairly small base relative to overall GMV, but traffic was up 2. 6x year. Orders were up 3. 3x year.
Uh and and more importantly on the conversion side, if you just compare conversion using Shopify catalog versus general AI search, which is effectively scraping, it's 2x higher.
And even in terms of what they're seeing from some of these like they're seeing more more more new customers come on.
>> But the real winners of of AI so far has really been kind of the long tail.
These you know like products that are outside our top 100 categories and we have a ton of them on Shopify.
>> Uh >> I completed my first Oh, sorry. You go. >> No, you go. Go.
John, I I can't wait to I completed my first agentic purchase on Shopify. >> What was it? I I bought a raincoat.
I It was raining and I told Dot uh in Chachi BT, I need a raincoat.
Go find it, buy it, send it to me, just do everything. And it used Shopay. And it has arrived.
It's >> It's great, right? It's it's so seamless. It's so simple.
seamless. It's so simple. One thing though that I would say John that people are missing about the Shopay integration whether it's Muse or Instinct or Claw 2 or Grockbot is that by using Shopay obviously the payment the transaction it's simple it's safe it's fast but as part of it also when you connect shop to these these personal agents your entire
shopping history the agent now understands they give way better recommendations obviously the same payment details with shop pay but also you now can track your package through the shop shop app as well so this idea of actually using Shop pay almost as an identity system for these agents mean that no matter what agent you use if shop is and shop pay are connected to it. It actually has a perfect
It actually has a perfect understanding about your buying. >> Yeah. Yeah.
I feel like uh we spent all last year talking about like are you seeing agentic commerce yet and it feels like we're still in the early adopter mode for sure but it's starting to be real.
I'm hearing more stories.
real. I'm hearing more stories. everyone has it access to at least a technology that can functionally do it whether they've leaned in and built the >> some of them are just using it still as discovery right I think I think BFCM 2025 which we spent we spent that day together um >> it was really about discovery I'm looking for new products I think now we
sort of move from just discovery to discovery and just buy from me >> like delegated delegated purchase >> yeah no it it actually is >> so so what what are >> what do you think's actually going to happened with this category because like I you know an agent cannot save me time versus just doing it myself with how good shop pay is like I'm like >> yes >> I think I think you go back to to Yeah. Go ahead. Go ahead. >> Yeah.
I just as somebody who's like very AI pilled but almost equally Shopify pilled, right?
pilled, right? like I've just been been buying things with Shopify for so long and it feels like every single year >> it gets like you know like significantly faster to complete a purchase like both like the >> I'll give you the example where I think um where I think it is where I think you'd use just go to the online store
you shop pay it's fast and easy versus John's example of of of the raincoat um if John had a favorite raincoat company I think he would have gone directly to the site in a matter of seconds would have checked out using shope I think because he doesn't necessarily have a specific company that he wanted to purchase from, there was a there was an element there of research. If you if I
If you if I look back on like the last bunch of of transactions completed by personal agents, it's companies like Robo Rock, which robot vacuums are everywhere. Everyone has one.
But this is specifically for a particular type of pet hair >> or We Things, which you guys know the brand like it's it's a smartwatch.
does not look like a smartwatch or a car seat for three.
This is uh literally a transaction happen.
A car seat for three kids.
A car seat for three kids that fit in the back of the so three car seats stacked.
>> This is very different than knowing I'm going to go to this car seat brand by it. I wanted to do it.
There's an element of discovery here. Yeah.
Whereby the AI whereby the agent can actually digest a billion different SKUs, look at product attributes, cross reference it with what everything it knows about you from shopper or elsewhere, and then say, "Hey, here's what you want."
And I think for those types of purchases, this gets into a really interesting new area. >> Yeah. Yeah. Yeah. Okay.
I think that's a that's a very that's a very helpful way to think about it because personally I would never just say like I need a rain jacket. Get me a rain jacket.
>> I would I would I would go to a handful of my favorite brands and >> I don't for the direct to consumer like the cool one like that.
Yeah, there there are there are a bunch of other products where that are not like brand driven purchases to me.
Uh that don't you know uh I'm now I'm so curious to John.
You got to wear the rain jacket tomorrow whether it's raining or not.
I I got to I got to see >> your jaw's going to be on the floor.
You're going to be like that's the best raincoat I've ever seen. >> Yeah.
I I I will though give a shout out though to you know to those that are are listening to the show and I know we have so many Shop merchants that watch your show religiously who are thinking about preparing for uh Black Friday, Cyber Monday.
What I will say to them is to the conversation we're having, it is really important that you look at the knowledge like the knowledge center inside the Shopify admin to figure out like what are agents seeing about your products because um remember like we have syndicated all of your products into catalog.
The catalog was built for agents so they can sift through everything real-time inventory, the right product attributes and details.
But if there are like details of your products that you are not telling it.
So if this rain coat is really good in in New York City but won't work in Toronto because it's too cold, that is a product attribute you should mention like this works better in cold weather.
Therefore, you will be found more if the consumer is searching for that.
That is a very November 2026 recommend suggestion that was never actually that important.
Um because eventually consumers can kind of do their own research.
But now it matters more than ever because the agent's going to surface whatever is the mo the most merit-based product. >> Yeah.
The big thing is uh measurements.
I mean, for me, like I I'm 6'8 tall and like have specific proportions, but I've given my agent >> Yeah, flex.
But I've given my agent like every possible measurement and so it can go and sometimes it can look for this particular size and then translate.
Maybe it's in centimeters and it needs to get over to inches or something.
It can do imperial metric trans translations, but some fits are different.
And so vending that to the agent in okay here's our sizing, here's how we benchmark against European standards, but then also here's almost like a 3D model of this item.
Here's here's the exact measurements.
And then the agent can translate that to match the human that's on the other end.
>> One of my favorite Shopify merchants is a company out of New York called Gold Belly.
Do you guys know Gold Belly? >> They're amazing. So I live in Canada now.
Uh but I love I mean you know I grew up in the US so I love getting like cats's deli delivered to to Montreal uh and stuff like that.
It turns out that some products are easy can go over the border easily from Canada from the US to Canada. Some cannot.
Um gold belly needs to be specific to say like this type of fish doesn't like because of some duties and tariffs doesn't work but this type of fish does work.
Again, that was something that you probably thought about in in, you know, in your um uh you know, it's some sort of uh line in your checkout where like, hey, like this won't work there.
Now, you actually have to merchandise your products to be able to say this ships crossborder. Again, very very new.
I I think what's also interesting is on the call today, I got most of the questions were really about um aentic of course, but it's interesting because the reality of what most merchants on Shop I need right now is actually like point of sale. It is international.
It is like B2B functionality.
It's sort of like the basic stuff.
And so I think at Shopify right now, we're at a very interesting time where we're we're sort of operating in two realities, which is building the future of commerce that people are excited by like Agentic and at the same time like actually powering like full stack proper infrastructure for physical retail or for you know for for things like that.
It's it's it's fascinating to see how different types of demographics care about different things right now.
And as shop, we're 14% of US ecom.
We have to sort of do both. >> Yeah.
>> You're the reason why Catz has a huge line.
We drove by Cats's Deli yesterday in New York and it was it was a line down the block and now we know >> great spot. >> It's a great spot.
>> Uh question on agent monetization.
So, I'll give you the two situations.
If I go to my preferred agent and I say, "Hey, I want this rain jacket.
please like go buy it for me and make sure that it shows up at my house, right?
The the the merchant is not going to have any desire to like pay any type of referral for that because the demand was already created like they did they they figured out how to reach me and convince me to buy their product some way or another, right?
Attribution is hard, but that they created the demand. I was on their website. I I found the product.
in John's case where the agent did the research did did all this discovery based on John's preferences and then John goes and buys that product and effectively like created the demand for that product.
There's a very strong argument in my view that that the merchant should pass back some of that revenue to the agent for basically generating that lead.
But >> yeah, I mean that that may happen eventually.
Right now the current economics remain completely unchanged for Shopify merchants like agentic transaction carry the same fees the online store transactions do.
If the transaction flows through Shopify payments the same economics apply.
So from a merchant perspective nothing really changes.
If you look across like Muse or Instinct or Google AI mode or Microsoft Copilot and GBT um you know like Shopify merchants default to Shopify payments.
Shopify payments. So the merchant should have you know it shouldn't make any difference whatsoever where it may eventually go and I think I talked about this last time I was on the show is what happened with the gentic ads because I think ultimately aentic advertising will be one of the ways why these how
these agents actually monetize sending customers to your store and I think the advantage in in aentic channels is actually relevant so I can I can sort of see a time and we've already integrated with um with chatgbt ads for example We announced in Q3 that all Shopify merchants can use the the chat GBT ads app to create manage campaigns right on there as well. Um but we're not assuming
Um but we're not assuming that Aentic shopping stays adree.
What we're trying to do is making sure our merchants participate in it.
So there may be a time where John buying his his coat if he actually, you know, does the search, has a conversation with his personal agent and it shows them two different things and he clicks on the one that actually is sort of in the sponsored bar that that merchant has to pay some sort of commission to the um to to the application. That hasn't happened.
I think it's going to be like Google where you've always had ads but most people click on the organic searches because it feels more legitimate. It feels more organic.
legitimate. It feels more organic. But I guess but I guess what I'm what I'm getting at is like I still think that in some ways it I think it would be relatively fair and I don't think merchants would have a big issue if agents are creating all this demand >> as long as they're showing as long as
they're willing to pull from the entire Shopify product catalog and they're not just picking and choosing right like if it's something that a switch you can toggle on on or off but >> um >> so I think if it's merit based I think you're right I think consumers is I think merchants are used to paying for traffic paying paying for for customers. I think if it's done I think right now
I think if it's done I think right now best I can tell the AI applications want to get as many consumers using these AI applications as much as they can.
So John having an amazing experience with the rain coat means he's probably going to come back for the next time.
I think right now that is what their focus is. >> Yeah, absolutely. Um >> well great stuff.
Thank you for coming on Massive Road to Road to uh BFCM >> BFCM.
It's going to be >> it's upon us. It's upon us. >> I can't wait. >> I can't wait. Um, awesome. Cheers, dude. >> Talk to you soon. Goodbye.
>> Uh, let us tell you about Cisco, critical infrastructure for the AI era.
Unlock seamless realtime experiences and new value with Cisco.
We have Bridget Mendler from Northwood Space back on the show.
We talked like a month ago. There's a new fun. Always a joy.
Bridget, welcome to the show. >> Out of control. Out of control.
You know, we got to we got to keep the regular cadence of conversations.
So, I got to I got to keep doing things. >> I know. >> Give us the round. Give us the details. The gong's warmed up.
I want you to >> rock on.
Um yeah, we just finished our $375 million series C fundra.
So, that's uh >> Thank you. Thank you very much.
>> And we talked last time, but uh what uh what what what what was the key unlock?
Is it is it technology milestone, financial milestone?
What really unlocked this particular fundraiser?
>> I mean, we're asking ourselves a series of questions.
The first one is like, can the tech do what we say it can do?
So, we have the two products.
There's portal, which is our C2 product, and then there is Prism, which is the high throughput product.
We've now validated them across not just LEO orbit, but all the way up to geostationary orbit.
And the real thing that matters for us is that capability demonstrating it, but also on a timeline that is relevant.
So we were able to turn that capability around um get it fueled in a matter of days.
So it's basically you know can we perform at speed and can we um actually hit the performance specifications that we are are saying we can do and then at that point it's a question of scale.
Have we kind of proven to ourselves that we can actually scale this product?
Um we have produced at this point 29 units of hardware across the products.
We've done two production runs.
we're now in our bigger scaled up facility 185,000 ft that's ready to produce hundreds of antennas.
Um and so we we kind of convinced ourselves we're ready for that scale and then ultimately it's been you know a tremendous year for the space industry in terms of just validating the original Northwood premise which is the volume and data is going to go up by a lot.
So um yeah we were ready for the next milestone. >> Super impressive.
And then uh demand like demand side like what's going on on that front?
Like I'm assuming a lot of it's still like kind of hey if we do this thing like will you give us a lot of money and so far it seems like the answer is yes but but break that down. >> Yeah.
I mean there there's the whole range from a lot of net new signed customers as well as um exactly what you're saying which is huge amounts of demand as new customer sets are examining space fresh for the first time on the commercial side um but also on the government side just huge waves of of funding and programs that are becoming more solidified.
So um all of those things at the same time um and a lot of different customer priorities but that's what we're you know built to support.
How are you thinking about the the breakdown of different opportunities in space?
Uh obviously there's a lot of satellite internet traffic. That's so exciting.
There's there's imagery that's happening.
There's also AI potentially happening, but unclear how bandwidth heavy that will be.
And then just uh yesterday uh SpaceX announced Starfall doing uh microgravity experiments for bio.
I have no idea how bandwidth constrained those types of experiments are, but what's on the horizon that you think will either be uniquely unlocked by Northwood or just be another tailwind for you?
>> Yeah, I mean, exactly like the industry is diverse and frankly like that's what we think is really important and what we're solving for.
So, the way that we've designed our product is so that we can fit as many industry use cases into Portal and Prism, our our two product lines as possible.
Um, and so when we're looking at those use cases that are coming up and are are interesting, the kind of like overall uh customer requirements that or or customer features that align with us is is it a customer that cares about scale and is it a customer that cares cares about reliability?
Like these are customers that are managing constellations that need to um hit that scale on aggressive timelines but also really assure high levels of uptime that historically you know third party ground has not been able to offer.
So that is things like comms.
Um comms is is a use case that has a particular amount of pressure there and I think as you said pushing into higher and higher throughputs is a really exciting problem and and for me I think that's especially exciting as we talk about overall volume of data in space becoming more important.
Um but use cases like you described whether it's Starfall or or others on the actual lower bandwidth side are also tremendously exciting and it's also a very key national security priority. Um, so yeah.
>> Uh, on the national security side, what's the supply chain look like?
I was I was reflecting on the fact that the the new iPhone, I think we're on 18.
It's been almost 20 years.
I think this one is the first one that has an Apple modem in it.
They finally moved on to their own uh their own hardware stack.
And I'm wondering like are are you supply constrained in any way or are you thinking about I don't even know what goes into building a ground station.
Are there rare earth magnets or is it just metal and aluminum or you get the licensed technology and IP from like Qualcomm or some other company?
Like what does that supply chain look like?
I imagine you're not bottlenecked right now because you're still growing the company, but what's coming down the pipe?
>> I mean, I don't think there's a single hard tech company that is not supply chain impacted.
Um, it's just like what are the strategies that make sense for the company to uh address that?
And I'd say that's where the the new capital comes in.
Um we're at a stage now as a company where we're making you know different sets of decision calls about how we can scale up production.
Um whether that be deepening vendor relationships.
So like we've had some relationships we've been building since day one with uh vendors that are non-traditional to these space problems.
Um and really and now we can indicate higher levels of volumes um which helps gets us you know special engagements with those partners. So that's great.
Um but also we're we're at this point able to implement bringing more things in house.
So, um, we're also doing that strategically with a whole wide range of, uh, pieces of our stack for exactly the reasons that you mentioned.
Um, and then also sometimes whether it be for like a performance or cost reason, but but also, um, you know, any, uh, any areas where we're kind of pinned down by a vendor's timelines, um, that's where we seriously consider insourcing. >> Yeah.
>> How many countries are you guys in now?
>> So, we're in building nine countries. Uh, yeah. >> Wow. >> Congrats. Wow.
Is that >> uh like what's the ceiling on countries?
Like I'm assuming that it's not like I'm assuming it's not like we have to 10x >> every country. >> Yeah. Yeah. >> Yeah. >> Yeah. Yeah.
Um so there is uh you know if you if you look at SpaceX for example like their market access program is tremendous.
I think they have nearly covered most countries which is insane and and super super cool.
Um but that's more on the consumer side.
So on the actual um you know back hall side which is the side that we focus on um it turns into basically like looking at a map and then plugging in a satellite elevation and determining um you know where you need to have visibility and where you need to have like uh extra density in a network.
So um I'd say there's there's still a lot more to go for Northwood.
I know uh we will be dramatically increasing that number in terms of both density and in terms of diversity.
>> So >> we should expect to 10x year over year.
>> I mean I was we're at a startup here.
We're we're trying to 10x everything year over year. >> There we go.
90 countries by by uh by next >> 9,000 countries. >> 9,000 countries.
And I hope you enjoy I hope you enjoy uh a lot of travel.
>> John, what were you going to say?
I was going to ask about hiring.
Um, just uh what is the balance between like >> PhD industry veteran, someone who's been working on the foundational technology for their entire career versus young person who might be able to get up to speed faster in the AI era, but at the same time they might not know to identify a hallucination.
How have you been balancing the role of young people in the organization in the AI era? Yeah. Yeah.
Um I would say if you have any interest in the work at Northwood um and if you have uh a lot of you know fire and what I like to describe for horsepower um to just get things done uh quickly um then we definitely are are really excited about new grad talent.
I think there's a lot that new grads can do and can contribute.
Um, you know, I I think there's a a spaceexism about what the exact ratio is of, you know, more senior talent to uh more new grad talent and really like what the optimal balance is there.
But um I think uh I think there's there's so much ownership that new grads are bringing to Northwood um when they come in the door.
So, we've been really impressed by what folks can get done. >> Awesome.
Well, congratulations on the new round and uh we'll talk to you soon, Bridget. >> Go for one more.
Go for one more before the end of the year.
There's still there's still >> still a couple months you got >> still like 50 days till Christmas.
You know, >> there's not a lot we got in the chamber, guys. So, we'll see you soon. >> Awesome. Great to see you. Congrats to the team.
>> Uh, let's tell you all about Figma agents. Meet the canvas.
Your AI agents can now create and modify your Figma files with design system context.
We got Tyler in >> Tyler taking over.
I'm going to drop off for this one.
Oh, Smi analysis crew hops on. Have a good one. >> Cheers, John. >> See you, John.
>> Uh, it's good to see you, Tyler. >> Likewise.
>> You are >> good to be here.
>> Physically around half the man John is, but spiritually 100%.
So, it's great to have you.
>> Um, we have multiple people from semi analysis joining. >> Yeah.
>> Any minute now, we can uh can head over to the timeline and see what's going on.
Um, Tay Kim was going off on on uh on Apple earnings yesterday says live by the sword, die by the sword.
It looks like services revenue from Apple's search royalty deal with Google will decelerate significantly going forward.
That's bad news for Apple. I don't know.
Like obviously not great because they just completely, you know, that deal was just like flow directly to the bottom line, but they had to, you know, all good things come to an end. So, >> it's true. It's true.
>> Um, >> I think we got the we got the semi's crew here. >> Let's do it. >> Coming in. Here they are. We got five of us. Incredible. Incredible. >> Going crazy. >> How's it going, guys? >> What's up? How are you guys doing? >> Welcome to the show. >> Um, thank you. Thank you.
>> Uh, gentleman on the left, introduce yourself because I think it's the first time.
And uh and then I don't know who this other guy is, but I think he I think he's a big yapper, so I'm looking forward to hearing from him, too. >> Yeah. Yeah, for sure.
Well, first just thanks so much for having us, guys. My name is Max.
I co-lead the tokconomics team here at Civ an Analysis.
Uh tokconomics is just the top of Civ analysis's full stack coverage and we focus on things like the AI labs, the clouds, the data companies, application AI companies, etc. >> Amazing. >> How's it going guys? I'm Andrew.
I uh recently joined the tokconomics team working with Maxer. >> Amazing. Uh who's this other guy?
>> Hey guys, my name is Dylan.
I'm a big loser, so thanks for having me on. Um, awesome.
Uh, you guys, uh, yeah, let's let's get right into it.
Why don't you introduce, uh, GDPX, uh, and then we can go into some of the takeaways from it. >> Yeah, for sure.
Uh, so GDPX is, um, in our opinion the single best benchmark for measuring frontier model capabilities.
Um we think previous attempts at comprehensive benchmarks most notably OpenAI's GDP bow are sort of fundamentally misguided uh because they pick their domains based on percentage of GDP or using things like the sock or on ONET taxonomy.
Uh whereas we believe that sort of the first trillion dollars of AI revenue are going to be extremely concentrated in certain domains.
uh the ones that can sort of uh move fast enough to actually adopt this new technology and also sort of have high enough of a skill ceiling to continue using frontier intelligence at high ROI.
And so we made a benchmark focus exclusively on those domains which are things like chip design, uh biology, cyber security, RSI, uh etc.
Um okay so my first question is uh how how did you think about the gap right now between AI's ability to execute on tasks and do them in an effective way versus roles right because I think like as you look out in as I look out into the next year and especially right now with the sort of market jitters that we're having it feels like we need a capability jump in the same way that We had like similar to
what we had in let's say Q4 of last year where there was like this momentary correction because everyone was sort of had this idea that that that chat bots were going to create you know trillions and trillions of dollars of value and then there started to be some skepticism and you remember there was sort of like um very momentary like correction and then we got agents and everything ripped again. And it feels like right now like
And it feels like right now like my question looking at this you know at the benchmark is like um are you guys thinking about obviously um there's a lot of room to run for all the models but like how are you thinking about like sort of broad capability increases versus just like scaling up intelligence in these specific domains. >> Yeah.
So I actually don't want to go into this too in depth because we're currently working on a demand model uh to sort of project 27 and 28 uh open anthropic revenue which is kind of is going to lay out in full detail the simply analysis.
I >> I do think we can say some things though right I'm going to give the overview for sure or you can go ahead too.
Yeah, I was just gonna say like I think um it's very like people people tend to think that AI will diffuse very broadly um and and that everyone will use a ton of AI and that's going to drive AI spend.
Um but we have a very different view of this, right?
We don't actually think that that's the case.
We think that there's a lot of uh challenges to AI being adopted this way.
We think that the general populace will not be able to effectively use a ton of AI in their tasks.
And given the fact that AI at a given level of capabilities falls anywhere from 100x to a,000x in cost a year because of the insane progress that uh these these firms are having in research, data, compute, etc.
Um it it it means that you know the the places where you actually can absorb all of this usage of AI is where you know sort of people people can take extremely large amounts of intelligence and get ROI from them quickly. Right?
replacing white collar work is something that will happen, right?
But that's not something that's going to drive the first trillion dollars of AI spend.
Um, those are not industries that are going to diffuse AI capabilities quickly.
Um, it's going to take them quite some time to, you know, reorder how their opex is spent, right?
Because companies just have a lot of inertia with regards to, you know, people's jobs.
Um, you know, changing how their compensation works, how changing how many people work on what.
Um there are areas where we think that intelligence is very very important though.
Um these include areas like chip design, cyber security, financial services, um life sciences, material science, um you know software engineering and in that you know there's sort of two you know there's you can absorb unlimited intelligence essentially.
Um and there's two reasons for that.
One is that it's just a domain where humans aren't anywhere close to the frontier.
Um, and there's just so much to discover.
Think material science, think chip design, think anything in the scientific realm, right?
We figure out new stuff, we deploy it, we make everyone's lives better.
There's the other world which is adversarial, right?
Think specifically finance and cyber security.
It's human versus human and we're constantly just going to be leveling up either trying to arb each other, you know, like quants do and, you know, make markets or it's going to be, you know, constantly trying to hack and then defense, hack and defense.
Um, and so there's this like constant demand for intelligence because when you have more intelligence than the other parties in the market, you win. Um, >> yeah.
>> yeah. So, so specifically though on on let's say like material science like if I throw if if I spend a billion dollars on tokens and I come up with a bunch of uh a bunch of like basically like ideas there's potentially like a duration I I
might have like a duration problem where I have hundreds of like sort of promising ideas but then that creates those ideas mean that I actually have to put spend behind each idea to sort of run them down and see if they're viable. And there still could be this like
And there still could be this like massive multi-year lag.
And so I'm wondering like I'm my question is like what is the willingness to spend for some of these companies that obviously have, you know, massive balance sheets.
They could either be heavily funded startups or they could be, you know, large sort of legacy players in these domains.
But I'm I'm struggling to see how you can get a hundred billion dollars of spend in one, you know, in in a domain like this just because very quickly you're not in a situation where you're like we have a shortage of ideas.
It's like we have a shortage of capital that we can put and and point at these different experiments. Does that make sense? >> Yeah.
So one one thing that I think we can sort of point to is there's there's two domains, right?
When you think about science, right?
There's discovery and there's engineering.
And everyone always talks about the discovery side of things, but actually the bulk of the spend is on the engineering side of things.
And you have to get really good at the engineering before you can discover something and bring it to market.
Um, and so when you look at um, for example, periodic labs um, they are a company that's specifically focused in this realm.
Most of their work is on engineering, right?
Increasing the engineering velocity of these materials.
Hey, if I make something, you know, this sort of deposition technique 10% more efficient, is that going to be what is the benefit to humanity?
And actually, it's not about inventing a new material.
Um, they're of course going for that as well.
Um, but if I if I improve the, you know, the let's just say, you know, in semiconductors, the deposition of something and so then your chip has less resistance or you can shrink it a little bit further.
That's so much economic value that is immediately delivered.
value that is immediately delivered. or hey I'm a battery company and I'm optimizing my you know you know my anode or my cathode or whatever it is and I've made it have 5% more capacity that is so much economic value delivered to society and it's and it's a bunch of small engineering and the big problem here is
when we look at software engineering everyone's like used to pretty advanced software engineering techniques but if you go look at the broad space of any of these companies in batteries in semiconductors in many in in in in automobiles in aerospace all of their tooling sucks and their research velocity is slow, their engineering velocity is slow. And so, you know,
And so, you know, using AI to create simulations to help uh judge all of the data that's coming out um you know, an Intel wafer, you know, they've made this comment, Intel when they make a wafer, they have about 100 terabytes of data that gets generated and they don't do anything with the vast majority of it, right?
There's there's small loops that get closed, right?
Oh, from this tool to this tool. There's a small loop. Use the data. Modify it a little bit.
Well, what how do you possibly analyze all this data when most of your, you know, PhDs in material science and chemistry and all this stuff are, you know, using Excel and mat lab, right?
And mat lab is if they're sophisticated.
Um so how about you have AI help this become you know something much more advanced in terms of you know the data analysis in terms of extracting information out of this stuff in terms of building models building simulations and closing that sim to real gap and then actually allowing you to make progress at an at a faster pace and and it could mean hey immediately I could figure out things that improve my yield.
um you know if I take for example and it's it's quite simple right the chips sit on substrates substrates route the wires in and out and power in those the substrates for CPUs this is stuff that isn't even that complicated you would think it's been around for a while the yields there are on the range of 40 to 60%.
What if those yields become 95%.
the economic value that's generated is massive.
Or let's take an Intel fab.
Their fabs are lower than TSMC's.
The yield if their yields are now all of a sudden higher uh than TSMC or equivalent.
Or you know, both of them just raise yields massively.
The economic value that can be delivered is massive. Um take Nvidia, right?
It takes them two and a half years to go from design concept to shipping a chip in volume production.
What if it takes them a year, right?
that architecture idea they had two and a half years ago or that change that they're trying to make you know a year into the process that's you know holding everyone up and and pausing things all of a sudden they can change it and do it much more quickly.
Um if they wanted to bring optics into the chip what about all the simulations that they can do and they can bring optics to you know co- package optics to reality much faster and get the step function improvement in performance of chips and networking speeds.
All of these things are rate limited by the by intelligence, right?
There's obviously a lot of physical difficulties, but the experiments that these companies are running are not able to close the loop quickly.
They're they're they're you know, the loops are very small.
The domains that people operate in, right?
The amount of siloing is just so massive because the complexity here is far beyond that of any software engineering.
And so the models need to get really really good at these things before they can assist these people and and you know help close this loop in a full circle, right?
And then optimize the heck out of these things, make these experiments that you're running um you know have have that data actually influence design and go towards the right process improve the yields, help you shrink faster, help you do whatever it is and and then actually generate hundreds of billions if not trillions of dollars of economic value to the world.
uh talk about the incentive for people building in these domains using the models to not want to talk about what they're doing publicly, right?
Because I feel like part of the current sort of disillusionment uh with with um with the state of AI is that you have a bunch of new software being created.
Software engineers are obviously happy.
Look, I shipped this new feature or or you have like really cool >> I don't know, bro.
I think I think software engineers are depressed, by the way. >> Yeah.
Yeah, but at the same time you have like the excitement of like um Carpathy making like hacker tube, right?
And everyone's like, "Oh, that's really cool that like you can do a two, three, you know, sentence prompt and like create this like sort of novel product um that that maybe wouldn't have wouldn't have been worth putting the effort into historically."
But I feel like when you talk about these domains like material science, like I know from having built businesses in the past, like if I figure out like one simple trick, I assume that like it's only going to be like you can only arbitrage it for like a certain amount of time before it becomes sort of common knowledge.
And so there's an incentive or or even another example like if I find an amazing freelancer like I'm not going to try to introduce them to like 20 people because I don't want my competitors or other companies in the category to have access to them.
category to have access to them. I assume over time they'll probably that that person will meet a bunch of people but like right now when you talk about you know these domains where you can have like a a 20% efficiency gain or things like that there's so much in like I assume that you guys are calling all these people and like trying to
understand the kind of like gains that they're getting but they're not exactly like raising their hand and going on LinkedIn or going on X and saying like oh we're getting this like massive speed up or efficiency gain here and we want to tell everyone because they want to like get as much out of it as they can before it becomes sort of more common knowledge. >> There's there's definitely a big
>> There's there's definitely a big difference in hustler culture, right, in in sort of software and founder land and SF and sort of these things versus um you know these industries that have that are manufacturing and making the world go right and work and operate.
Um I agree that they're not going to stick their hand up and scream uh what they're doing.
Um they'll they'll do some vague stuff.
Hey, we improved our our manufacturing efficiency.
Hey, customer, here's what we did.
It'll get out there, right?
You make your product better, everyone will understand that you made your product better.
Um, your financials get better, everyone will understand that your financials got better.
These are just like obvious facts that are public.
Um, and so, you know, we're in the very early innings of these things to be clear.
Um, you know, whether it be material science or semiconductor engineering or whatever whatever very hard science thing it is.
Um but our our sort of like one of our main you know focuses from GDPX looping it back is to focus on these areas where we think AI has to improve drastically in order to deliver a trillion dollars of revenue.
Um and now we at semi analysis are very bullish on this happening um on AI delivering a trillion dollars of revenue uh you know just openthropic and the leading labs within the next couple years.
years. um you know the two you know Max and Andrew will call me bearish for saying in two years but uh generally um you know if if those areas don't deliver then yeah we're we're going to like we're going to saturate all the existing tasks pretty quickly and spend and and diffusion of like white collar work replacement will be too slow and these companies will actually not even maybe
be able to make their compute commitments right it's like that's that's like sort of and I think the AI labs I think us at semi analysis I think the people who are really in the know see the tremendous progress that's happening I mean Max or Andrew, why don't you talk about the material science eval and and some of the things that uh our PhD uh collaborator sort of found. >> Yeah, sure. You want to take this one, >> Yeah, sure.
You want to take this one, Andrew? >> Yeah.
Basically like our our collaborators like found that like the models are much much better than they previously expected.
Uh and so they had like a moment where they have to like sit back and be like wait like this is actually better than a lot of like the the stuff people are putting out out there in papers.
uh and like in fact in like one of the problems like Astra for example solved like a subtask of a task better than like a state-of-the-art uh state-of-the-art reference uh published reference and so that was like a a big moment for them.
Um so it's very clear that the models are very spiky in these categories and are increasingly better and reaching like above you know superhuman um baselines but you know still like even on our benchmark uh the best model Astra material size only reached like 28%.
Uh >> can you talk about the the value of like using uh human baselines in these?
I noticed there was no human baseline in these.
Um, is that because this uh bark is just like so general and like it wouldn't make sense to for someone to be so good at all these different things or why is there no um kind of human baseline compared to a person here?
>> Yes, someone actually tweeted about this and I replied to him.
It was uh it's funny because it's very it's very hard to put a human baseline on these, right?
Like if Astra just gets better than like state-ofthe-art like teams of researchers, like what what is a human baseline there?
Um, I mean maybe for software engineering you could say like this could take x days of software engineering.
Uh, but like some of the stuff is just like teams of hardware engineers would have had to do to get done.
Like in our hardware design task or our chip design task like that that's not one person.
That's like many many uh people working together.
And it's very difficult to even put a number on like the amount of hours or uh days that it would take for those people to accomp like finish even those tasks that we have.
And we're well past the point where we're trying to measure, you know, our models similar to humans.
It's like we're trying to measure when will models be superhuman because we're not going to pay for models that are similar to like regular humans.
People are only going to pay for models that are superhuman in certain domains and then utilize them to advance humanity. >> I'm so curious.
We were talking about this earlier on the show.
How how should people read into Elon's post?
uh I believe it was Saturday that uh that like Grockbot is just too expensive to run.
Sorry, tangent, but um I'm curious what the what the house take is on that.
>> Like it might be too expensive to run today, especially when you consider sort of like the CPU and you know sandboxing costs needed per user.
Uh but I think one thing to understand about these personal agents is that it's very likely you will not need sort of the smartest frontier model a year from now in order to like serve the majority of people.
Um and sort of the the cost to serve will actually undergo this you know potential 100x plus uh like follow the same trend is essentially a decrease cost of intelligence that Dylan was talking about.
So even if like personal agents are really expensive to serve today, um I think it's very likely that they are totally economical to serve you. >> I know. So what's the logic?
You have a product that people love that's like, you know, gaining traction and then you basically like have everyone hit their usage limits, you know, suddenly and then say, "Oh, it's too expensive."
Like Elon has Elon's been on the record saying that he thinks that Grockbot will one day and again I don't know how much to read into this but he's like he thinks that Grockbot will will consume the majority of their compute over a longer time horizon.
So like what what what you know assume the cost is going to drop 100x like why do you why do you in stop like subsidizing that when it's seemingly working?
>> Well Elon you have to remember what Elon's trying to do right now, right?
He's taking SpaceX and turning it into a Neocloud, right?
Um in in in some sense, right?
He's building ridiculous amounts of compute.
And you know, he's only got, you know, if you talk take his balance sheet, you know, he's got what, like $80 billion of cash on it.
$20 billion is going to go to um you know, rockets and stuff like that.
But then the other 60 billion he can spend on compute.
He's raising another 40 billion right now.
Um at the same time, he claims he wants to build 13 gawatt.
build 13 gawatt. um 13 gigawatts if you build all of that out is is you know you know 400 plus billion dollars right um you know 500 plus billion dollars so so where does he get the cash right he has to monetize his compute as effectively as possible um to able to pay for all the capex that he wants to do um and
given what's going on in the spot compute market right you know being able to sell GPUs for you know $15 an hour uh when the TCO of them is only like $4 an hour um you you have this massive delta Um, and so he's exploiting that and he needs to focus on how do I build the most compute period, right? And that's
And that's that's step one.
Step two is okay, how do I use that compute most effectively to dominate the universe, right?
Is sort of like I don't know if you're like take a evil look at it.
I'm sure it's just to it is just to benefit humanity.
But the only way he can build >> a little bit. >> Yeah.
But the only way to build 13 gigawatts or even 10 or even seven financially is if he makes as much money as possible from the compute he has.
And so if that means, hey, he ripped some compute out of the XAI's team to sell it to Anthropic and Google, he had to do that, right?
Because ultimately a year from now, his XAI team will have a lot more compute to deliver by doing that.
Um and and and so he needs to be able to generate all this cash flow and and and sort of bridge the financial gap to get him to being able to spend 34 $500 billion on capex um over the next year or year and a half. >> Yeah.
And he's got to convince people that the the going rate for these gigawatts is not the 12 to 15, you know, billion a gig that people are signing 5-year contracts for.
It's the, you know, 40 billion plus a gig that he was able to sell this like short-term contract to anthropic for earlier this year.
Um but if you look at people who are actually all in on the personal assistant like you know a meta news I I do not see them seeing you know stopping the subsidation anytime soon because they know it's going to become economical like a year from now and all that really matters at this current point in time is we we get the adoption you know we get the users. >> Makes sense. >> Makes sense.
>> Um any other questions Tyler?
>> I think yeah this was great. >> We covered it. Yeah.
Uh what else is top of mind for you guys before we jump off?
>> I I think one other thing that's really interesting is um you know the frontier models are actually not the best at certain tasks from openthropic because they're safety classifiers, right?
Google and and meta and the Chinese models are are better at like cyber tasks, at biotasks, at certain alignment tasks.
Um and that presents a really interesting opportunity that certain markets should just actually not use open anthropic unless they get into some specific sort of tent uh that open anthropic are willing to give uh which plebs like uh us at least are not able to.
>> Um on uh just reminded me something I wanted to ask earlier.
Uh it feels like there's a massive incentive right now for companies to come out and say, you know, we're bringing compute on prem and we're running open models and we're getting uh uh and and how much should people read into that?
I think the headline yesterday, I couldn't tell if it was real, is that gold uh Goldman is sort of banning uh the use of of Claude code and codecs internally and other coding agents.
To me, that seems kind of like an insane move given that they're, you know, underwriting underwriting the IPO.
And it also feels like you're you're sort of betting against Frontier progress and and capability increases, but like how should how should people think about those kind of headlines?
>> I think this is really funny because uh >> Go ahead.
Yeah, I was just going to say like it's really funny that they would do this because like even on our finance like benchmark, it was very clear that none of the models were even able to like successfully do a task of an analyst.
Uh like the closest one got like 50% of the credit or something.
So like even like if you use open models, you're just like shooting yourself in the foot because you're just like using worse models and like even the frontier models aren't there.
Like imagine how much worse like an open source model would be. >> Yeah.
But I think I think onremopensource it is a real market.
It is going to keep growing like pretty fast um in absolute terms.
terms. Uh but if you sort of believe that the world has these fundamentally intelligence ball domains which like Dylan was explaining at the beginning and that people will be able to figure out how to use frontier intelligence in order to sort of earn high ROI in those
domains relatively quickly then you would expect sort of uh the market for frontier intelligence to grow faster on a relative basis than sort of the on-prem/opensource market even if like both of them still do pretty well kind of just in isolation. >> Yeah. >> Yeah. That makes sense. Uh, cool.
>> Yeah, I mean, we we do believe the bulk of tokens will be uh non-frontier models.
We just think the bulk of dollars will be frontier models.
Um, and so like, you know, great if Goldman bans their random, you know, year one, year two analysts from using cloud code.
Um, you know, that's probably a short-sighted mood move, but you know who they're not going to uh ban from using cloud code? All of their pentesting.
Um, and they're going to have to spend a lot of freaking money on pentesting because if Goldman's not secure, that's really bad. Um, >> yeah. >> Yeah. Well said. Uh, cool, guys.
Very, um, thanks for thanks for coming on and breaking it down.
Um, and, uh, I'm excited to see how quickly all of the different participants can uh, can hill climb this.
You've certainly given them a lot of work.
>> It's going to be exciting. See you guys. >> Awesome. Cheers, guys. Have a good one.
Uh, let's see if there's anything else we want to cover.
I think we got John coming back on. Oh, bring me back. Oh, there he is. I got the flashbang.
The the the weather changed outside the hotel room.
So, it's full much brighter than before.
>> But you can do a three up if you want.
We got to go through a couple uh a couple more news items.
We got to talk Crowd Strike.
First, let me tell you about MongoDB.
What's the only thing faster than the AI market?
Your business on MongoDB.
Don't just build AI, own the data platform that powers it. Uh, no, I like this.
So, Crowdstrike's acquiring two AI security startups for a combined uh $495 million.
Fiddler AI and Dread Node. Such different names.
One is like the most whimsical. Oh, we just Fiddler.
person who plays the violin particularly in folk bluegrass country or traditional rural dance music.
The fiddler >> I love it. I love it.
Yeah, it's it's it's funny.
They're uh they're taking security so seriously that they're just doing bulk acquisitions announcing them simultaneously.
But George Curts had a great post as well saying that uh he's creating the the the emergency button.
Uh what I forget I forget exactly how it >> the kill switch which is something like every I think like Bernie Sanders has literally asked for that and like many other people have been like we need an AI kill switch and and the labs have
said that there is like effectively a nuclear football that there's a device that can you know take everything offline but uh I think every company wants an AI kill switch so good good market positioning good branding you know people are rebranding uh comput
data centers to compute campuses is uh our buddy over at now the Atlantic Mike Isaac wants to use the term doesn't want to use data centers he doesn't want to use compute campuses he wants to call them data bases the base for all the
data I like that I'm still supercomput camp but >> yeah we we we've had the perfect name for decades the supercomputer >> I know the supercomputer it just sounds so good and it even fits with like the new the new super intelligence push. Uh
Uh the other news there, there's a bunch of OpenAI news here.
They hired Dan Driscoll, former TVPN guest.
>> We got him, US Army Secretary.
This goes deeper than I thought.
>> I I I really enjoyed talking to Dan Driscoll when he came on the show.
Uh it was a very funny moment for us because it was like we were just having fun on our show.
There was like now we're talking to the Secretary of the Navy or the Army of the Army.
And uh yeah, the reactions that day there were some quot being like TBNN's talking to the head of the army cuz it it had turned from like us just purely joking around to like us joking around with like our friends and obviously we had some very successful founders and venture capitalists on, but it was still like very uh very very light-hearted and pretty low stakes interviews, just people yapping about their their takes on the market business.
And then this one was like so so good to see him hopping in.
>> Uh SpaceX, >> yeah, >> uh has introduced Starfall, which enables affordable routine access to a microgravity environment, providing unique opportunities for manufacturing scientific research and testing in low Earth orbit.
Um SpaceX says Starfalls for biomedical research, developing advanced biomeaterials, medical implants, stem cell organoids, and replacement tissues.
Tyler, you would want to you were looking at um developing some stem cell organoids.
Um um so I'm sure that's exciting for you.
Pharmaceutical discovery, growing more uniform protein and drug crystals and microgravity to accelerate drug development and materials science.
Um manufacturing nextg materials and semiconductors using the vacuum and microgravity environment of space.
Um I always thought with Varta like the the the the the sort of like debate around Varta since the very beginning was that like okay incredible incredible team like very very unique >> sort of like market that they're going after always questions.
It was like so difficult for people to even believe that like this was super necessary and and next like what would the end market even be?
even be? It just it seemed so like experimental and sci-fi >> that it was almost like it it was hard to it was much harder to get behind than like the typical like manufacturing startup which is just like we make widgets and we're going to sell this many this year and we're going to
hopefully sell more next year and many many more in you know a decade right and I always in the back of my mind was like well I don't know what this market turns into but at least like this feels like very complimentary to SpaceX and it doesn't seem like something SpaceX is going to do. Um, and turns out uh turns
Um, and turns out uh turns out they are going to go compete in this category.
Um, but if anything, it just feels like incredibly validating to >> Varta because SpaceX just wouldn't do this unless it was like they saw it as like a a massive, you know, opportunity, right?
That was going to be a very important part of the overall space economy.
So although Varta now has some real competition again this just feels like a huge like vote of confidence in the business and the in the overall market opportunity.
market opportunity. So um uh pretty pretty cool >> questions that come out of it like one is uh is this something that can be somewhat complimentary to Varta in the sense that I I it's unclear if SpaceX is building just the launch trajectory the launch capacity to send something up and
bring it back like that is a different workflow that Vartis had to build and then Vartis also had to build a capsule for doing the biomedical research and has a bunch of deep partnerships with bio companies so there is a world where they wind up working together in some capacity. And then also we've just seen
And then also we've just seen time and time again uh companies not wanting to dual source from or single source intelligence Frontier Labs or uh launch capacity.
And so Rocket Rocket Labs going and AS is still you know trading really high um because uh there's just this expectation that not not everyone is going to be signing up for Starlink because some of them are going to say I want an alternative.
I want to be able to dual source.
I want to have some negotiating leverage.
So, um, overall, uh, exciting to see it, uh, the whole market like really heat up and and yeah, total validation that the thesis, uh, was something that that was, uh, interesting and viable and that there would be real demand behind because, uh, SpaceX wouldn't be going after it unless there was serious demand.
Uh, let me tell everyone about Console.
Console builds AI agents that automates 70% of IT, HR, and finance support, giving employees in resolution for access requests and password resets.
Uh, >> something going on in um in politics today, John. >> What's that? >> I don't know.
I just I keep seeing random posts about I don't know.
A lot of lot of blue and red.
>> No, I I What's politics? >> Oh, yeah. Yeah.
>> I mean, that that might we might need to save that for another episode. The TBPN deep dive. >> Yeah.
>> Yeah. history not not on any policy or politicians but just the history of politics the concept >> where does it come from it start with >> Rome Greece democracy uh yeah get out there vote do whatever you got to do on on uh election day and uh there's
there's some good Nate Silver uh podcast going around he was on Odd Lots breaking down P polling and uh and uh prediction markets are fitting together today and some of his thinking on the the benefits of each of them and where some of them break down, but it's all interesting. Uh
Uh we'll we'll have a bunch of updates tomorrow, I'm sure, as as the polls roll in.
Uh although some of the some of the elections could be like takes time to count the votes, so maybe go ongoing all week.
But there are plenty of shows where you can go get updates on what's happening with the midterms. So go check it out.
Uh yesterday we were at the New York Stock Exchange.
Let me tell you about the New York Stock Exchange.
Want to change the world?
raise capital at the New York Stock Exchange. Just do it. Uh don't Yeah.
Get your mind off the midterms.
Go IPO at the New York Stock Exchange.
That's If you're frustrated with the political climate, just take your company public. It's easy. Great.
Uh I want your take on this on this font.
Are you Are you a fan of this font?
They're calling it the new comic sands.
This speaks to me as a Doritos American. This is me. But what do you think?
Is this is this uh going in your repertoire for brand building?
>> Yeah, I mean I think every every company in the world should sort of adopt this at the same time and just level the playing field and just compete entirely on on uh product specs >> in every in every domain from like >> Yeah.
>> from from haircuts to c you know to coffee, you know, it should just be a stats driven market. No more branding. >> Yeah.
Yeah, >> everyone adopt this.
>> Everything needs to look like you're announcing an illegal street race. That's the goal. Um, no.
Uh, the chatbt app allows you to change the color of the bubble. Default is blue.
Blue bubbles going back blue and gray for what you type, what the agent checks back to you. You can change it. I changed it to green. TBPN green. I was very satisfied.
Then they updated the app.
Different shade of green.
It's completely rugged me on my green.
I have a nice I have a nice forest green going. It got a lot lighter. Don't like it as much.
But I want them to go further.
I want to be able to set this font as my default font in chat.
Every response coming back like this.
That would that would fix me.
That would get to three billion dau.
I I I think this is the unlock. No, >> I I agree. I agree. >> Fantastic.
Last last post, last ad read.
Let me tell you about Shopify.
You heard about him earlier from Harley that Shopify is the commerce platform that grows with your business and lets you sell in seconds online on in store on mobile on social on market places and now with AI agents.
So Jordy, is there anything else in the timeline that you want to go through?
Any other uh details that we didn't get to? Uh there's some debates.
Uh oh yeah, OpenAI also partner with Cathedral Systems.
everyone's ramping up cyber security stuff, but this was cool because uh Cathedral was the one that was able to break into the anthropic slack and there was a lot of debate over that.
But um in general, it seems like the folks at Cathedral are are good at penetration testing and have and everyone wants the agents to be ultra secure since they be they are quickly becoming like the master key.
Everyone's giving Muse all their passwords, baud all their passwords, chatp all their passwords.
And if you break into that, you're kind of in the email.
you're in all the services, you're in the bank accounts pretty quickly and you can imagine wreaking havoc pretty quickly.
So, uh cyber security is of the utmost importance Frontier Labs, not just because people want to steal the weights, but also uh the actual uh user experience for sure.
>> Uh we'll close out with uh the largest domain sale ever >> from our friends at Shopify.
They spent $99 million on shop.
com and I think they stole I think they stole it.
>> Yeah, >> I think they stole it.
>> That's a great deal for that money.
>> Amazing uh amazing domain >> and um I'm sure they have big plans for it. >> Yeah. >> Yeah. Very exciting.
>> Um anyways folks, >> that is our show. >> Yeah.
>> Um John, let's uh We made this work. We made this work.
I thought you were here right next to me the whole time. You tricked me. >> No one noticed. >> No one know. No one.
>> The empty chair is so bad. >> Oh, it's so brutal. So brutal.
I I I I you know, there's a reason that we could do remote shows every day and we don't because it's not the same. But uh we still had fun. We still had fun.
>> Production team learned a lot.
I think that there's a whole bunch of improvements that they have in the works.
Uh I think we're going to get this dialed for the occasional uh for the occasional moment where I got to go to the opera. You know, it happens.
I got to go to the San Francisco opera.
Marriage of Figura is playing and uh it say no to all sorts of tech conferences.
Oh, this is happening that. No, no, no.
But you can't say no to an SF opera invite.
So, that's where I'll be.
Time to change the tuxedo.
I I I'm I'm kind of dressed casual right now, but I'm about to class it up and >> Thank you.
>> get to the opera in my tuxedo, my >> bow tie.
>> Could you Could you Could you have called in with a bow tie?
Would it have killed you to do that with the >> Maybe Maybe that's the secret.
Maybe if I put on the tuxedo regularly, you'll put on the suit because right now you're like there's a comfortable level of distance.
Don's in sort of a business casual.
We're getting out of here.
>> Leave us five stars at Apple Podcast. com. Goodbye. We'll see him.