Model Mayhem, NVIDIA x Hugging Face Deal, GPT-6 Astra | Pablo Torre, Mohit Aron, Akshay Narisetti, Hari Ravichandran, Matt Caldwell & Jordy Leiser, Jeff Thornburg, Charlie O’Neill, Sridhar Ramaswamy

You're watching TVPN.

Today is Thursday, September 3rd, 2026. We are live from the TVPN UltraDome, the temple of technology, the fortress of finance, the capital of capital. Let me tell you about ramp.com. Time is money, save both.

Easy-to-use corporate cards, bill pay, accounting, and a whole lot more all in one place.

It's model mayhem, folks. It's model mayhem. We got tons and tons of new AI model releases. It's a great week to be into AI. I think all the lab leaders, they got together, they said, you know what, people just love AI, let's give them more.

Even more.

Let's all team up to launch new AI for them the same week so that everyone has something just to be happy about.

That's right. That's right.

We got Anthropic Fable 5.1, we got MuseSpark 1.3, we got Gemini 3.8 Flash, OpenAI's teasing Astra. Is it GPT-6? There's a 6 where the S goes in one of the videos. People will figure it out. But the model mayhem is continuing. Everyone got back from their long summers, their vacations, and they said, we gotta launch something new.

Time to ship.

We gotta update this stuff.

And Grok, Claude, and OpenAI were all down this morning.

So much mayhem.

People thought Astra might have escaped.

Kind of a deflock moment for AI maybe? What do you think's going on?

Possibly.

Do we actually understand why all of the different models went down at the same time? Because it's easy if it's like AWS went down and it took down a bunch of stuff.

Yeah, it was East.

Yeah, but why is Gemini down?

I think it was US East 1.

Why is Gemini down then?

No, no, Gemini wasn't down.

Oh, Gemini was never down.

It was the whole thing.

Oh, okay, okay, okay. That makes sense. So Amazon clearly very critical to the global internet. Good luck to the folks over at Amazon that are, Couple big announcements.

One, Chamath's birthday.

It is.

Big 5-0.

Should we sing full happy birthday?

I think full happy birthday. Happy birthday to you. Happy birthday to you. Happy birthday, dear Chamath. Happy birthday to you.

Fantastic. And the All-In Conference is coming to Los Angeles, I believe, a couple weeks. Very exciting. Sign up, get tickets.

So wait for our invites. So wait for our invites.

I think we— I think you've been invited to pay and go if you want to go.

Good to know. Equally important.

Yes.

TBN's Road to Christmas.

How many days we got?

112 days out. 112 days out.

112 days.

I gotta say, it's feeling like it's going a little slow.

Yeah, it—

I—

I wish there would— I wish I was feeling more pace.

Okay, think about it this way. We're only 13 days away from double digits. That's a big moment. That's a moment everyone's gonna be talking about on the road to Christmas. When we get to 99 days till Christmas, that's when you can start a countdown.

That's big.

If you get a really big advent calendar, you can basically start that. Typically, advent calendar starts December 1st. Why not 99 days away?

A Q4 advent calendar would be pretty elite by a TVP. Little treats along the way, help you hit those KPIs.

That could work. That could work. Let me give you the roundup on the model mayhem that's going on. First, let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. So Anthropic launched Claude Fable 5.1 alongside the restricted Claude Mythos 5.1. Google released Gemini 3.8 Flash and a cybersecurity-focused version. That's good news. Meta released MuseSpark 1.3. So It might not be that much of a surprise that Anthropic seems to have the strongest model of the 3 with Fable 5.1 scoring 66 on the artificial intelligence index. That is the bar chart that everyone has been posting in this cycle. It feels like we're sort of maybe getting to the end of the benchmark era. It feels like when these models are released, it's much better to solve a novel math problem or do something else. The pelican on the bicycle is still one of my favorites.

Yeah.

I love that one. But—

Yeah. It changes everything.

It does. But the benchmarks, they've been accusations of bench hacking, hard to interpret.

I just think people have very, very low trust in benchmarks.

They do.

At this point, everyone has had enough experience using various models.

Yep.

They have their own sort of internal benchmarks.

Yeah.

And so the demos of like, I built this game, I did this thing with it. And then also just the trusted voices of people who use a bunch of these models and they kind of give you the breakdown of what they like, what they don't. That has been where people lean a lot more. But the artificial intelligence, the Artificial Analysis Intelligence Index, this bar chart that you see, has been a good way to kind of compress down a bunch of benchmarks into one meta benchmark. So Fable 5.1 got the highest result ever on the index. The score's also ahead of Opus 5, which got 63, and Fable, which got a 62. Anthropic says Fable 5.1 is also cheaper and more efficient, made possible by an improved caching system. should make ordinary workloads 25% cheaper and long-horizon agentic jobs 45% cheaper, the company says. That's good news. Interestingly, as Ben Thompson pointed out, Anthropic is also sort of dropping its no-zero data retention policy, which there was a whole news cycle around a few weeks ago. People were saying, you know, why is Fable not taking off in adoption? It's a really great model. And there were a bunch of different explanations. One of them was companies demand zero data retention. They don't want closed-source AI labs to be hoovering up their private information.

Yeah, and I think they said that there's testing functionality that will allow for—

Data retention, but it's on servers and infrastructure that the company owns.

Yeah.

So you do keep some of the data, you still are monitored for hostile usage, but it's not going straight into Anthropic's databases. So Alex Karp and Satya Nadella both warned against this idea that models that companies should have data sovereignty. So the policy is going to be replaced. The No Zero Data Retention, No ZDR, is going to be replaced with something called EFS, Enterprise Frontier Safeguards. And that may have contributed to lower Fable adoption among enterprises. And it sounds like it was a direct response to user feedback. So good news that the people spoke and the companies listened. So over in Google world, Gemini 3.5 8 Flash is the company's 3rd Flash release in 6 weeks. They are flashing out these Flash releases. It scored 73.7% on DeepSWE, just behind Opus 5 and competitive with models that cost several times more. Independent testing gave it a 59 intelligence score, which isn't the absolute frontier, but it's a great result for a model generating roughly 300 tokens per second. So very quick. very cheap and very good at coding, at least on this particular benchmark, DeepSWE. We'll see what adoption looks like and where enterprise spend goes. Ara Karazian has some very interesting data from the RAMP Economics Lab you can go check out. He also has a new post that's very interesting that we can talk about in a second. But last model, Meta Muse Spark 1.3, did very well in benchmarks, scoring 75.4%. higher than Gemini 3.8 Flash on DeepSuite, beating both Opus 5 and GPT-5.6 Sol. It didn't sweep the board. Opus still beats it on several professional work and computer use evaluations, but it got 62 on the intelligence index, which is only behind the newest Claude models. Tons of stuff to think about and discuss here. So the interesting post from Ara Karazi, and I don't know if we have it in the timeline, if we can pull it up, but he was saying that there's a lot of concentration in the enterprise AI revenues right now. OpenAI and Anthropic, 80% of their enterprise revenue comes from just 1% of the companies. And I was like, 1%, that seems crazy. And he notes that this is uncommon for software categories. Like if you look at CRM, if you look at databases, if you look at all sorts of different software spend, typically you don't see as much concentration. You don't see 1% driving 80% of the spend. And I was wondering about this. And so, I started looking up like, where else do we see this type of inequality, if you can call it that, this distribution, this power law? Power laws are everywhere, but where else does this exist? And you might go to hiring. Like, is AI a drop-in replacement for hiring? Is it going to be proportional to hiring? And in fact, 1%, the top 1% of biggest companies in America, they do hire a ton of people. The top 1% of American businesses employ 65% of the total workspace.

Wow.

Workforce, not 80%. But interestingly—

The concentration risk there, John. 65% of jobs are tied to just 1% of companies.

There is, yeah, there is. And I mean, yeah, you definitely see that. Although the top 1% companies tend to be pretty— Lindy, you're talking about—

No, no, I know.

Yeah, the government and whatnot. But the interesting 1%, 80% correlation comes from sales. So the top 1% of American companies by sales generate 80% of total revenue, total revenue. And so there's this weird dynamic where I don't know exactly how correlated it is, how causal it is, but there is an interesting dynamic there where it feels like if you look at the total AI spend, it's around $150 billion a year or something like that. And then you look at total revenue for all US businesses, AI is roughly a quarter of a percent of total US business revenue. And it tracks fairly closely to the revenues of those individual firms. So, you see that 1% of the top businesses generate 80% of the revenue. They also spend 80%. They also generate 80% of the AI revenue. And so, there's this interesting dynamic where Because enterprise AI particularly, you're not gonna be on the $20 plan, you're not gonna be on a $200 plan, you're gonna be consumption-based, and you're going to look at it a lot more like a marketing line item that's proportional to your revenue potentially. That's at least one interpretation of this. Another fellow over at Ramp said that this is a Rorschach test for how you feel about AI. Either you look at this and you're like, it's great, or you look at this like, it's over. But Fun, fun, fun, fun chart to dig into. Anything else on this, you guys? Read any of this?

Nah?

Let me tell you about Codex.

I just was realizing—

Codex is a powerful workspace for getting work done with AI agents. Whether you're writing code, analyzing data, creating content, or automating business workflows, Codex helps you move projects forward from start to finish.

Our guest today, we forgot to cover, Pablo Torre, joining at 11:30. 11:30. To talk about the Clippers and Kawhi Leonard.

So these are people that go, they watch live streams and they clip them and they put them out on social media.

Yeah.

Why?

I think that's what the team— why they named the team that.

Yeah.

Kind of an homage because there's a lot of clipping that happens in LA. TikTok clips, Instagram clips.

So they call them—

yeah. This whole Ballmer-Kawhi Leonard thing. You had an interesting pronunciation of Kawai's name earlier, because I don't think you'd ever heard of him.

Well, I was calling him Steve Balme. I dropped the R, because I thought it was French.

Oh, nice.

No?

Yeah. And then we got a bunch of others, somewhat of a lightning round. We have Mohit from Siphon. He's got a big funding round from Altimeter. We have— Coming out of stealth. Akshay from Pocket, sold over 200,000 devices. We were talking to Jimmy yesterday. He was saying, why can't— Big companies do hardware. Well, Pocket's doing it. They're making it work. Jimmy did say that he sold— he was like, oh, Meta sold 2 million pairs of glasses. I sold 2 million pairs of headphones in Brooklyn, which was a great line. And a bunch of other great teams joining. And we'll cap it off with the CEO of Snowflake coming off a great quarter.

Yeah, stock is way, way up. Very exciting. Let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Just build AI, own the data platform that powers it. So the other big story in the news, the open source community is stronger than ever. You saw 3, 4, who knows, 5 closed source releases. All the, all the big labs are duking it out. Meanwhile, Nvidia is going even bigger on open source with the $13 billion acquisition for Hugging Face all over the timeline. This was leaked. A couple weeks ago, I feel like rumored. What are you laughing at?

No, just every single day I would see a headline about Nvidia hugging Face.

Yeah.

Think, okay, now it's official.

Yeah.

No, now it's official.

Today is the day they can talk about it. They can explain it. And there's a lot that makes sense. There's not too many questions. It's just a great outcome generally. But it's an interesting story because it's a true 10-year overnight success and, and they're having fun with the acquisition price. Nvidia agreed to pay $12.9303 billion for Hugging Face, which just happens to be the exact decimal code for the Hugging Face emoji, which of course is the icon used by the company. And then also, if you take that number and you turn it into a color code, I think you get a green that sort of hints at Nvidia. So they're having fun both ways, like, yeah, this was always in the plan symbolism. And it's just funny to be having fun with a price this big. I remember the Instagram acquisition and the idea of a billion-dollar outcome as being insane during the social media boom. And now we're seeing like decacorn liquidity events every couple of weeks. I'm, of course, thinking of OpenRouter and—

It's honestly an incredible time to be investing in AI. 7 years ago.

Yes, yes. Best day to plant a tree.

3 to 7 years ago was an amazing time to be investing in AI.

It was. And so obviously this shouldn't come as that much of a surprise because Jensen has been probably the loudest voice on open source. He put out that open letter that everyone signed on to and he wants to maintain NVIDIA's dominant position in the AI ecosystem. both selling chips to closed-source labs who might wind up making their own chips as well. And there's a whole tug of war there. But for open-source AI development, he wants to be the place where developers and companies go to pick their models and then hopefully rack them on NVIDIA GPUs. The simple distillation is just Hugging Face is the GitHub of AI. It's a little more complicated than the Microsoft GitHub deal. But it still makes a lot of sense. So Hugging Face doesn't own the smartest models. They don't even try to build them. And there's some interesting financial dynamics there about how capital efficient they were because of that decision. But they created this nexus for people to upload, discover, test, modify models. And the numbers are good. They have over 18 million developers, 200,000 companies using the product, 3 million models, and over half a million datasets. And so, They have certainly created this vortex of activity that's really valuable. How much, how strong is the network effect? It's certainly cooking and it's certainly driving a lot of value here. So the interesting thing about Hugging Face is that it did not start as an AI GitHub for AI. It started as a completely different idea. The founder worked at a French computer vision startup called Moodstocks that was eventually acquired by Google. And in 2016, he teamed up with 2 co-founders, one who was a mathematician and the other one who was a scientist who had worked in patent law apparently. And they started building an AI that could basically talk about everything. So this was post-Siri, post-Alexa, but instead of focusing on like, tell me the weather and be an assistant, you know, set a timer, he wanted just to be able to talk to you.

Still not solved, by the way.

Wait, which one?

Siri.

Yeah, yeah, yeah. So yes, it's pretty good at setting timers. But so the goal was to build something like a Tamagotchi, something very cute, hence the Hugging Face icon. A funny, emotional digital friend targeted at teenagers. The app let their name— let users name the bot, text it, send selfies, trade emojis. It was explicitly marketed as an AI best friend friend for bored teenagers. And they scaled it. I think this is pretty significant. It was doing 1 million messages a day. They had more than 100 million messages in total by 2018. That seems pretty significant. That doesn't seem like, you know, you're languishing in the App Store with no downloads because how many messages a day can a bored teenager possibly put up with an AI agent? Even if it's like 1,000, you still have, I guess, 1,000 users, power users. I don't know, probably the average user's doing 20 messages a day. So you're seeing pretty significant adoption. And so they were able to raise a series of financing rounds. The big one that grabbed headlines was Kevin Durant was in the $1.2 million— I think it was pre-seed round. Betaworks and SV Angel were also in there. That was 2017. Then they did a proper $4 million round led by Ronnie Conway's A Capital in 2018. The company had raised money and the technology worked well enough to feel sort of magical, but it was still 2018. This is pre-GPT-3 and it didn't become a durable consumer business. So in 2018, Google released BERT, which was sort of the first language model, very primitive, but people were really excited about it. But it wasn't delivered just as weights that you could download on the internet. It was delivered as a paper. From Google. And the paper was implemented in Google's TensorFlow framework. And people liked PyTorch, so the Hugging Face team converted BERT from TensorFlow to PyTorch and released the conversion for free. And so developers really liked that. And that became sort of like the initial go-to-market flywheel for developer adoption. And eventually, they added more and more models, eventually thousands, now, I think they have millions of models, which is sort of crazy. But when you think about all the forks and fine-tunes, it makes sense. Eventually, the team stopped trying to build this one application and focused on building tools, became the picks and shovels trade. So instead of trying to pick a winner, you just host every model. They became the Switzerland of AI to some degree. And so the flywheel started compounding, more models, more developers, more model creators, more companies. And it was the same basic network effect as GitHub. GitHub was the default home for open-source software. Hugging Face very quickly became the default home for AI models. Over time, Hugging Face grew from a code library to a place where developers could publish models, version them, attach datasets, discuss changes, and they even allowed them to build demos. Hugging Face eventually launched a Spaces product where you could demo the different models. Companies could maintain private repositories, same GitHub strategy. So 2019, Lux comes in with $15 million.

Ooh.

Then, yeah, Lux got in early. Series A, $15 mil. They also came back for the Series C in 2022. That was $100 million at a $2 billion valuation. Sequoia and Coatue were in that round. There was also a Series B in 2021. That was $40 mil. And then the big step up was in August of 2023. Hugging Face raised $235 million at a $4.5 billion valuation. And it's a murderer's row of potential acquirers. So you got Salesforce, Google, Amazon, NVIDIA, AMD, Intel, Qualcomm, and IBM. So it's not like they were doing a roadshow to sell the company, but it's very much like, we want to be the Switzerland of AI. We want good partnerships with everything. We're going to be chip agnostic. So yes, we have NVIDIA on our cap table, but we also have AMD and Intel and Qualcomm. So you can count on Hugging Face as being like an independent place. We're not purely NVIDIA-backed, which is maybe one of the things that they'll have to deal with now, but they're not purely NVIDIA-backed at that time. So, it's very much like, oh yeah, we'll host a model that runs well on AMD, we'll host a model that runs well on NVIDIA, we'll host models that are from Google, from Amazon, etc. And so, it looked like this like peace treaty moment from the major AI infrastructure companies. You get everyone around the table, everyone everyone's aligned with the mission. And Hugging Face becomes this neutral territory where it supported competing clouds, chips, frameworks, models. No simple— no single company could control the platform. And so, even though they did a number of rounds, Hugging Face, I'm going to say, only raised under $400 million, which is a lot of money but not at a $12 billion outcome. And it's pretty small considering the outcome. And it was very capital efficient because they weren't actually buying chips or serving models directly. And they had this flywheel that sort of spurred growth through the network effect naturally. So not a lot of cost in the business. They became profitable in 2025, still had half the money that they raised. So Nvidia came in to offer $500 million late 2025 at a $7 billion valuation, but they turned it down.

Whoa.

Turned it down because they said, hey, if we're going to go deeper with one particular area, it's got to be the whole chip. Bang. And so, that's what wound up happening. So, pretty high revenue multiples.

My question is, I wonder what Jensen's vision for Hugging Face is. They do offer model routing.

Yeah.

And they rank a bunch of inference providers. Is this something that— could we see Hugging Face and OpenRouter and RAMP's router competing more and more? Yeah.

Sure.

I think it's 2 things. I think one is, If, like, closed source is already its own business line, sell chips to the labs, but the labs are building ASICs, they're doing a lot of stuff, and there's this whole back and forth tug of war there. But on the flip side, you have open source, which is continuing to grow. And if you can be sort of the front door to that and then say, hey, you found your best model, your framework on Hugging Face, and now you're ready to go and buy chips or buy inference or buy compute, and NVIDIA's right there. That is a very logical flow. And anything that they can do to make open source powerful, exciting, a place where you can build a career, have a great outcome, I think that's beneficial to NVIDIA because a lot of people will see this and say, yeah, like you can go and build a great company and have a fantastic outcome. I mean, the retention packages are apparently $1 billion for A pretty small team. I think it's in the hundreds still. And so if you're— if NVIDIA is just trying to send a massive signal to the world that you can make it in the open source world, that's a really good signal to send. And it feels like it's landing loud and clear, especially today. What else is on your mind regarding Hugging Face and NVIDIA? And while you think about it, I'm going to tell everyone about Console. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. What do we got on the timeline?

People are still waiting for an official launch from of Astra. Lisan Al-Ghaib is sharing some Astra benchmarks. Arc AGI 3, 98.6. He had, uh, they had previously said, are you ready for a nuke to hit Arc AGI 3? So almost fully saturated. Frontier Math Tier 4 V2 gets a 97.6. DeepSWE 74.1. X-Plate Bench also saturated at 100%. So, wow, seems pretty good, but still no official announcement.

Okay, well, we will keep monitoring it. Let's go through what else is in the timeline. What is John Palmer saying these days? He says, I actually think Snapchat for work might be a good idea in today's big companies. Work platforms, work communication platforms have always followed what teenagers were doing 10 years ago. I did use IRC when I was a teenager. Unks. Tyler has to look it up. He doesn't know what Internet Relay Chat is.

Wow.

I, for what it's worth, I didn't use IRC either.

No. Did you use AOL? You didn't use AIM, AOL Instant Messenger?

Wow.

Youngin'.

No.

What was your first communication platform on the internet?

Email.

Email? Gmail?

Yahoo?

I remember being inundated with emails. I remember there was like a summer— there was a summer—

He's laughing at me because I had a Hotmail.

Wow.

Rude.

Well, I just remember spending a summer as like Yeah. Not even a teenager yet, just being super stressed about my inbox.

Yeah.

Because like every kid had just started using email.

Yeah.

And so they were just sending these super long emails. And I would, and I would be thinking, I'd just be like playing outside in the grass and thinking, man, I gotta, I gotta check my email.

Yeah. Do you think there's anything actually to this? It sounds like he's being serious. In the age of AI slop, the most efficient form of communication is just short videos of yourself speaking. What do you think?

I would love to use Snapchat for work.

So if we, if we said, hey, as a team, We're going to communicate through short selfie videos.

Yeah, with the filters.

With the filters. You got to have the dog filter or whatever, the hot dog filter, whatever those were. He says he's serious. A 2-minute demo video. Yeah, I guess in terms of actually just taking a video of your screen, showing people what you're working on, there's a lot of different things you can do. There's a town that's For sale, six million bucks gets you an entire city in California, no less.

Not a cheap place to live. You know what I'm thinking?

What are you thinking?

Data center developers have been having some issues, right? Towns—they they try to move into a town, start building the data center. Town says absolutely not. Well, here you go. Buy the whole town. Buy the whole town. Who's going to tell you no if you are the king of the castle? The town next door, maybe.

They might be annoyed.

Boss. I don't know.

Anyway, king in the castle. King in the castle. There's a there's a. Video here on Good Morning America breaking down the details. An entire town went on sale for $2.6 billion.

TBPN City. I did get excited about the potential of just welcome to TBPN, California. But yeah, I'm not super eager to move.

This has Riley Walls' project written all over it. Yeah, he started with the street. He's got to do— who bought the street, by the way?

Notion.

Notion. The Notion Way. I think that was what it's called. So Notion, California, maybe it's right there. It's official, hot bot summer is over. Supergrok is saying goodbye to companions. I remember we were debating, you know, there's obviously a lot of people that had ethical concerns about AI romantic companions, but we were more discussing just like, is there actually a business here? once you break the seal of like, okay, we're doing it, he did it, would it actually be successful? Because Replika has seemed to get to scale. There's been other products that have played in this world and seemingly reached adoption and scale. But it's very— this is the same thing as the Sora discourse where everyone was caught up in Sora is is either going to be the most powerful thing ever and we're not gonna be able to stop watching it, or it's gonna be good and so fun and it's gonna be amazing and dominant. But no one was counting just like, oh, it might just go away in 6 months. And it's the same thing here.

I don't know.

Internet porn.

I think I was saying it was gonna go away.

Gonna go away, yeah.

Just because I thought it was gonna be a tool.

Yeah, and I think we benchmarked the market for this. Like, if you look at other romantic stuff, it's just not a trillion dollars of revenue.

Yeah, so when Elon first said, okay, I'm going all in on on adult entertainment. I did think it was a potential path for Grok to get into the single-digit billions. I probably overestimated the market there, but it felt like one of the plays that he had to get back in the game at the time.

Yeah.

But again, this was last year. And back then, if you could get into the single-digit billions, you were doing pretty well. And then the game obviously really shifted down.

Yeah, it makes sense to wind it down, focus on enterprise software. That was what was in the SpaceX S-1 was the— what was it? $13 trillion market he was going after, something like that. It was a shocking, shocking number, maybe $20 trillion or something. Absolutely huge, huge numbers, and it makes sense. There's a lot of value in the enterprise, much less in this controversial topic. Well, Phish.audio ran a billboard campaign. We love out-of-home. This one sort of confused people. Heshy Brody says, $100 for anyone that can explain what this company does without looking it up. This is on the New York City subway today. And we'll read it to you and you can take a guess. Fish.audio says, we put voice AI on a silent sign. You see the problem? Is this one of those jokes though?

Because there's a joke.

I think this is an ElevenLabs competitor. We know, but is this a real ad from a real company or is this a prankster making fun of tech ads? Because there was that Prankster who put up a bunch of billboards, like fake billboards, remember? Like, 2 2 12.

I think it just makes you— it really makes you think. It really makes you think. It made me go to the website. OK. So what do they do on their website? They make text-to-speech, speech-to-text, audio separation, voice changers, translation. And they partner with global innovators, John. They're working with HeyGen. OK. Retail.

That's a real company.

A bunch of games companies. It looks like Clout Kitchen. You ever heard of Clout Kitchen, John?

No.

You ever been in the kitchen cooking clout?

Okay, so isn't it like they put voice AI on a silent sign? You can't— like, if you have a billboard, you can't listen to it, but their product is audio.

Oh, so if you get—

You can't listen to the billboard. That's the problem.

Yeah.

Right?

I guess I would expect—

Your explanation is a problem.

I would expect the application to allow you to take a picture And it reads it to you. That's what I would expect. Anyway, we do have our next guest in the waiting room. Let's bring in Pablo Torre from Pablo Torre Finds Out. He's a host, investigative journalist.

What's going on?

And he's on an absolute tear. What are you doing?

How you doing? Pablo, it's an honor to have you. I was telling John before the show started, we podcast for a living. I'm not a big— I don't follow very many sports very much. But every time there's a sports story, I go find your content. And when I watch your content, you make me feel like an amateur, because you're just so good— Really good.

—at what you do. So good at yapping.

But then you're also, also an elite investigative journalist, and we wanted to take a victory lap with you. Yeah, it feels like—

yeah, my show is designed— thank you, A. B, my show is designed for people who don't care about sports at all to find it even mildly entertaining and interesting. And this is a story that has immersed me in the world of securities law and the funneling of money and billionaires and Especially one particular billionaire who used to run a certain company called Microsoft. And so all of this, I think, is actually in your guys' wheelhouse. So happy to be on with you.

Yeah, and I remember, I remember I watched some of your content around a year ago when you were originally breaking this story. And then I'm surprised that the story like at least went away out of the, out of the public eye for some time. I'm sure you were still following it, but take us maybe From back a little bit to first kind of uncovering this, and then we'll get all the way to the present.

Yeah, so this was an investigation that started with a tip, and the tip was there's a weird company out in LA called Aspiration. It was a tree planting startup, AKA a carbon credits company. The ESG thing, as you guys may recall during the pandemic, was a real thing. We're gonna be good guys, right? Everyone's gonna be a good guy now. And, uh, you know, endorse your industry.

So cynical. Insanely cynical.

Hey, hey, hey, I love trees. I love planting trees too. I love what they do to carbon and oxygen, reportedly. So Aspiration has all these endorsers that are public. There's Robert Downey Jr., there's Leonardo DiCaprio, there's Frank, there's Cindy Crawford, for people who grew up in the '90s.

Six God.

Whoa.

Toronto figures into the story in lots of ways, it turns out. But the point is that there are all these A-list stars, Orlando Bloom, I didn't even mention, all these A-list stars that this good guy company is paying to tell everybody about what they are.

Yeah.

And what I get a tip on is, hey, you should look into this company. It's kind of weird. They just signed a deal, $300 million, to be the jersey patch sponsor of Los Angeles Clippers. And they have another endorser that you'd be interested in.

Yeah.

And when I look into this, it's not immediately obvious what they're talking about. But when this company goes bankrupt because, and this is gonna be in your wheelhouse again, they were gonna go public via SPAC.

Yeah.

What a time, right? What a time. SPACs, ESG, we're all gonna get rich being good guys and no one really needs to ask many questions. Just trust us, it's gonna be fine. And what turns out is this company is co-founded by 2 people, 2 big Dem donors, classic good guys, went to Harvard. I mean, I resemble the remark. I wanna say that this has been also awkward for me. Steve Ballmer, Harvard, all these guys, all of this, right? Pedigree, all of it. The thing that's interesting is that when you dig into how this company fell apart and they go into bankruptcy, you of course get to examine some public filings. And one of their big creditors was a company called KL2 Aspire LLC. And if you're a basketball fan who's aware of the Clippers, you may realize that KL—

Yeah.

Sounds like Kawhi Leonard. 2 sounds like his jersey number. KL2 Aspire sounds like a vehicle you've invented to accept money from Aspiration.

Yeah.

And he was owed $7 million outstanding. And the weird part was that this dude had no public record of being associated with Aspiration personally at all. And so when you dig in and you begin to ask questions of people who used to work at a company that has now gone into bankruptcy, in which the co-founder Joe Sandberg is now, by the way, serving 14 years in federal prison for fraud. You get people who are interested in explaining how crazy their life has been.

Yeah.

And what they say and what they provide in the tonnage of all the reporting I did for months, 7 months before we came out with part 1 of the series 1 year ago today, was documentation that attested to the fact that Kawhi Leonard was paid, according to this agreement, a total of $48 million, $20 million in stock, $28 million in cash, to do nothing.

Yeah.

For a deal that never got announced. And the question was, why?

Yeah.

And so you follow the threads and you get to, wait, there's a massive salary cap circumvention scheme in which the richest owner in American sports, Steve Ballmer, is trying to use all of his wealth in ways he's not allowed to, to get money to a guy that he needed to take away from the Lakers and the Toronto Raptors in order to make his dream of owning a professional basketball team exactly the dream that he imagined.

So how strict are these rules? Because I can imagine— that's a funny image.

Yeah, so we covered the whole, like, the AI talent wars last summer every single day. It was crazy. And we kept coming back to this idea. Part of why the talent wars worked in many ways is there's no salary cap, right? So Zuck could just spend exactly as much money as— The players or the talent would accept. But yeah, in this case, rules are— there's actual laws.

Yeah, I guess I just mean like if I'm a, if I'm a player and I know that the owner runs a hedge fund and I see the benefit of putting some of the money that I earn directly with their hedge fund, it gets me access. Or, oh, I know that the owner of the team is friends with the CEO of Nike and I might be able to get a shoe deal. Like there's some level of just like doing business above board that's probably acceptable. But is this a blurry line? Like, how defined is this? Like, how clear is it to you?

So 2 things to know. One is that this is a cardinal rule of sports in which you raise a very astute point. If you are LeBron James, it's kind of insane. No disrespect intended to the latest person who got bidded up between, you know, Meta and OpenAI and Anthropic or whatever. No disrespect. LeBron James is probably thinking to himself, Why can't I make $100 million a year?

Yeah.

His salary is capped. And if you have problems with that, I get it. Sports is a fun mix of socialism and capitalism that is convenient typically for the owners of these teams who are wealthier every time you check the news.

Yeah.

So point A, well taken. Point B though is that it's a cardinal rule of sports, meaning that sports is one of the few places where your spending power is not automatically supposed to let you just buy whatever you want, right? This is how big markets and small markets ostensibly get to some level of parity. This is how, and this is debatable, of course, we're getting into more philosophy than I expected, but it's a fair point. It gets into the question of like what's actually moral and not. What Steve Ballmer did, given that blurriness around the ethics of it, was violate, and really, I would say, blow past any semblance of plausible deniability that, hold on, I thought this was something that we could do here. And I say that because according to the NBA's investigation and my own, this was not merely a scheme that he ran to funnel money with one company, Aspiration. He did it with 4. He did it with the scoreboard manufacturer, multimillion-dollar deal for Kawhi Leonard to do nothing. He did it with the insurance company, Locked In Insurance, which was the insurance company on the build-out of the Intuit Dome. He apparently did nothing for that. And he did it with— and this is a fun one— Boingo Wireless.

Oh, yeah.

Which you may know as the wireless provider of Los Angeles Clippers. And the thing you say to yourself, do I really need to pay for Boingo Wireless?

Yeah, I'm like, oh, they don't have Starlink.

Which is my personal review of the product.

But nonetheless— So when your original story broke, you got a bunch of pushback, even from a bunch of people that are pretty tapped in and knowledgeable. And some of that pushback was just that there's no way that Ballmer would be this dumb, right? So this story couldn't possibly be true because who would ever risk it all in this way? Is that somewhat accurate?

I think that's even the most generous defense. I think there are a lot of people who are just being pulled by the Clippers and Ballmer's crisis PR handlers. This weird podcaster is clout chasing. None of this makes sense. Steve would never do this. He's Steve Ballmer.

Yeah.

The less, I think, surprising version though, which you just articulated, is a common one, which is he can't be this dumb.

Yeah.

And I dare say that one of the things you learn when you investigate financial impropriety in our great country is that that is something that you should never assume. Sure. There are many phenomenally successful people who, because they are desperate to get the thing they can't just buy, do things that they think will never get found out.

Yeah.

And the problem in this case across these 4 companies and across these employees and across my reporting is that he trusted people that he should not have trusted. And also he maybe shouldn't have tried to do it in the first place.

Isn't there one weird trick that would have made this all work, which is just Ballmer goes to Boingo Wireless and says, yeah, we got this deal. It's expensive, but you will get some Instagram posts. And then Leonard actually follows through on some of the promotions. And yeah, would that have been enough that he was actually Yeah, like it could work.

Yeah, if he, if he had actually carried out the endorsements, would that have been enough? Or was that still crossing the line?

There isn't a— so I've debated on my show literally Mark Cuban about this whole story before it became obvious what this was, and I appreciate the apology he gave me on Twitter yesterday, genuinely. But there's a Shark Tank episode we could do just like pitching ideas for how to circumvent the salary cap that would have worked. And, and I think one of them, by the way, just to play that game briefly. It's like telling Kawhi Leonard about crypto.

Yeah.

Right, there are untraceable flows of money that I am told are reliable at this point in the calendar that you could use, right? Just one example. But what they did was actually almost too clever by half. What they really did was they created fake jobs for Kawhi Leonard and fake consulting agreements, consulting fees for the companies. So they were trying to create separation.

Okay.

A familiar term in the investigation of financial propriety, but they're trying to separate where the money was going from A to B to C. The problem was that everybody in the course of doing it, people they trusted, left a trail of paper as well as testimony, as well as a lack of rational explanation as to why this existed if you never were to announce the deals. And Kawhi, and this is the great comedy that is sports in which salaries are capped but egos are not. Kawhi Leonard. Kawhi Leonard said to everybody, I'm not doing a single thing for these off-the-books payments. I am not doing work for them, but I want them. And Steve Ballmer said, got you, we'll make that work. And because these deals could not be explained as endorsements because he literally never endorsed them, it became this comedy of errors.

That's very, very odd. All right. Let's get to the punishment. What's been your reaction to Kawhi got a $700,000 fine? Doesn't feel very significant, at least, you know, compared to the actual payments and what was received. And then you have a 1-year ban for Ballmer. And then— but what was your reaction to the verdict from the NBA?

Yeah, this was the biggest punishment for an owner in the history of, I think, American professional sports. And the NBA took 5 first-round picks, which is the most you could imagine. They took $30 million, which is a rounding error, of course, for Ballmer. They made him pay the $50 million legal fee to Wachtell Lipton, the NBA's outside counsel, which is also a rounding error, but still.

Yeah.

Pretty annoying, I would imagine. But the real thing that I thought was meaningful was that they banned Steve Ballmer from his own building for a year. He loves basketball. He sits courtside. He measured literally the toilets at the Intuit Dome when he was designing the whole building. He was a micromanager who knew everything about every piece of the building, and they banned him from being inside his own building, which he funded for $2 billion. Privately, to his credit. And they also banned his president of business for a year, and they banned his GM, his president of basketball, for 6 months. And so Kawhi got effectively a slap on the wrist, and that is wild on the merits. The reason they did it briefly is because when Ballmer now is rattling his saber, because he is the former CEO of Microsoft who, when presented with antitrust findings by the literal federal government, also fought, right? This is unsurprising if you know the trajectory of his life.

Yeah.

What he has threatened is litigation. And what he has not been able to pursue is arbitration, which is interesting. Arbitration, briefly, in the world of Sports Court, which is another TV show we should pilot together. In Sports Court, you have an arbitrator, a system arbitrator, But that is an arrangement between the player, Kawhi Leonard, and his union, as well as the league. And when Kawhi Leonard got a slap on the wrist, only $700,000, no suspension, no voiding of the contract, no nothing. He agreed to a settlement that was quite favorable. And the question was why? Well, Kawhi Leonard's settlement meant that the arbitration option was legally removed. So now they boxed in Ballmer. Which is again no small thing. My reaction was it's no small thing to be at war with a guy with $145 billion depending on Microsoft stock, and to say you got one move and it's to enter yet more discovery.

Yeah, no, he doesn't want to be discovered. It sounds like yeah, it sounds like a nightmare situation for Ballmer at this point to get more of the details out. Like he kind of I.

There's more. There's more.

So, so your, your read on it is he just has to basically accept the consequences and, you know, get to watch as a fan for the next year. But how much does this set back the Clippers? Like, how impactful? You know, in many ways, like, he tried to cheat his way, you know, a couple of steps forward, but in the process of that, set the team back, like, basically took the team out of probably contention for, I want to say, a decade. Like, what, what, what player would want to go to this organization now knowing that they're not going to get any first-round picks for the next 5 years? Like, it just feels like a really tough situation.

The next Clippers first-rounder that they will be able to select hasn't even sniffed puberty yet. Like, we're just in a timeline that is a nuclear winter kind of feeling if you're a Clipper fan. But I wanna say that at a certain point, perhaps things get so bad that sort of things almost need to reset. And so actually the thing I'm wondering aloud with you guys is, as we contemplate what is Ballmer gonna do next, he has now threatened personally with litigation Adam Silver by name in a letter through his lawyers.

Mm-hm.

He is signaling it is wartime, and we just gave you some of the stakes around what that would entail practically. But historically, if you're gonna go to war with your sport and with the commissioner and with other owners, by the way, that's the sort of backstage politics of this. Adam doesn't do this as a crusade. He does this because he's read the room of other billionaires who own these 29 other teams. If you're gonna go to war with the other 29 ownership groups and the commissioner and the league and journalism and everything. Typically, it stops being as fun for you.

Yeah.

Typically.

Yeah.

And so does Steve Ballmer even wanna consider perhaps selling and taking a punishment that is a humiliation on one level? But also, you know what the Lakers just went for, $12.5 billion. What could you make on the back end if money was a thing you cared about?

Yeah. So yeah, that's another calculation. What does this do to the value? Like, there's no way you could value the Clippers at least within 10% of the Lakers, right?

No. Doing back-of-the-envelope math, they're not, they're not getting double-digit. They're not getting 8 figures. You know, one of the— sorry, excuse me. They're not getting 10 figures. What I think is interesting, though, is they're gonna make more than people I think are prepared to intuit based on the fact that, intuit is a pun not even intended, based on how horrific this franchise has been. We're just at a place in sports where the scarcity is so clear that these are detached valuations from revenue. Again, there are a couple places in this world where that is true. You guys cover a lot of them. Sports, though, is on its own crazy ride that is getting divorced from the reality of things.

I don't know, I think sports and the private markets are pretty neck and neck tied at this point. But I take your point.

No, and by the way, that's the story of the Lakers and Mark Walter, which we are also investigating. My point being that they're not gonna get 12.5, they're not gonna get 10. But for Balmer, it's a trade-off between what do you really want? You have all of the money. What do you want? Do you want to fight? Do you want to surrender? Do you want to walk away with something like your, you know, ego or dignity, whatever? I don't want to be too over the top here, but like, he has a calculation that is not a normal calculation to make. Because right now it's signaling that he's gonna fight.

Yeah. Yeah, fighting, fighting seems like, uh, rough. Going to war with you seems pretty rough. Do you feel like this has probably been happening? Do you think it's possible that Ballmer felt confident in doing something like this because it was happening at a larger scale? And if you actually look at a bunch of other teams, it's really much more widespread. And if you start looking at a bunch of these endorsement deals, There's definitely a better way to go about— you said you could have a whole show about getting around the salary cap. And I just have to imagine there's so much money involved. There's so much ego. If you own a team, you wanna compete. You wanna win. And I would guess that there's a bunch of other owners and GMs and things like that that are kind of a little bit nervous now. Because they may have done something similar or even at a smaller scale?

I think 2 things are true. One is, by degree, lots of other examples of salary cap circumvention, funneling money to players, have happened, will happen, are happening. But the other thing that's true is that in terms of the scale and ambition and the paper trail and the sort of almost just like subplots Nothing is quite like a story in which a guy who doesn't wanna do anything, who has zero discernible endorsement value even when he does something, has leverage over the richest owner in sports who is willing to spare no expense, almost literally, to funnel money through 4 different companies. One of whom, by the way, is a public company, Daktronics, that has now acknowledged in yesterday's earnings call that their CFO, according to him, is now dealing with an SEC investigation. So there's also a civil suit happening in LA court against Steve Ballmer personally by 11 Aspiration investors about fraud. My point is there are lots of other examples about how to circumvent the cap that are, I'm sure, getting no attention, and maybe we'll get to some of them. But in terms of what this one is, the reason it's the biggest punishment is because they found stuff that is singular in terms of all of the webs that it entails. And that is Well, I just—

but now the people that— the person that is most incentivized to uncover all these other salary cap circumvention, you know, potential deals that have been happening is actually Steve Ballmer. Because you imagine in his war, the only thing that he's— he's like, he's got his back against the wall. He's sitting there thinking, the only thing that I could possibly do to kind of clear my reputation is make it come out that—

You think he's gonna start Are you gonna be feeding Pablo tips?

Yeah.

Is he gonna be your number one tipster?

I, I, I would gladly welcome backcheckable information from any source. I also welcome the idea that maybe the biggest consequence of my reporting is that Steve Ballmer becomes himself a podcaster, and that would be something that I welcome as well.

That'd be fantastic.

Final job. Once you have everything—

Well, he's got a lot to learn.

It's only you and the mic.

I will say Based on what I've learned from your show, that trajectory is also an increasingly common thing for a billionaire to do. So perhaps we're not so different.

Yeah. Well, thank you so much for coming on the show.

This is fantastic. Honor to have you. And congratulations. I'm— thank you, guys. Yeah. Congratulations. I mean, it's a strange thing to say congratulations for a story that is ultimately quite sad and unfortunate. But yeah, credit to you for this.

But the craft of journalism is on display here.

Yeah, look, as always, um, comedy is when bad things happen to other people. Um, so there are Laker fans who think this is the funniest story in the world. I do shed a tear for the Clippers fans who are like, yeah, really? Again? Again? This again? Another owner? Another owner that we might not— anyway.

Yeah. Uh, last question, is it possible that he sells the Clippers and overpays for another team just to get right back in the game?

Look, what he wanted from the beginning, what he was interested in, of course, because of Microsoft and geography, was a Seattle team back again. Now, I will advise this just as a podcaster to a potential future podcaster. I am told that one way to be approved to buy a new team is to not go to war with the 29 other ownership groups that will need to approve that. So just a consideration as he continues to examine contingencies.

That makes sense. Well, thank you so much for coming on the show.

Yeah, thank you so much. This was a— anytime. We really appreciate it. Congratulations on your overall run. Everything's fantastic.

We'll talk to you soon. Anytime. Thank you. Goodbye. Let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Up next, we have Mohit Arora. Ron from Cyfin, coming out of stealth, live on CBN.

Finally.

How are you?

Here he is.

Welcome to the show.

Hey, thank you for having me here.

Thanks so much for hopping on. I would love to start with a little bit of your background. This is your latest company, but not your first. Can you give us the career highlights thus far?

Yeah, absolutely. You know, my last company, was called Cohesity, and the one before that was called Nutanix. So I'm the founder of 2 previous companies. Both of them fortunately are doing north of $1 billion in ARR. The first company is already public. Yeah. And this one is, Cohesity is yet to go public.

So you had a tough time raising for the new company, I'm assuming.

Sure, it was a quick turnaround.

Really, you really had to, you know, do all the laps up and down Sand Hill. Yeah, scrape, scrape together pennies to get this done, I'm sure.

No, the round came together, I'm sure, pretty smoothly. But, uh, what were you laying out? I imagine it was like, I'm— I've, I've gotten to a billion of ARR twice, here's the plan to get to a billion of ARR really, really quickly. What, what is the plan? What was the pitch?

Yeah, the pitch was that, look, in my first company, uh, I brought together or converged infrastructure.

Yeah.

In the second company, I converged data. I'm here to converge context for go-to-market team.

Okay. Somebody had to do it. Somebody had to do it.

So I imagine with the connections that you've built through your previous companies, you can start upmarket with your go-to-market. Is that the plan? Are you thinking work with bigger companies, bigger deals, or do you still want to scratch the bottoms-up itch, the smaller, more agile early adopter crowd? What's the go-to-market here?

You know, I'm very passionate about serving customers. I think the problem is all over the stack, both small companies and big ones. My sweet point is obviously going to be the the big enterprise, but I don't want to leave the smaller companies behind. So it's going to be all of them.

How do you think about the actual value that the product will bring to these teams? Because there's been— there was a little mini boom cycle for like AI SDRs, which feels adjacent. And a lot of people were saying, hey, it's way too early to let an AI, any AI, even if you're wrapping a frontier model, I don't want it emailing my customers who I might have been golfing with this weekend, and you don't have that context. And so I don't want you embarrassing me, but it feels like that's where we're going. How do you see the process developing over time?

Yeah, I focus on the problem. And the problem is something that we describe as the context gap. And I've seen this problem you know, running my previous 2 companies, even pre-AI, right? We execute our go-to-market operations without enough context.

Sure.

The context is all over the place. It's sitting in multiple tools, people's heads, and it's not even accurate. It doesn't reflect reality. And we are trying to make decisions without it, right? And the way we try to do stuff is we maybe do a lot of meetings. 80% of the time spent in meetings is to gather the context, Yeah. Not change the news, right? Gather the news, not change the news. And so this is a unique opportunity, especially with AI, to actually solve that problem.

So how much— yeah, because I feel like if you were able to create an 8th day in the week and sit down the whole team and say, everyone dump everything into some voice notes and some shared text files, the frontier models can summarize and put all of that together. But it's really a data acquisition problem. And so how much of this is about enabling teams to do change management to the way things work versus autonomously going and gathering data? Both? What's the actual solution?

Yeah, the actual solution is all of that. So automatically gathering the data, making sure that humans are in the loop so that AI is not just hallucinating and you know, putting garbage. It is also in helping people in real time, real-time coaching and stuff, where they can ask any question or know anything they want from the context. So the solution includes all of that.

Well, it's— oh, sorry, Jordi, please.

Key lesson from the first company that you brought to the second company, and then key lesson from the second company that you're bringing to this company.

Key lesson from the first company Actually, both my companies, I would say, is it's all about the people. It's all about the people inside the company. It's all about the people outside the company, which is our customers. And so you focus on that and the right things will happen.

Bearish on AI. Just kidding. It's your 3rd company, but it's your first gong hit here on TVPN. How much did you raise for this one?

I think it will be the last.

How much did you raise?

How much did I raise? $44 million.

Well, thank you so much for coming on the show and breaking it down for us.

Let's do it again soon.

Great to meet you. I'm sure there'll be many milestones. We'll talk to you soon. Cheers. Have a good rest of your day. Let me tell everyone about Figma. Agents meet the canvas. Your AI agents can now create and modify your Figma files with design system context. You don't get a lot of CEOs, a lot of, a lot of Successful founders that say, you know what? It's really it's it's not about the people at all. It's it's actually just the code. It's just the AI. It's just the AI agents. That's all that matters. No, it of course is the people. Well, and we have a great person next.

Often the best advice is the most simple, Lindy.

Straightforward. Yeah, it's true. Well, we have Akshay from Pocket. He has sold over 200,000 of these devices. What's going on? You what they look like on the screen. Welcome to the show, Akshay. How are you doing?

Hey, John. Hey, Jordi. How are you guys doing? It's great to be in TBN. I'm glad to have you here.

Yeah, no, we, we, we started hearing and learn— learning about your company probably a year ago at this point. We're like, what's this company that's just selling a lot of hardware? Yeah, somewhat, somewhat under the radar. I'm sure it doesn't feel like that to you, but you guys haven't really chased the spotlight as much as other companies that sell hardware that haven't sold any hardware.

Yeah, I mean, the, the, the phrase in Silicon Hardware is hard.

Don't go after it. It feels like AI has unlocked new opportunities, both in— there's demand for new devices, and then there's also AI that you can ask to help you procure things and help you design things. But what has the road to actually building the first device, then scaling to 200,000 devices been like for you?

Yes, interestingly, we actually built an app before we actually built the hardware.

Yeah.

And the simple thing was that, hey, it's the most obvious thing to do. It's the most easiest thing to scale. So let's just go build an app. And we built a meeting note-taker app and we released it to a bunch of people and there was not a lot of response. And being surprised, we thought to ourselves, hey, what about that device that we were all thinking about that we should go do that'll be like a necklace? And that's kind of what we did with open source work at OMI. We went and did a bunch of Kickstarters and we printed a bunch of these devices and gave it to people. And we actually saw that people were using the device 10 times more than the app. There were a couple of other challenges regarding being, being always on and being a wearable. So when I started Pocket, my first immediate thought was this should not be always-on. People are still not comfortable. There's still dialogue that needs to happen before always-on wearable becomes a thing. And we said, let's go build something that's available for a phone. So, so the people on Wall Street are happy to use it and it doesn't augment their style of their suits and coats that they wear.

And that's interesting.

That's kind of where Pocket came around.

That's very cool. So yeah, walk me through the current product specs, price point, uh, what the favorite features are, sort of like how you think about where you are now and then where you want to go.

So currently, this, this is Pocket, and this retails for $129. The hardware comes with a freemium subscription, so you kind of get to do unlimited summaries, unlimited transcripts. You get 3 messages you can ask about your conversations per day. If you would want more, like speaker tagging, advanced speaker recognition, all these kinds of things, and advanced AI models for reasoning and summary agents that can make better details out of your summaries and have access to 300+ professionally sourced templates like SOAP notes for docs, Oh, interesting. Case notes for lawyers, and CRM entries for salespeople. You can get on our pro plan for $20 a month or $200 per year. And that's kind of the whole product, basically.

For healthcare, sorry. So the thing that stands out to me is, like, you launch the initial app. You're not seeing the kind of growth that you wanted to see. And so then you're thinking, OK, we should launch— the solution is to launch a hardware device.

Let's turn up the difficulty level to 11.

Which normally, normally you'd think like, okay, the product's not working, add a hardware thing that functionally does a lot of the same stuff. It's probably not necessarily gonna— I don't know. My intuition would be that it wouldn't solve the problem, but I keep coming— like, what is— why does it make sense for this to be a separate device from the phone? Because I'm sure, you know, if you met with investors, a lot of investors would be like, well, I have a device that's a hardware device and it has battery, it has a microphone, it has a screen, it has all these things.

Yeah.

So, why are your 200,000+ customers saying like, no, I want this separate device? Because clearly, the customers are right.

Yeah, absolutely. It's 300,000 now.

And you got to hit that gong first. GONG!

Right. So, when we actually went and asked these people like, hey, Why are you using this 10 times more than the app that we had? We actually had one of the users who was actually using a different phone he just bought for recording conversations.

Whoa.

And he also had a battery attached to the phone on top.

Sure.

So that's kind of what Pocket ended up being a product by just watching how people use these recording apps on the phone. There are a lot of limiting factors for recording on the phone. A, you cannot record your own Zoom meetings and Google Meet if you're already on the phone. You cannot start—

That's right.

Another recording. So there are a lot of people who have issues with that, and there are people who get incoming phone calls all the time. So if you're recording something and you get an incoming phone call, it interrupts the recording. And if you take the phone, your recording is stopping essentially. Apart from all of these technical reasons, the real reason is actually that The people who take 10, 15 meetings a day, these doctors, consultants who are like, you know, patient goes out and the patient comes in, the client goes out and the client comes in and there's 10 to 15 back-to-back meetings. They feel awkward in the first 8 seconds to take their phone and start a recording by unlocking their iPhone and setting up the recording and hoping so that the operating system doesn't kill background running apps for a long time. So people just felt that, you know, this is much more reliable thing, a quick access thing that, that can immediately start recording, and I don't need to feel awkward on placing it on a desk. So people don't take, take their attention away. Why is this guy looking at his phone? We just met. Yeah. So that's kind of what we came around, that, you know, convenience beats intentions. So if your users have intention to record, And if you don't make it convenient enough, they will not record. But if you actually make it convenient enough to record and they want to record even a little bit, they'll 100% record.

Interesting. Should more people be taking notes with and recording their conversations?

Oh yeah.

Because I haven't recorded a conversation in years.

Yeah.

But what am I missing?

Yeah, absolutely. I think like there are a lot of—

I guess I've recorded hundreds of hours of conversations through the show. So, yeah, maybe I have to count that.

We have people recording, you know, tens of conversations almost daily. And most of these people are actually field workers, you know, consultants, salespeople, real estate agents who are showing people's homes that are noting down requirements, places where note-taking is actually useful. So for example, DoorDash is one of our customers. They, they use Pocket to go to restaurants for, for the salespeople to record their pitches and then come back to their HQ and sync with everybody else in a knowledge base of, hey, how did my pitch go with this vendor, with that restaurant owner? So I think like there's, there's definitely like a use case for field note-taking, and that's kind of where Pocket comes in. Sits in. But, but yeah, I think like in a broader sense, your AI needs a lot of context about you to help you. Yeah. And every single time you essentially use Pocket, you can connect your Claude MCP or OpenAI's API and your ChatGPT already knows about all your conversations and you can, you can pretty much manipulate all your transcripts and, and understand how people are talking to you and all these cues. For example, if you go to a VC meeting, the VC says 100 other things, like, it actually sounds so interested, like, I felt they almost invested in the startup, but they never said the word invest. So, the founder is going back thinking, oh, I have a term sheet or something, but the truth was, it was just a hi.

Yeah. No, no, I've seen that play out where someone comes away and they say, oh, yeah, this VC VC said they're going to invest, they tell it to another VC, they call each other, I didn't say that. And it becomes—

Yeah.

Who said? Are you broadly bullish on new AI hardware? Because I think one of the reasons that people have been generally bearish is you've had the Rabbit, you've had Friend, you've had a bunch of these shots and attempts.

Humane.

Humane.

Yeah.

So there's been a handful that haven't got the level of traction that you have had. But given your experience so far, do you think there's A lot more devices to build that are AI native in the real world?

I think so. And I think like, you know, Rabbit and Humane died so, you know, Pocket could live. And we learned a lot from all these other devices that essentially died. And they primarily died because they were trying to replace the whole smartphone game.

Doing too much for sure. Too much stuff.

Yeah, exactly. They were actually just going against going towards real-world work, the way that the workflows work for real people. They don't want all the other things like ordering Uber Eats from AI. All they want is that AI to fit in their existing workflows.

Yeah.

What we did was we looked at a lot of people, had a lot of meetings, and And we didn't go into things like personal note-taking, for example. All these others are like, you know, dead, basically. Whoever went and did anything other than note-taking— and note-taking was essentially a product-market fit. It was almost like fitness for Apple Watch kind of product-market fit. And everything else was basically like people didn't want them.

Yeah, yeah. I— when I saw Rabbit R1, I never got a chance to actually use it, but I was I think, like, it's such a beautiful design. I thought it would be perfect for kids where if they're just going around taking some photos and learning a little bit about the world. And I actually wound up buying a different product that's a circular device with a screen and a camera on it and it has little games where the kid can go and say— it'll say, like, find something flat, find something round, and they'll go and find it. And then that other company, Stickerbox, that we've had on the show is doing really well in what I would call AI hardware, but you push the button, you say a prompt, and then it prints a cartoon for the kid and it makes a little sticker. And it's these like very focused things that have— it's just, do you want this at this price? It'll do this for you. It makes a lot of sense. What are you doing or what do you need to do on the regulatory side? Because I imagine if it has a radio in it, needs FCC approval. But you mentioned the medical discipline. And HIPAA, does that— is that relevant to you? How have you solved that problem?

Yeah, so we're HIPAA compliant and we essentially have HIPAA compliant servers that we run HIPAA workloads on specifically for medical use cases. And yeah, like from day one, we thought about being HIPAA compliant and SOC 2 compliant for enterprise as well. So FCC is definitely like if you're making a device with with Bluetooth, radio, you pretty much need FCC. So from day one, you would need that as well.

Sure.

So yeah, in terms of like, you know, consent is what comes up most of the time for these kind of recording devices. And what we did with Pockit was we said we would like to bring the risk level down to almost a user action. So everything happens with, with a user action. So you're supposed to like tap a button to start and stop recording, although it might be a friction step to do this. But we would love for, for the user action to be the risk level where we are for Pocket. So everything starts with the consent of the user. At least it starts with one, one party, one state consent. And then You know, it's on the user to basically ask for other people's consent. And surprisingly not, people actually are very open when you actually ask them to record.

Sure.

And they're actually happy that you would share the transcript and the meeting notes of the recording.

Sure, sure.

Do you think you'll ever launch something like a coworker agent or assistant functionality? Is that— because I imagine if you can give your device a bunch of context, hey, I want to do this thing, eventually it could get to the point where you can send off tasks in the background that the user can check on maybe on their mobile device.

Yeah, so that's actually on roadmap in the next 3 months that we would actually like to build agents that would actually take context directly from your conversations and actually act on those, make reports, make docs, PPTs, presentations, slides, all kinds of things from your meetings. So this is like increasingly becoming like the regular workflow all kinds of consultant sales pitches that people take the conversation and make slides out of them and come back for the next meeting. So we would love to make an interface for that. So in the next few months.

How do you think about picking different AI models for various pieces of the workflow? Because I can imagine that you're selling something pretty valuable, so cost might not be the most important thing. At the same time, you can just wait and things get cheaper often. Switching from one model to another might inject like a different flavor or vibe sometimes. So how have you thought about this? Are you fine-tuning your own models using open source, bouncing back and forth between whatever frontier models are available? What's your decision criteria right now?

So surprisingly, we have the highest margins in the subscription industry for AI.

I can imagine.

Roughly 75% margin in software subscriptions, essentially. So we do this with different kinds of routing.

Oh, okay.

Yeah. So we have our own version of OpenRouter where we essentially route to different kinds of providers and models. We use a section of open source for transcription. Sure. Because people love to choose what model they would like to use for summarization.

Oh, they do?

Okay. Yeah. So transcription is pretty much left to us. So we do it on ourselves. So that's kind of why our margins are super high because we get to use whatever we like. Sure. And we use our own models that are fine-tuned on top of OpenAI's Whisper, uh, because a lot of these online models were trained on YouTube datasets like WalkCello, uh, which are extremely clean datasets, um, and work really well for Zoom meetings and meetings of online note-takers. But when you come for offline, there's like a train going next to you, there's, there's interesting things happening. So they actually go, go go crazy when, when there's like an offline recording. So we need to fine-tune for them to work for, you know, offline recordings essentially.

That makes a lot of sense. Uh, thank you so much for coming on the show. I forgot that I am a featured customer on the Zapier blog from 2015 because I developed a workflow that would record all of our conference calls Send them to a human transcription service and then upload the transcription into Google Drive. And then you could search by keyword just using Google Drive search.

Powerful.

Nowhere near the power level now. But I was thrilled.

Interestingly, Y Combinator actually currently uses Pocket to record all their interviews.

Oh, really? Cool.

And send them to their online systems through our APIs.

There you go.

So yeah.

Very cool.

Something similar.

Well, congratulations on the progress.

Great to meet you.

Thank you so much for coming on and breaking it down.

Love the approach.

Have a great rest of your day. We'll talk to you soon.

Thank you, guys.

Let me tell everyone about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web apps, servers, databases, and more, while Railway automatically takes care of scaling, monitoring, and security.

Who we got next?

Our next guest is Hari from Aura. He's the founder and CEO with great How are you doing? Welcome to the show.

Hey guys, good to see you.

Since it's your first time on the show, would love an overview of the company and yourself, and then we can go into the actual news.

Sure. The company Aura, we're in the consumer security, consumer safety space. So basically, if you thought about all the various different ways that a family could get impacted by using services online, scams, spam, transaction fraud.

Yeah.

All the way from that, all the way to making sure that if you have teenage kids, for example, if they're using devices too much, we have ability to kind of figure out how best to sort of guide and navigate them around digital detox, et cetera.

Okay.

So yeah, so that's the core of the product. And I've been an entrepreneur my whole life. This is my third company now, so.

Awesome. How big is the problem? Now that I'm thinking about it, I mean, we were just at CrowdStrike's Falcon on Tuesday and we heard that phishing open rates, phishing email open rates have jumped from 12% to something like 60% in the AI era. And when I actually go back to the number of cybersecurity scams, I have just as many examples of friends who thought that their grandmother was asking for gift cards as I do in a corporate setting. And so how prevalent is this? What are the— what's the data that you monitor to understand the scope of the problem?

Look, I think that over the years, both the breadth of the problem and sort of the number of people it impacts has gotten astronomically bigger, basically, where certain scams used to be, I don't know, a few thousand dollars worth of scams. Now we see that's gone up to like $25,000 per scam on average. The incidence rate is quite high because Now that there are a lot more AI-based tools, the bad actors also use the same tools to do deepfakes, to do ID theft. With a lot more data out on the dark web, that ends up becoming a precursor to basically being able to use that to do nefarious things to families. And I think the other big thing is if you thought about kids where when I was growing up, we weren't native to phones and native to online services. My kids, I mean, they are digital natives, so they basically, you know, are on their devices all day long, which means the incidence of them putting information out there is just much higher. And so as a result, there's just more and more and more of this happening, which is pretty scary.

Talk about the recent hack leak. There was some ID verification service that had a breach, and I think from what I could see, it seemed like Half the people in the US were impacted, something like that, with their now, you know, image of their ID out on the web. But yeah, break down the breach and then, and then what people should do.

Yeah, look, I think, you know, starting with even a couple of years ago, there's the NPD breach, the national, the national levels of Social Security number, etc. Now there's a lot more ID digital image theft, etc. I think the key thing I would say is The incidence of these is so high now that when people hear about it, you just start kind of immunized. You're like, okay, well, whatever. I'm sure my data is already out there. It's what's going through people's minds. The trouble with it is this is just an early tell for the next set of things that are going to start happening. So for example, at Aura, we have a fully trained foundation model that is looking at sequences of stuff, right? Hey, what's the first thing that happens? And what's the second thing that happens? What's the third thing that happens? How can you sort of analyze risk with systems? And we see that with data breaches and information getting out on the dark web is a pretty early and good tell. If it's happened multiple times, you see that the incidence of actual identity theft is bad on the back end. So the very tangible things I'd say, you know, monitor your credit, make sure that you're looking at your credit card numbers and statements. Don't wait for the whole statement to show up.

Mm-hmm.

If you start seeing odd things on physical mail, like, you know, keep an eye out for that as well. you know, weird situations where people have stolen information and, you know, it gets sold to folks for physical crimes as well. So we see a lot more intersection of physical and digital happening a lot more these days as well.

So interesting. What is the industry side of the business look like? Like, I imagine that as you're protecting homes, schools, individuals' identities, There's the flip side, which is if I'm a business, I don't want someone purchasing with a stolen identity and having a clear communication line between those 2 parties probably makes sense. But what does that actually look like in practice?

I mean, I think that's a great question, right? 'Cause if you thought about your own life, like, or me, the surfaces I spend time on are my home, my workplace, and a parent, as a parent, my kids' school. So there's a lot of data that kind of goes across that stack. And even more than just, you know, are you using a nefarious card for, you know, for purchasing something from me? Even more basic than that is a lot of the enterprise scams like the ones you were talking about, even with Falcon, for example, many of those originate on the consumer side. And many of the breaches that happen on the enterprise side is social engineering attacks anymore, right? So it's basically going through people who are employees.

Yeah.

And then kind of using that to get into the enterprise. So on both those pieces, I think getting the consumer to be more aware of these issues ends up becoming pretty critical. We see a lot more overlap now where, you know, we do sell our product through large enterprises as well. And so we talk to a lot of CISOs of big companies and the feedback we get is, look, the lines are very blurred between what's at home, what's at work. And so even besides sort of the loss prevention piece of it, we just say you use a fake credit card to buy product. It's even more imminent that they want to make sure their enterprises are kept safe. And sometimes, you know, keeping the employees safe is the best way to do that.

Yeah. Yeah. How much time do you spend working on streamlining onboarding? When I think about these security products, we've talked to the CEO of 1Password, a fantastic product, but onboarding can be like a stumbling block for people that You know, want to secure their life, secure their identity, but the actual integration of getting deep into all the accounts, collecting everything, making sure that everything's set up properly, that can be a hurdle. I imagine that you have low churn, but the conversion to, I heard about this, to actually, I'm using it properly, has got to be a hurdle. How much are you focused on that?

We spend a lot of time on that. Time to value is a huge metric for us, which is, you know, how How quickly can you get the customer to something of value, basically? But I'll say 2 things. One is on the way that the industry works today. If you thought about a product you bought, you know, for, for either of you guys, right, it's going to look the same. Like you buy something that's going to keep your identity safe or you buy something to keep your passwords locked up or even something for kids, for example. The solution doesn't really understand or know you. It knows the problem, right? And so one of the things we have done which is very different is we're probably the only product that puts the actual family at the center. So there's actually a large sort of family graph that maps out all of the relationships for people. Who are you? You know, who are the people that you sort of, you know, connect with? And then all of the reasoning and the inference systems that we run on top of that have good ability to have a view of you end to end, which means that when you go into onboarding, we can do a better job first getting you on board and then as we're watching and observing, we want to make sure that, uh, that, um, you know, you— it's very personalized for you and your digital footprint.

We have a massive gong here. Any, any recent revenue milestones or growth, growth milestones you could share that we could, uh, use to hit this gong for you?

Oh yeah, sure. We, uh, we hit— we're about $340-some-odd million of ARR. So So that's a great accomplishment. Thank you. Thank you.

Not too shabby.

And thank you so much for coming on the show. Yeah, great to see you guys.

Have a great rest of your day. We'll talk to you soon.

Take care, guys.

Let me tell you about public.com. Investing for those that take it seriously. They got stocks, options, bonds, crypto, treasuries, and more with great customer service. Uh, GPT-6 Astra has landed. The blog post is up. You can go check it out on openai.com.

To the stars.

To the stars. And the headline, I believe, good performance on a bunch of things, but the ArcAGI-3 number is crazy. The score is 99.9%. So it feels like they just beat that. They just beat ArcAGI-V3. So that's the video games, the ones that Tyler was briefly—

Globally ranked.

Globally ranked. You're out of a job, Tyler. Say goodbye. Hang up your ArcAGI V3 hat. Don't worry, the team over at ArcAGI is working on V4. We're going to move the goalposts soon. We're going to move the goalposts soon. But this is very impressive if you've played around with ArcAGI V3. It requires some real creative thinking to actually learn how the games work, how to be efficient. What else is sticking out? Has anyone been able to monitor the timeline at all?

Very chaotic launch. So I think it's hard to get a real reaction yet. Tyler, what are you seeing?

Yeah, I mean, there's still no actual post on OpenAI's like X account. It's just the blog post that's now live.

Okay.

But I mean, there's a bunch of benchmarks in there that are—

Yeah. Terminal Bench Science. We're going over into the world of science. GPT GPT-6 Astra scored 64.6% on reasoning effort max, cost $26. I'm sure there will be a lot more. Also, the Exploit Gym honeypot, lower is the better, GPT-6 Astra, 0%. So good performance there. The world's best computer use model, I'm putting that to the test ASAP. Let's see if it can 1v1 me on Rust, because if it's good at using the computer, it should be able to noscope, right? That's the bar. That's where my goalposts are. Agent's last exam, good performance. So you can go check it all out. I'm sure we'll be digging more.

Sam is going to be on Bloomberg in just a few minutes over with our, with our buddy Ed Ludlow. So that'll be fun.

So we'll be digging into this, and we have some special guests lined up to talk more once we've been able to digest, take it for a spin. And have a lot of fun with it. We have our next guest joining in just a second. But first, let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in-store, on mobile, on social, on marketplaces, and now with AI agents. Very exciting. We have 2 guests joining at the same time, Matt Caldwell and Jordi Lycer from the Minnesota Timberwolves. and Lynx and Jump. They're both CEOs, and we're going to have them both on at the same time. Let's— we need to check in on Ed Zittrain, see what his reaction to Astra is. Is it any better, or is it just more of the same junk? Because he has been a bear for a long time. He's coming up on the 3-year anniversary of the bear post. Dan Liu has collected all of Ed Zittrain's position.

You think he has sort of a gambler's mindset? Like 80% of gamblers quit right before they're about to hit it big?

Yeah. I think it's just an entertaining shtick. It gets views. Kevin Roose, formerly of the New York Times, now independent Hard Fork co-host, who's writing the AGI Chronicles, On sale October 10th. He says, anyone who has interviewed Ed Zittrain or had him on their show in the name of AI skepticism has made their audience dumber and less prepared for what is actually coming. Wow, that's a crazy thing. Eric Newcomer chimes in and says, it's his audience, they follow him around, which is interesting because I believe Eric's best episode of all time was with Ed Zittrain. And he was like, the Ed Zitron fan base like stuck around for a couple more episodes and was like, I don't like this guy. I don't like this at all. I don't want reason. Wait, reasonable takes?

And Eric, of course, is— I don't want to see both sides.

Sometimes Eric's skeptical. Sometimes he's optimistic. But he uses these and he doesn't believe that it's like AI's not—

RSC says when you ask Ed Zitron why he keeps writing all His predictions are— when all his predictions are wrong. And there's a quote from Dion Waiters. I'd rather go 0 for 30 than 0 for 9, because you go 0 for 9, that means you stop shooting. That means you've lost confidence.

That's pretty good. I like that. Well, speaking of basketball, we have the CEO of the Minnesota Timberwolves, Matt Caldwell, here with us, and Jordi Lysaker joining as well. Welcome to the show. What's going on, fellas? How are you doing?

Doing great. Hey guys, see ya.

Great to have you.

Thanks so much for hopping on. I would love to start with the news. Can you both introduce how this all came together and the various roles that— how you guys see yourselves working together?

Yeah, Jordi, you can kick it off.

Yeah, well, great to meet you. Thanks for having us on. Big fan of what you guys have built and been doing here. Basically, I started this company Jump with Mark Lore and Alex Rodriguez. I think you had Mark on actually.

I saw—

Yeah, we did.

Talking about— you also had Alex Bohanian, one of our, one of our boys with Jump too.

Yeah. Nice. I love him.

Yeah. So the whole theory was that Mark and Alex were gonna buy a team, and if they were gonna buy a team, these guys are not sort of in the business of doing things the old way. And, and so buying the team, actually, I just saw the Shopify plug. We're trying to do for sports teams what Shopify does for merchants, which is basically totally unlock direct-to-consumer proposition for a sports team. And so Matt was hired, you know, right when Marc and Alex took over the club. And so we got to know each other and then we were able to launch our platform in the NBA with the Timberwolves, with Matt, who's, who's an unbelievable leader and sports executive under the ownership of Marc and Alex when they finally took control of the club.

Yeah. When you think back to sports ticketing generally, just the different eras of technology, the risks, what they unlocked. What do you draw on as things that organizations got right? Any organizations that jumped on waves early and saw success? What are you pattern matching against when you think about where you want the team and the organization to go from here?

Yeah, Matt. Yeah, sure. Thanks again for having me on. Listen, traditionally, teams outsourced ticketing, right? If you look at it from a team perspective, like there's a lot of work to be done. You're trying to win games, you're signing players, you're marketing the brand. And, you know, to create your own ticketing system that's transacting and selling season tickets and single-game tickets, you know, it's a lot of work. You know, Ticketmaster has built a great business on this with both concerts and, and, you know, sporting events. And, you know, I think I think in the last 10 years or so, teams are saying, OK, we have a lot of season ticket holders, but we don't really have data on people that just come to one game or pick up a couple of games during the year. And when all that data is outsourced to another provider, it's hard to figure out the behaviors of that fan and how do you market to them? How do you make them a season ticket holder? Essentially, just grow with that fan. So what Jordi has done with Mark and Alex built a ticketing system that's in-house that when someone goes on our app now, and traditionally we've had our schedule and some content and highlights on the app, but now you can just scroll and buy tickets on the app. You don't even know that you're on Jump's platform. It looks like the Teams app. So it's a way to have the data on your side of the table, on your app, and You're working with a great partner, you know, with Jordi and John.

So how do you think about actually scaling the marketing now that you have more data, more of a direct relationship? I've thought about times when I've gotten invited to a sporting event and my friend might have bought all the tickets. Occasionally they'll say, oh, to get in, you actually have to download an app and claim the the ticket, and then you'll have it on your phone. And then I should be getting retargeted to return the favor. Maybe I should buy them a round of tickets next time. And I have probably been served a few ads. I don't know, in one ear, out the other, a lot with this stuff.

Right.

But how do you think about building your marketing organization internally? Shopify has some buttons where once you're set up, you can just say, oh yeah, run some Facebook ads. You're at a much larger scale. How do you think about the rest of the organization internally executing this strategy?

It's a great example. Traditionally, the experience you talked about, your friend transfers you a ticket, you download Ticketmaster or StubHub, SeatGeek. All that data, that person's information, cell phone, email, all stayed with the third party. Now we're getting access to it so we can have our sales staff call. We can call them. We can have automated emails going to them.

Yeah.

We can go visit them in their seat. We know someone came in. We know their name. But something that's really helped for us going to the next level is we've gotten to the point with Jump where even if someone puts a ticket in the shopping cart—

Sure.

—but doesn't close it out, or even just browses over a game, we get that information. We get an automated email to that person. and, hey, these 2 tickets are still available. Or even better, hey, those 2 tickets were just sold, but there's 2 other ones that are similar. So you just have just more engagement, especially with what I would call the casual fan, single-game buyer. Like, our season ticket holders, we know them really well. They sit courtside. They're in suites. You can visit them at the games. Every season ticket holder has their own individual concierge sales rep. But it's the people that, like yourself, got transferred a ticket, come to a game. We want to make them longtime fans. And now we just have so many ways to reach out to them now.

I went last time, I went to an NFL game, and I swear I had to download 3 apps to get the actual ticket. And even by the end, I'm like, I hope I got all the apps I need to get in to use my ticket.

The parking ticket and then the hot dog app.

And it was a free ticket too. I was like, it shouldn't be this hard.

How are you, Jordi?

It was a lot of work.

Yeah, yeah, a lot of work.

Time is money.

A lot of work.

Time is money.

I was committed though. Jordi, how are you thinking about like using this to deepen the like fan engagement, knowing Alexis, like he's obsessed with collectibles? I imagine by having that direct relationship, you can unlock different products and things like that for fans based on their actual engagement with the team so that You know, fans have to actually like earn certain things. But where— what does that look like?

Yeah, you nailed it. And actually, this hits the problem of, you know, the 3 apps and the new login and the Lakers do some deal with some startup and then you got to create a login for that thing to like— it's just the whole thing is just one Frankenstein tech stack, you know, mess. And so again, Shopify is the best comp. Like it sort of consolidates that into one seamless this experience, then everything else plugs into you.

Yeah.

So in that same vein, collectibles, merch, concessions, add-ons, parking— you should have, you know, the Timberwolves login, the Timberwolves account, add to cart cool jersey, add to cart cool collectible. Oh hey, I know that you, you are a season ticket member and you came to— you know, KG's coming back for the Wolves this year, it's a big part of the story. So they should know that these people who were there during the KG era You know, it's not just that they know the CRM like we know Bob Smith and we know where he lives and we know it's like, did Bob go to those games? Was Bob, you know, kind of there during the era? And then you can nudge and you can, you can sell and you can upsell and you can engage. And so it's content, it's merch, it's media, it's everything sort of in one system.

Are you thinking about cross-selling across different sports? When I think about the Shopify example, they've launched the Shop app. You know, you— oh, you bought this pair of boots, maybe you want this pair of pants. Is there a world where you bought this ticket to a basketball game? Here's, you know, it's the offseason, here's a football game or a baseball game. Is that something that's a little bit more—

That's a little more the market. That's kind of the marketplace model that, you know, that's the— there's a bunch of ticketing marketplaces and there's YC versions of that and there's stuff up, you know?

Yeah, yeah, yeah.

And I think that that's the thing that, you know, Jordi, you were saying It's just kind of the beginning of the mess of this whole thing where if I want to buy, you know, Delta Airlines, like if I'm going to fly on Delta, I get the Delta app, you know everything about me. Yeah, they went direct to consumer 20 years ago and like, that's, that's it. Yeah. So I think that's the missing piece in tier 1 sports at least. But it's really like if you're, if you're in Minnesota, you've been to a Wolves game, hey, let's kind of like get to know you as a Wolves fan, just like you were, you know, a customer of any other brand.

Yeah. Yeah.

How sharp are the elbows of the, of the other players in ticketing? I mean, I can't imagine they're thrilled that a platform that's allowing people to go D2C popped up with a bunch of backing.

What do you think?

Yeah, I imagine they've got like little knives right here and they're just—

Yeah.

But yeah, it feels like something that's so aligned, so aligned to the team, so aligned to the fan. It's just misaligned with the the legacy players that no one actually, I think, likes?

Yeah. Well, that's a great question. If you're investing in startups, do you want to invest in a platform that is beloved by its customers and beloved by its customers' customers, but not beloved by the 50-year-old incumbent? I like being on that side of history, and I feel like we're in a good spot. But at the end of the day, I mean, I'm in Los Angeles. I just think you guys are here. Sure. This is a town that just— and you're in this space, sports, entertainment, I would say also in like the team owner space. You know, the headlines are pretty, pretty wild recently. Yeah, it's just a lot.

There's a lot of going on.

Yeah, I was gonna say, I don't, I don't, you know, we're not super tapped into the industry, but I was gonna say, if you do any endorsement deals with athletes, make sure they actually endorse the company. Uh, just, I, I think that's what I've learned. We had Pablo on the show earlier, and that was one takeaway. Make sure they actually talk about Jump and what they're doing.

Oh, no, that's Matt's point. No comment. No comment. What else is happening in the world of technology and sports that you're tracking? I mean, AI is getting slapped on everything. And I love this story because it's just a very clear software and enablement, like the actual relationship. It gets to like the core of a business problem that you're trying to solve. And then there's technology that is really powerful, but it's maybe a little bit further out. But what else are you harnessing at the Timberwolves that you think is interesting or you're starting to explore? Like, talk to us about technology's role over the next year or decade.

Yeah, Matt, go ahead.

Yeah, I mean, the first thing that pops to mind for me is the pricing of our tickets. If you think about, we have about 20,000 seats in an arena, maybe half of them are season ticket holders, right? So you have 10,000 seats that you're selling at least 41 nights a year. And if you have a manual process where you have different price codes and there's supply and demand, and every couple hours or every day or so, you're updating prices and there's transactions happening, you're missing out on margin, or you're missing out on lowering pricing for a certain game and getting more people in the building. So I mean, AI is going to explode sports, just like it's doing for every other industry. And to be with someone like Jordi that's very AI-forward, our pricing has gotten so much more better. Automated, um, you know, we're pushing revenue. And then, and then we talk a lot about open distribution, you know, with all these different ticketing systems. You can imagine, you know, uh, there's some tickets that are on Ticketmaster, SeatGeek, StubHub, uh, they're all kind of spread out. You're not getting the full marketplace looking at your tickets all at once. So what we've done with, with Jump is, you know, any ticket we put on the Jump platform say, you know, a single-game ticket in the upper level, you know, that same ticket is now being put on Ticketmaster's, you know, TM+, you know, uh, web, you know, secondary market, SeatGeek, uh, yeah, StubHub, all the secondary markets, so that you're getting all that demand, all those eyeballs looking at the ticket. And interesting, you know, you're hopefully getting, you know, hopefully driving up pricing if the demand's there.

So, and that's not possible in the old way because obviously Ticketmaster wants them on Ticketmaster and SeatGeek And so our model is like, you know, if you got AirPods, they control the distribution. Apple says, I want some in Best Buy, I want some at Target, I want some Amazon. Like, we give the teams total control to distribute. And that's a, that's a totally different model.

Our head producer, Ben, from Minnesota, he just texted us and said, I actually bought Wolves tickets last night using Jump. It was great.

All right.

Two thumbs up. That was my plan. Yeah, I texted Ben.

I saw on your feed there was some drop about a little bit of a negative vibe about maybe the wolves won't get all the way to the other side. I saw that recently. I didn't want to call it out, but since you mentioned it, like, what's going on with that?

No, no, no, Ben's a fan.

Ben looks confused. He's a true fan. True fan.

All right.

Well, thank you so much for coming.

Yeah, great to meet you guys. I love a business that just makes so much sense. You guys were probably like feeling out this idea. It's like, is there a reason this thing doesn't already exist? Seems really obvious. Seems really aligned to everyone involved. So very, very cool. Great to meet you.

Cool. Thanks, guys. Thanks so much.

Cheers. Appreciate it.

We'll talk to you soon. Let me tell you all about Cisco. Critical infrastructure for the AI era. Unlock seamless real-time experiences And new value with Cisco Team Deathmatch. Up next, yeah, wait, we're heading to the portal. We gotta get to the bottom of this. Ben actually— this is true, Ben— you bought Wolves tickets last night using Jump?

Yeah, yeah.

Okay.

Yeah, I did. It was great.

Wait, are you a Wolves fan?

Uh, yeah.

Okay, wait, why the hesitation?

Not huge.

Not huge? It's a surprise for someone, so shh.

Oh, okay. Oh, sorry.

Okay, edit that out. It's not like it's a livestream or anything. Anyway, I believe we have our next guest already here. So let's bring in Jeff Thornburg from Portal Space Systems.

It's Portal time.

How are you doing?

Fantastic. Great to have you.

Welcome, Jeff. Thank you so much. Let's start with the decision to start this company. you've experienced from SpaceX. What was the original thesis? What was the first prototype pitch deck? How did you even begin this journey?

So many things, but I think the simple answer is, after 30— over 30 years in the space business, I had a lot of colleagues in the defense side that was really struggling with how do they combat adversarial advancement on And we're running our spacecraft like hot air balloons, and they're running theirs like fighter jets. And so we really set out to build the fighter jets for orbit here at Portal while also allowing more space exploration for NASA and the civil space community.

Yeah, what is the benefit of being able to maneuver in space generally? I mean, we've seen, you know, the trajectory of traditional rockets, ballistic missiles, Seems like a lot of stuff can happen in space, whether it's the Starlink constellation or the International Space Station. None of that required too much maneuverability. Why maneuverability? Why now?

Well, you've all seen the movies, right, where the adversarial country runs out in the field and covers their stuff with tarps because they know the satellite's coming overhead to take pictures. And so most things on orbit are very predictable. And, you know, to mix it up, we used to not have any real competition on orbit. And now guess what? China and Russia are challenging us—

Mm-hmm.

On the defense side. And, you know, we don't want people throwing tarps over stuff. We wanna show up and do the things we need to do when we wanna do 'em and not be constrained by the lack of fuel in the gas tank on orbit. And a lot of the ways this has been looked at in the past is it's like you went to buy a new car and you throw that new car away as soon as you use the first tank of gas. And, you know, we really set out to build platforms where gas is not in the equation anymore. Use it, move it quickly. You know, there's some generals out there that have this phrase I really love where, you know, we drive our spacecraft like we're going to church and our adversaries drive them like they stole them. And we really want our customers to drive them like they stole them. And that's really what built the foundation of the business 5 years ago.

I love that. So, yeah, what has progress been like over the last 5 years. Is this— are you at manufacturing stage? Is this prototype level? I imagine that, you know, you got to actually send stuff up to at least low Earth orbit to really test the capabilities. But what's the progression been like?

Yeah, I first have to really give a shout-out to my team because none of what we've done would be possible without them. And, you know, we're about 60 people now here north of Seattle. And the progress has been amazing. I've got to work with some amazing people over the years. The biggest compliment they've given to me is wanting to come back and work with me again. And we were able to launch our first mission to orbit in March and successfully test a lot of our electronics hardware for the brains of our systems.

Mm-hmm.

Uh, and then we just finished our first, uh, small Starburst spacecraft. It's about 600 pounds. It's gonna go up and really demonstrate our ability to do some of the things we just talked about. And that's gonna launch October 1st. So the team has been grinding for the last several weeks to get ready to ship to the Cape, and we're going to do that in the next few days. And they couldn't be more excited or more fatigued all at the same time, but they're doing a fantastic job getting us ready.

Very cool. Space, you know, conflict in space is something that you kind of— we hear about from entrepreneurs and and various people that are in defense tech, but it's not something that you necessarily read about in the newspaper. It's happening way up above us. And it's not something, you know, it involves private companies and governments, but no one is really incentivized to come out and say like, hey, we had this issue. It seems like it just kind of gets talked about behind closed doors. Do you think this is something that over time private companies will be like end up disclosing? Like, when do you think there's more of like a conversation around this, or is it something that just kind of stays talked about behind closed doors?

I think there's changes happening right now in the conversation, but it's tough, right? Because when national security is impacted, you don't necessarily want to be telegraphing all of the things that are going on. So, I think there'll be some element of protecting that information. But I think there's going to be more companies like Portal, and you're seeing them now, that are going to have a significant part of their business be defense-related. And the Department of War and the Department of Defense in general are really trying to bust up the legacy compartmentalization of information so that we can support the Defense Department and the agencies in ways that frankly, they have to have. And I think what's happened with a lot of the acquisition reform over the last few years is trying to get capability in the hands of the warfighter as quickly as possible. So there's a much more willingness to partner with industry. A lot of the decision makers are saying, hey, bring it all, we want it all. We want to maintain our number one position in the world and the Western world with our allies in controlling the space domain. So you're seeing a thaw in that. But I think I think there'll always be elements of national security where you will know there are things happening, but you won't necessarily know all the details, just so the public can understand why it's important. But what I would point people to is just take a look at that 60 Minutes report from a couple of years ago where they were with the Philippine Navy in the South China Sea and they were actually filming how China was engaging the Philippine Navy. It's all public domain. It all gives you a great idea of what's actually happening in other parts of the world with adversarial aggression and tactics. And I think it's a great example that, you know, people could then kind of take forward. And why is this important? It's important because we protect our interests in land, air, and sea, and now space. And space has to become a domain that we defend for future commercial and commerce activity.

What's the biggest lesson that you're taking from SpaceX to Portal?

Tenacity. Not taking no for an answer. Every problem has a solution, and with the right team, you can get there. And I think those are the biggest lessons I took away.

Do you have to show that to the 50 employees that you have on your team? Do you— is it enough to say it? How do you actually display it if you're a CEO? Because tenacity can look different in this particular seat.

Yeah, I think it's tenacity coupled with maintaining a cool head, working the problems as they come. I joke with my group, how many times do you see me running around with my hair on fire? That's not the right approach to inspire confidence in your team. So, I think the bottom line is you've got to project Um, a calm but aggressive attitude to work through these issues. And I think just having experience and doing that in the past, which certainly all of us that came out of the SpaceX environment have an incredible amount of experience in, in solving impossible problems and moving forward with a great team.

Well, Jeff, congrats on the progress.

Very, very cool.

Thank you for everything you're doing, and thank you for coming on the show.

I'm glad. Well, I'll be driving it like we stole it.

Awesome. That's what I want to hear.

Yeah, I want to hear about people in space. You know, you know, they say in AI, they say you don't need to drive a Ferrari to get groceries. But I hope we're going to be driving Ferraris like we stole them. Yes. In space to do simple things like getting space groceries. Absolutely.

Kind of lost the plot. Anyway, thank you so much.

Great to meet you, Jeff.

Have a great rest of your day. We'll talk to you soon.

Up next, we got People are not talking about space groceries enough.

Really mixing metaphors there, I think.

We got Charles from Base10.

Charles, the head of AI model training.

Hi, Charles. There he is.

Welcome to the show.

Hi, everyone.

What's going on?

What's the best model you ever trained?

Best model I ever trained? I can't talk about it.

Name every model you've ever trained.

The traditional answer.

I didn't realize this.

Name every parameter.

I think OpenPhysics has 5 million models trained. Is that enough? Do we need more?

Well, yeah, I guess our bet is that we're going to end up in a world with some great models from the frontier closed-source labs, but we're going to end up with not 5 million, but probably hundreds of millions of models, possibly even one model for each person, possibly tens of models for each person. And I guess that's the future that Base10 is betting on.

Yeah. So what is the framework or what is the next model, the thesis behind model development that you want to execute on in this role?

Yeah. So I think, to be honest, and the closed labs will tell you differently, but everyone is scaling the same recipe right now. There is no difference between OpenAI's RL stack and the Chinese open source RL stack and the American open source RL stack. We've got the same recipe. People believe it scales. We first scaled pre-training and model size. We're now scaling RL.

Yeah.

And people are going to keep doing that till the cows come home. And the bet is that you hit higher and higher levels of intelligence and you'll be able to do more and more economically useful things. I think to us, we care a lot, for instance, about continual learning. So we announced BaseLabs today, which is going to be doing less myopic, longer-term research around what models can actually do. And continual learning looks very different for us as it does to the big labs. Like, the big labs are already in this kind of continual learning loop of, you know, they train a model, GPT-N or Claude-N, and they release it into the world. They collect feedback on what it can do, what it can't do.

Yeah.

Then they build a fuck ton of RL environments to like patch the holes in it.

Yeah.

And they go back and like, from a high-level view, bird's-eye point of view, that is continual learning, right?

Yeah.

The continual learning that you or I might imagine is very different. It's where, you know, you have an open-source model, that you specifically are using either as a firm or a team within a firm or even an individual. And that model is organically adapting to the information that it's learning. It doesn't have to, like, you know, write everything down in memory markdown files because it learns about, you know, your business and so on. And then we think there's like major paradigms to be unlocked within that regime. And it's not necessarily going to get the focus from the big labs because that doesn't necessarily benefit them, right? Like they want to serve one big model at scale and the pitch they've, they've sold to their investors is that, you know, you do that at a large enough scale and you don't need this, this kind of continual learning.

Yeah. How do you, how do you process their claim? Because a lot of the big labs, there's this dance between, oh, well, there's a fine-tuned model. It was really good at this one benchmark. Then the next version of the big model that does everything is better than the fine-tuned. And it feels like this horse race where I can totally see the cost argument. And I can see even like the, yeah, for 3 months, this fine-tuned model was better. But if we're talking just like raw capability, how do you interrogate that claim that the big god model will always be better as long as you give it time?

Yeah, this is where I think people are thinking about intelligence capabilities in the wrong way. People think about intelligence relativistically. They say, okay, you know, the open source gap is like 6 months behind closed source. And, you know, GLM 5.3 is at the point that Opus 4.8 was, whatever.

Yeah.

I think the best way to think about what models can do for you and for the world is like, absolutely. So for any given task that you want to do with an LLM, there is some intelligence threshold that below that you can't do the task, and above that you have very diminishing marginal returns to more intelligence on the task. And so when you think about it that way, The game of LLMs over the last 5 years has been, okay, we have these things we want to do with them. Closed source hits it first, which, to be honest, we think is a good thing for many reasons, which we can get into. But, you know, open source eventually, like 6 months later or 9 months later or whatever it is, can then do that task. And then for many reasons, like whether it's to control your own intelligence, whether it's to improve at that task specifically, once you have the base level of intelligence required to do it, you probably do want to swap to open source. And so It's not really about the God model being better. If I'm filing a tax return, there is just a limit to how much intelligence I need to do that particular thing. And so I think the world is going to look like the frontier closed source labs are going to continue to push the frontier. You do want to use the most intelligent model. You have very inelastic demand for intelligence when you're doing frontier science or frontier maths. But for a lot of the economically valuable things, it looks a lot like, okay, you know, I'm a Cursor or I'm one of these big companies who are now realizing like, I can't just be a wrapper anymore. I've been through the lifecycle of building a great product that people love and they tell me what they love and hate about it. And I should be using that information to, you know, make my model better at the things that I care about and not at anything else.

Yeah.

And so that pattern was executed fairly well with Composer. It's obviously early days and the ecosystem isn't mature enough for anyone apart from, you know, the Cursors of the world and a few other big companies to to go about training. But I think eventually we'll get there where it's much more organic.

Can you talk about the goals of this project? I mean, I understand the research direction, continual learning, but is the goal more to advance the research or produce a continual learning model product that is then sold? How are you balancing discussion that needs to happen within the ecosystem versus productizing something that could be a really meaningful breakthrough?

Yeah, it's a good question. I think the founding mandate of Base Labs is to do whatever we can to make open-source models, and probably bleeding into closed-source models eventually, as useful as possible. And so it's very difficult to specify a priori what that looks like. One obvious place to bet is, yes, this continual learning paradigm and doing research around that. And thinking about that in a different way to how the big labs think about continual learning on millions of people at once.

Yeah.

But we're not hubristic enough to say that that is the only way in which we're going to make models more useful. So for instance, another thing we're thinking a lot about at the moment is everyone talks about aggregating compute for open source to keep up. You need a certain number of chips to be able to train these things, but not as many people are talking about aggregating data. The closed source labs are spending billions a year on RL environments. And again, RL is the next paradigm that everyone's scaling. Distillation and cheap data and centralized data. There's a lot of discussions around these things about how the open source labs are currently keeping up. But what keeps me up at night is I wonder if there's a point where it bifurcates and you can't get all the way there from distillation and whatever data you have access to. Someone needs, without a commercial incentive, needs to be making these incredibly complex RL environments that normally cost billions of dollars. in the aggregate and open sourcing them to the world. So a key part of BaseLabs is like, can we make those environments? And we've been doing this for the last few years, like we've been making them for individual people. Why don't we just make them for everyone and release them? So any open source or closed source provider can train on them and you kind of cut the gap that way. So aggregating data is another big bet that we want to make. And there's a few other things. One thing that we're starting to notice a lot is as the capabilities have risen and you know, even the open source models have kind of subsumed all the economically valuable tasks that like most people are going after. And it's very difficult to tell the difference between an open source and a closed source model. Then you start to think about what the values of these models are. And you can't not have an opinion when you're training a model about what the values and ethics and morality of that model is. Like in your pre-training data and your mid-training data and your post-training and your classifiers and your safety stack you run on top of it, everyone has an implicit or explicit opinion about what the values are. It's a reason why so many American companies don't want to use Chinese open source is because we're not very clear what those values are. So how do you organically shape those values to what you want them to be, both as a country, as a firm, as an individual? Like, can we do post-post-training to elicit the right values in models?

Trust in benchmarks feels like it's at an all-time low and maybe is headed lower. What do you— what other— do you have any other ideas of how to communicate the sort of— maybe it's still that they work well enough, but any other ideas around how to communicate the value of certain new models? Yeah, it's a combination of the hack—

the benchmarking, bench hacking, and then also just like inundation with so many that a lot of consumers and enterprise buyers, they sort of just roll their eyes when they see a new one because they're like, I got— I've seen 100 of these. I know that you probably do well on one. But yeah, I would love to hear your thinking on this.

Let's think about who makes benchmarks. It's typically a bunch of research fellows from Stanford or what have you who've pulled together, and they're very smart people. And they're like, OK, this is this weird way in which we're going to trip this model up. I think ArcAGI is a very good example of this weird pocket where the models traditionally don't do well. And they've obviously updated that benchmark. more and more as the labs have kind of caught up. The common saying is like, you know, once you've benchmarked it, you can RL on it. So, you know, releasing these makes the models better. I think the best way to do this is aggregate the economy. And this is where we actually see us having a really big advantage compared to particularly the closed-source labs. Like, I do believe Anthropic and OpenAI are generally good. And like, you know, they don't look at user data if they say they don't. And like, they have this very like bird's-eye view of like what the models are good at and what they're bad at. It's why they go on Twitter and ask for feedback. We have a real advantage in the sense that we have thousands of people using them for different things. And each of those individual companies has tried in some way to build their own evals and benchmarks to test how good a particular model is at the task that they care about. And so rather than— there's no silver bullet here. There is no set of 10 benchmarks that is going to tell you exactly the core and the jagged frontier of whatever model it is you're testing. The best way to do this is to look at how everyone is using them in the economy for real things that people are paying for, not weird niche things like ArcAGI, and aggregating those benchmarks. And I guess that's part of what we're trying to do with the RL environments and pumping them out at scale. But yeah, we get a lot of really good signal from that.

For the record, here at a TBPN, we had Tyler do ArcAGI V3 tasks manually as a human. He technically got paid for it. So, you know, when you score top, he was, he was, he was globally ranked.

Globally ranked.

For like a week. We were very early. I think the day of launch, he was like number 7.

Charlie, what does it mean to feel the whole elephant?

Oh yeah, great one. I've gotten a lot of flak for this and people have interpreted it in quite inappropriate ways, including my CEO, actually. Yeah, but there's this old analogy of like, you know, blind men touching an elephant. And like, they don't really know what it is they're looking at. And one's feeling the trunk, one's feeling the leg, one's feeling the tail. I think it is a little bit like that with things like LLMs and particularly like continual learning.

Mm-hmm.

You know, some people say that continual learning is just like having a god model with, you know, a million tokens of context and it can search whatever it needs to and organize the information that way. Other people really believe that no, every single token it should be updating, you know, its information. We know that that fries the model in different ways. I think Thomas Kuhn wrote a lot about like the structure of scientific revolutions. And I think we are pre-paradigm when it comes to things like continual learning. Like no one can agree on what the definitions of these things are. No one can agree on, you know, apart from the core recipe to produce the base LLM in the first place, like what the right ways to be scaling this is, you know, the ecosystem around it. Like compaction is a great example of this. Like OpenAI and Anthropic have gotten really good at compaction in Claude Code and Codex. At the moment, that's just summarizer models. There's probably really, really cool things that you can do if you train neural compactors, if you have models themselves doing compaction in KB cache space. So I think it's just a lot of work to make this a science because so much of it is inside the closed source labs. And I guess that's the mission of BaseLabs is how can we bring as much of this into the open as possible without a commercial agenda? Even the open source labs have a commercial agenda. They need to produce the best model this quarter. We don't have that pressure. And so I guess we can pick the most interesting scientific problems and work on them for as long as we need to before we feel like we've made traction.

What is your—

Very cool.

P brute force, the probability that the answer to continual learning is just brute force. When I think about what OpenAI and Anthropic do for updating a model, you mapped it out. It's a couple of months of data collection, building RL environments, retraining the model. That takes GPU hours, but if you get a 10x speedup in the amount of time it takes to train the actual model versus— you can build the RL environments 10 times faster, you can generate the data 10 times faster, and you start working at this over maybe it's a decade, is there a world where you could just do exactly what we're doing now, but every second and it feels continual? Because when I talk to— anyone, they're continually learning, but they take a second. They say, oh yeah, okay, I'm updated. I now know that fact.

Yeah, I think there's a few ways to break this down. From the perspective of RSI, like recursive self-improvement, I actually think that continual learning is either unnecessary or it is brute forcible.

Sure.

Like, we're gonna asymptote towards models that have access to the whole OpenAI Anthropic training stack. They're gonna make slight architectural improvements. They're gonna drop pre-training loss. They're gonna be able to pump out RL environments, scale by themselves. themselves. And then that process will speed up and speed up and speed up as we get more and more compute. But for the everyday person and the everyday AI-native startup or enterprise or whatever, continual learning doesn't look like that. At the scale that they're operating at, you can't afford to— OpenAI and Anthropic, they wash out all the noise in the gradients. They take all the useful stuff, and these big pre-trained runs makes it work.

Yeah.

But when you just focus on, like, I have an agent which is a legal associate, and I'm trying to fine-tune it on all these complex relationships and all the things that the firm does, and this implicit behavior that we want it to have. Continual learning breaks down. We don't have an answer to it. We can't SFT. It degrades the model. We can't RL because it doesn't give knowledge acquisition in the right way. There's a lot of work to be done there. So I guess it depends on, again, which part of the elephant you're touching. What definition of continual learning do you care about?

We got to get an elephant in the studio.

We need an elephant to sit on board.

We've got dogs. We've got horses.

Which part of the elephant you're touching. Thank you so much for coming on. This is a fascinating discussion. Congratulations on the progress.

Yeah, great to meet you, Charlie.

Very excited for you to solve this. Come back soon.

This was fun. We'll talk to you soon.

Have a great day. Goodbye.

Up next, we need an elephant. Snowflake.

By tomorrow, we got the CEO with us here live on TVPN. Sridhar Ramaswamy, welcome back to TVPN. How are you doing? Thank you so much for joining us and congratulations on the fantastic success. Break it down for us. What's working? Is there anything that's not working? It seems like everything's going really well. How are you doing? Hi, am I live?

Yes, we're live. We're live.

Welcome to TVPN. Thank you so much for hopping on.

Oh, thank you. We had a pretty amazing quarter. $1.49 billion, 37% year on year. Our AI products, Coco and Cowork, getting really broad adoption. We feel very good overall, but this This is also a time of intense competition. There's the hyperscalers, but also the foundation labs. So we think there's lots of business to be had, but a ton of change to keep up with, both for our product teams and for our go-to-market teams. So yeah, it's a day after earnings are done and it's back to work, pedal to the metal.

How much time are you spending trying to identify like a venture capitalist, the next company that's going to be a major Snowflake customer? Because I feel like the big labs, you identified them, they're obviously huge consumers of data, huge businesses there. But it feels like every time we talk to somebody on the show, there's a new business that's getting venture funding, they're getting users, they're getting revenue. And I can tell that they're just generating a ton of data.

Yeah, they're ramping, ramping revenue way faster than any company.

And they're probably ramping data 2 orders of magnitude faster than they're ramping revenue even. So what— so what is— what does that look like for you and for Snowflake?

Yeah, we spend a lot of time both with our own innovation. What are great new ideas to be had at this moment? Because so much is possible. AI has is industrializing software. It's much easier to create it than it was before. So I spent a lot of time with the product and engineering teams, back-to-basics innovation. We also meet a lot of startups and companies, both established ones, talking to them about the opportunity that is possible with AI, talking to them about what we have done with our own sales teams, but also brand new startups. I met a startup, I think it was day before yesterday, barely 3 months out, but They have dozens of customers all doing reinforcement learning on their platform. We were talking about how we could better partner together. It's a good amount of time spent on both sides of the cycle, but what I think is unique and cool is how quickly ideas go from it's just an idea to prototype to it's a product feature to getting adoption with a lot of customers.

Yeah. With all of the growth, I imagine that there's new challenges. Where are you going to be focused on unbottlenecking the organization? Are you going to be hiring, putting more AI, as you mentioned, in the hands of even an even broader swath of the organization, even though I imagine everyone has access to some level of token budget? But what does— what does the growth side of the business— what are you trying to unlock in the next quarter?

Quarter or the next year? For the majority of the company, the biggest mind shift, mindset shift that they need to go through is scale is not just about people any longer. What AI makes possible is stuff that needed individuals, people to go and do, can largely be automated. It's much more about the judgment. What's the work that you're trying to get done? What do you— how do you want that, you know, how do you want that done? And then setting up a framework by which things can get stamped out. But that's a pretty tough change for an org or a set of people, even you and me, that have traditionally looked at organization size as an important barometer of growth. And that's something that we are stressing with the entirety of the Snowflake team.

Yeah.

expect some functions to grow, but these are typically the ones that do things like interact with the external world.

Sure.

So account executives, I can see that growing for customers that are spending. But for a lot of other functions, including in engineering, there is so much leverage that you can get by being at the forefront of what is possible with agentic AI. We continue to hire folks, especially young folks who know of no other world, because they bring a perspective. that is fresh and they know how to go from 0 to 100 straight off the bat. But a lot more focus is on how do we become effective as a company. And a lot of it is also about how do we set up new efforts and give them space to experiment so that they can find that magic in a bottle.

Yeah.

AI is not going to help us create great new products just like that. There's an amount of experimentation to figure it out. And giving space to small teams to execute is among the biggest challenges that we need to make sure that we solve. Coco, for example, it's a breakout success, but for much of its existence, it never had more than 5 people. And even now, the team that works on Coco is tiny. But that's the kind of impact that is possible today.

Are internal meetings Is meetings underrated? Would you expect that in a world where AI can do more of the work but lacks some of the context and taste and decision-making and deciding what to do, if you woke up and you said, wow, we're, you know, as a company, Snowflake is having 10%, 20% more internal meetings between various members of the organization, would that sit well with you? Because there's been a long history of people, oh, internal meetings are such a waste of time. It can be very cumbersome if everyone's calendar's just full. That's just layers and layers of management. But is there some world where that's actually the correct way to be running a business of this scale in this era?

I think bringing people along is an important function of every company. I wish I could tell you that we've solved this information exchange problem perfectly. There are a number of things that we are working on. We have an internal Enterprise Brain project, like many other companies do, in order to capture all of the relevant knowledge about the different things that are happening within the company so that the right piece of information gets to the right person.

Mm-hmm.

So I do think of projects like that as greatly enhancing internal communication. Remember, in a regular company, people go to meetings so that they don't feel left out, which is something that I really discourage people from doing. My attitude is like, if I have nothing to contribute to a meeting, I should read the notes.

Yeah.

I really should not have to go to that meeting. But I would say it's a work in progress to make sure that we do a good job of Transmitting internal information. It also has more profound implications. I think things like how many reports a manager should have needs to be dramatically different in the age of AI. Of course, making sure that people feel happy and motivated about their work, that's always going to be the role of a good manager. But, you know, information about what exactly did you get done last week, that's the kind of stuff that AI can facilitate a lot. It's a work in progress. I truly hope, you know, I've not measured it recently. You bring up a good point. We should go look at it. I hope we are not spending more time with internal meetings, but we should also not pretend that communication is a solved problem even in the world of AI.

Yeah. How are you thinking about your own internal AI spend and budgeting heading into next year? I think a lot of You know, companies planned, spent a bunch of time planning next year, and then there was a capability jump and they blew through their budgets more quickly. I think as a leader, you have to assume there's gonna be capabilities jumps, which may mean that you spent more, but at the same time, there's a lot of drive for efficiency, cheaper models, open source, et cetera. But how are you planning around token spend, you know, looking forward?

My take overall is that the money that we are spending on AI tokens is well worth the cost.

Mm-hmm.

We continuously optimize. Absolutely. We don't tell people to token max or do dumb things like that. It is about driving real results, impact, as, as it were. And when our cost does go up, we have a good team. That focuses on optimization, everything from what's the default model to can you create task graphs where the simpler aspects of solving a problem are handled by models that are not quite as expensive. And my take is that we gain a lot by having people embrace the technology and feel like it can make a real difference to their job. I feel pretty good about optimizing. We also practice what we preach where we have things like the AI gateway that's meant to make model access more efficient. We are absolutely experimenting with open weight models because they can be a good vector for lowering costs because we also control the harness, which is CoCo. A lot of our internal teams use— there's a lot of instrumentation that we have for what exactly are people doing, how can we come up with better ways for doing the same thing. Our support and our SRE teams, the folks that keep Snowflake up, they spend a lot of money on tokens. But on the other hand, if they are running through tens of thousands of alerts that are coming in every day, makes sense to go there and optimize because that's very leveraged work that can benefit everybody. So, you know, yes, we are planning for it, but AI token costs It's not the top thing on my mind. Creating great products, getting our customers to adopt them, having the framework by which these things become more self-correcting. That's kind of how I think about it.

Based on all your experience in the enterprise, how do you think the model routing landscape will evolve? This seems like something that a lot of big companies are excited about, you know, getting a slice of that market. You saw Stripe's deal with OpenRouter focusing more on developers? Ramp, our partner, has a product. Everybody wants to be in the token flow, but how do you think that it'll evolve?

That's not the highest value creation point for Snowflake as a company. We have products like CoWork that can literally get deployed to every employee in a company. So we spend a lot of time thinking about What does it mean for an enterprise sales team to be AI-built, to be operating at the edge of what is possible with AI? And we are seeing deployments of Cowork go to thousands of users within, within companies. It's operating at a much higher level than simply model routing. I think it's a good infrastructure capability, but I focus a lot more on things like How do we get every data engineer within a company to be using Coco, our builder product? Or how do we get large fractions of employees within companies to be using Cowork? What are high-leverage, high-value projects that we could be doing for customers? Because our customers have always trusted Snowflake with all of their most important data, and focusing on how we can drive real business outcomes is, is where a lot of our energy is at. Absolutely. The Gateway is a good product and it's not just model routing. We are also offering things like access to tools, MCP tools as it were, that provide governed access to lots of different applications within an enterprise. We are also actively experimenting with is there a security solution around agent trajectories to make sure that people are not misusing models? So there's a slew of these things that are there at the infrastructure layer. But I think our bigger prize is in delivering value for our customers.

Last question. There's a lot of M&A news. Obviously, it's a very exciting time. You spent 15 years at Google, very acquisitive company. How do you think about M&A? What makes for a successful acquisition? How is it unique at Snowflake when you consider it, when you don't?

Our strength is as a data platform. People trust us with their most important data. People trust us with helping them get insights and drive actions from their most important data. My primary lens is, will a company being part of Snowflake accelerate that mission? We are thrilled that we bought Natoma because MCP is increasingly really, really important for Snowflake and all of our customers because it provides the real-time context of everything that's happening in your and my life, whether it's Slack or email, right in the harness. And that was a great acquisition to make. That's the kind of lens that we bring, which is how does something being part of Snowflake accelerate both the company that we buy, but also the larger mission of Snowflake as the AI and data platform platform for every enterprise that there is.

Thank you so much for coming.

Yeah, great update. Great quarter.

Congratulations on the quarter. Yeah. Congrats to the team. Can't wait to talk to you.

Thanks for having me.

Have a great day. Bye. Didn't get a chance to hit the gong, but very gong-worthy quarter.

Huge update in artificial intelligence world. Someone has created a product that speeds up Lex Fridman when he asks questions No way. LexSpeedman.com. Mario Dion says, I wanted to watch the Lex Fridman episode with DHH, but Lex talks way too slow for me, so I built Lex Speedman. It speeds up Lex to 2x speed and keeps the guest at 1x, cutting up to 20% of the episode. Wait, so he— so Lex will talk for 40% of his episodes? That's not possible. I feel like Lex is like, he asks one question, he really lets the person talk. That's kind of a benefit of that.

Yeah, he does.

He talks for a while?

Kind of drifts off.

I wonder, we should pull some stats on like, on like who, who talks more, because sometimes we have good back and forth, we're about— we're all talking with the guest. Sometimes it's just us, short question, they go on a rant for a couple minutes, you know.

Yeah, sometimes we get some very angry, very angry comments About what?

Cutting each other off?

Would you please let the guest talk?

Oh, they say that? Interesting. I remember early on they were saying, you guys gotta stop cutting each other off. And we— do we ever— was I cutting you off, or were you cutting me off? I feel like this is just the way we talk, and like, it never—

Our real fans went to the rescue.

Yeah, we never really got to the bottom of it. I think it was me cutting you off sometimes, but especially with the ads. I love cutting You off with an ad. That's the best. That's the best. But also, no dead air, you know?

I love cutting you off with a flashbang.

No, don't do it. Don't do it. Flashbang. We're going to the stars. Astra GPT-6 is out. Go check it out. There's a crazy video by Matt Schumer. Uh, he one-shotted a game in Unreal Engine. Now Unreal Engine has MCP, so you can actually build the game of your dreams basically just by chatting with the model.

And, uh, Mike over at Arc Is moving the goalposts. Moving the goalposts.

Moving the goalposts. They're moving.

Thank you.

Let's do it.

Thank you for not waiting to move the goalposts either. You know, he said Astra is, is the new state of the art on Arc AGI-3. It's a qualitatively large leap towards AGI, and the pace of progress is frankly surprising. That said, we lack evidence to call this AGI yet. When Mike says it's AGI. While we are still studying the human capability gaps, we believe open-ended invention is unsolved, and this will form the new basis for ARC AGI-4. So now it's like, you gotta invent new physics, new science. You gotta go to that. You actually— Astro has to actually go to the stars.

Yeah, he really said, what have you done for me lately? I love it, I love it. It's AGI when Mike says it's AGI. It's a good time. Anyway, very interesting. You can dig into his post. He gives a lot more context there about ARKGIV3 benchmarks, Astra. You can go check it all out.

And of course, we'll be discussing it tomorrow.

We have a bunch of special guests. So have a great day. We'll see you tomorrow. Leave us 5 stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. And we will see you tomorrow.

Thank you.