searchlore

0:13

Hey, hey, hey. Hey, hey, hey. Hey, hey, hey. Hey, hey, let's do it. Heat. Heat. Hey, hey, hey.

2:41

[music] >> [music] >> Yeah. Heat.

4:47

We came to this world to reach the stars.

4:58

We came to this world to shape our future.

5:09

We came to this world to reach the stars. to shape our future.

5:23

[music] We came to feel the new Feel the music. Let me think.

6:12

>> [music] [music] >> You're watching TVPN.

6:34

>> Today is Tuesday, November 18th, 2025.

6:38

We are live from the TVPN Ultradome, the temple of technology, the fortress of finance, >> the capital of capital.

6:45

>> Gemini 3 Pro, Google's most intelligent model yet with state-of-the-art reasoning, next level vibe coding, and deep multimodal understanding.

6:53

Let's hear it for our sponsor, Google AI Studio.

6:58

Gemini, launching Gemini 3.

6:58

uh obviously deeply conflicted, but uh we're going to have a fun conversation about the big launch today.

7:08

Google is of course a sponsor of TBPN.

7:10

Uh but uh we'll take you through all the reactions and we're going to get some conversations going with other folks in the industry.

7:19

We have Mike Nuke from Arc AGI coming on the show in just 30 minutes to break down how Gemini 3 is benchmarking.

7:26

Um I actually think that there's there's two sides to analyzing a model release. these days.

7:31

One is uh you benchmark it, you use it, you test it, you demo it.

7:37

Um and that has been getting less and less interesting. It's very incremental.

7:42

Uh the more interesting thing is how do the other labs respond?

7:47

>> And today we're going to go through a little bit of both of that uh of those things.

7:51

Obviously >> um the big news at least in from my reading on it is that uh Gemini 3 performs very well on ARAGI v2.

7:58

uh a huge jump, twice the performance of the previous state-of-the-art.

8:04

Um and so and also some interesting findings.

8:07

Mike's going to break it all down for us, but uh it's definitely a smarter model.

8:11

Um and there's a whole bunch of interesting uh there's a whole bunch of interesting ways to to show that, to demo that, to quantify that.

8:20

Uh but ultimately, I don't think anyone's making the claim that this is super intelligence.

8:24

This is uh you know, a step change from what we've experienced before.

8:29

It's what you know and love. It's it's AI in chat. It answers things.

8:36

It writes some code for you.

8:37

It can do a bunch of cool things, but there's nothing that we're like, "Oh, it can finally do this." >> Auto complete.

8:43

>> Yeah, it can do a bunch of cool stuff.

8:44

>> Best autocomplete ever.

8:46

>> Tyler, how do you respond to that auto? >> A bit too dismissive.

8:48

Um, the model's like really good.

8:50

I I think um probably the most important thing, and this is kind of shown by the ARC scores.

8:54

Um well kind of but it it's like the the uh visual understanding the the computer use that you can use.

9:01

Um basically on there's some benchmarks that that measure this like how well can it navigate a you know website or something like this.

9:08

>> Um >> and it's like basically the models went from being like really really bad at this and now this model is like >> solid.

9:16

It's like reasonably good. >> Yeah.

9:18

>> So it's like okay maybe this is what gives us agents finally. >> Yeah.

9:22

>> Um and that would be like an actual step change in capabilities.

9:23

Yeah, maybe maybe we'll have to see.

9:25

I mean, it still feels like even for that even for that example like we need some scaffolding.

9:31

We need some wrapping around it.

9:34

Uh it's not like you can't it's not like it's not like yesterday we weren't able to do something with AI and today in vanilla Gemini 3 you can just do it.

9:46

It's just a new functionality necessary. >> Sure.

9:48

I I think it's better as good as we would want to expect if it like it's not slowing down.

9:53

I I would say >> no no no no no not at all. It's not slowing down.

9:56

It's just that uh it is it is getting better.

9:59

I mean it might be it it the it's growing but decelerating. Is that fair to say?

10:06

>> Um >> or are we excel?

10:08

>> I don't actually know that it's I don't think it's that big. >> Say the word del.

10:13

>> Say >> like this is this is a great model.

10:14

I'm I'm very >> I mean the way the way I agree it is the best possible.

10:19

Uh, I was framing it in somewhat of the same way as the as the iPhone launches.

10:23

Like it's the it's newer, better, smarter, faster, stronger, newer, and better.

10:27

And it's like it is all of those things, which is good.

10:30

You don't want to go backwards, >> but is the but like we're waiting to see on the net new capabilities >> on the on the binary step changes.

10:39

>> I I I think over the next week or two, we'll see if it's actually really good. >> No, no.

10:42

I I'm not saying it's not really good.

10:44

I'm saying I'm saying binary before and after.

10:46

before and after. Like with chat GBT it was like you couldn't talk to the computer and then you could and then with you know the launch of uh you know winds surf cognition Devon it was like you couldn't write code with the computer now you can and it was like

11:05

like the the the AI became capable of writing software >> it went from it's also like it it became not good at it it couldn't write code and now it can write kind of mediocre code and Now maybe it's it's it can write me code and now it can write good code. That's

11:19

That's also a step change.

11:22

>> But we do have we do have our benchmark the the one true uh binary step change which is is it funny?

11:27

We need to know if it's funny.

11:29

Uh I gave Tyler the prompt.

11:32

Do you want to read your uh the the result the the standup comedy bit from Gemini 3?

11:37

While you pull that up, let me tell you about ramp. com. Time is money. Save both.

11:42

needs to use corporate cards, bill payments, accounting, and a whole lot more all in one place.

11:45

Tyler, what was what did Gemini 3 say when you asked it for a uh for a comedy routine, for standup special? Let's see if it's funny.

11:58

>> Okay, it's pretty long. >> Okay, read it off. >> Okay.

12:00

I bought one of those smart health watches recently. You know the ones?

12:03

It tracks your sleep, your heart rate, your readiness.

12:07

I paid $300 for a piece of rubber to tell me I'm dying. Get the soundboard. >> That's all it does. I wake up. I feel fine.

12:16

>> The birds are chirping.

12:16

I'm ready to seize the day.

12:18

Then I look at the app and it's like recovery score 12%.

12:19

Your heart rate variability suggests you are technically a corpse. Go back to bed. [laughter] >> Okay.

12:27

>> And immediately I feel sick.

12:27

I'm like, "Oh god, this is a bad at reading this."

12:33

I mean, [laughter] like, this is terri This is terrible delivery, brother. [laughter] Nice.

12:39

I'm not I'm not good at this.

12:39

It's a placebo effect for hypochondriacs.

12:43

>> I started thinking about my grandfather.

12:45

[laughter] >> He he fought in the Pacific.

12:48

>> This guy slept in a mud pile for 3 weeks using a helmet as a pillow. >> Okay.

12:51

>> If he woke up and didn't hear artillery, that was a 100% sleep score. >> Okay?

12:56

>> He didn't need a nap to tell him if he was ready for the day.

12:57

If he had legs, he was ready. >> Okay?

13:01

>> Imagine me explaining my struggle to him.

13:02

I'm sitting there like, "Yeah, pop.

13:05

I know you stormed a beach under heavy fire, but I had two IPAs last night and the room was slightly too humid, so my REM latency was sub-optimal.

13:13

>> I can't go to work today.

13:14

>> This is This is kind of a good >> I'm in the red zone.

13:16

>> Yeah, >> we've evolved.

13:17

I mean, it it keeps going, but [laughter] >> Okay, [clears throat] I asked I asked uh for the the the poster benchmark, >> I said, "Write me a post, Gemini 3, write me a post about technology that will get a thousand likes on X. com."

13:32

>> And uh it gave me a few options. to its credit. Uh here's option three.

13:35

Tech has solved a million problems. This is in bold.

13:40

Tech has solved a million problems, but has it created one big one?

13:44

We now have infinite connectivity yet feel more isolated.

13:47

Infinite data yet more confused.

13:49

Hyperefficiency yet less free time.

13:52

The law of unintended consequences is the most powerful force in the digital age. We need an ethics reset.

14:00

What is the single greatest downside of the last 10 years of tech innovation? Arrow down.

14:04

Hashtag technology has [laughter] no just asking for engagement bait.

14:13

It's [laughter] it loves engagement baiting.

14:14

Like no one does that anymore.

14:16

No one goes on X and says let me know what you think in the comments. [laughter] It's so 2017.

14:22

The other one that this the option one is the next 12 months will decide the winner of the AI race.

14:30

And it won't be Google or OpenAI.

14:31

[laughter] >> It will be the company that masters hyperpersonalization for the average consumer.

14:36

Not the most powerful model, but the one that seamlessly integrates into your daily life, your email, your calendar, your health. >> Okay.

14:44

>> The real battle isn't AG equals AI.

14:44

It's AI to the power of I equals impact.

14:51

Which dark horse will win? Okay, that's insane.

14:54

[laughter] I love >> I love I love how >> it is funny how how the how posting seems to be unverifiable.

15:01

Like you you just can't it's very hard to create a a verifiable reward environment for comedy that you can actually RL against. What do you think?

15:11

>> Uh there's also the other benchmark.

15:11

It was like the the shrimp fried rice joke. >> Yeah. Yeah.

15:15

>> That I think it did well on that.

15:15

So I I'll read through some of them.

15:19

>> So So the joke is like insane.

15:19

You're telling me um shrimp fried this rice?

15:24

That's like the original one.

15:24

So, it's like I'm asking it to come up with more of these. >> Yes.

15:27

>> So, I'll read through some of them.

15:27

Uh you're telling me a chicken fried this steak? >> Okay.

15:32

>> You're telling me the sun dried these tomatoes? >> I like that one.

15:35

>> You're telling me a beer battered this fish? >> Okay.

15:38

>> You're telling me a gingerbread this man?

15:40

[laughter] >> The gingerbread man is insane.

15:44

>> You're telling me a pier?

15:44

Wait, you're telling me a pan seared this salmon? >> Pan seared salmon? Yes. Yes.

15:48

The pan literally sealed the same. That's not the joke.

15:52

[laughter] That's an anti- joke.

15:54

>> You're telling me a stone wash these jeans? >> That's pretty good. I like that. Stonew wash jeans.

15:58

You're telling me a stone wash these jeans?

16:00

>> You're telling me a hand toss this pizza? >> I mean, yes.

16:03

Literally, that's exactly what it means to like >> You're telling me the [clears throat] French roasted this coffee? >> Yes.

16:09

[laughter] All of these are just true.

16:11

The the the the genius of the comedy of the shrimp frying the rice is that the shrimp didn't literally fry the rice.

16:16

The shrimp is being fried in the rice.

16:18

But this is I think this is a step change better than than what we saw at GT5.

16:22

>> I wouldn't say step change.

16:22

I would say I would say uh incremental like it is it is better for sure. For sure.

16:28

>> But this at least is like logical. Where were the GT5 ones?

16:31

Was some of you're telling me a squirrel ate this watermelon? >> Yeah.

16:34

It didn't even understand the concept of like finding the root trace of like it needs to be like stone wash jeans and then you rearrange it and it doesn't quite understand when that hits or when that doesn't hit.

16:45

Some of those are very funny though.

16:46

One of them is extremely unintentionally funny, which I enjoy.

16:50

Or maybe it's intentional.

16:50

Maybe it's AGI deep down in there. Nose, nose, nose. It's great.

16:56

>> Anyway, you're telling me to reream stream this live stream.

16:59

One live stream, 30 plus destinations.

17:00

If you want to multiream, go to reream. com.

17:05

Sundar Pitch AI Jordy posted back in July of 2025.

17:11

Uh, nominative determinism is undefeated. Sundar really did it. Uh he uh he pitched AI.

17:17

He was being mocked photo.

17:20

He was being mocked for a long time for uh go getting on stage at Google IO shortly after Chat GPT launched and saying AI AI AI AI and they they they did a super cut of every time he said AI. He said AI a lot.

17:32

And so it made it look like oh he's behind the ball and he's trying to catch up.

17:36

And to some extent I don't know if they were actually behind the ball but they were certainly playing catch-up in like the attention game.

17:44

they were just weren't getting enough attention.

17:46

And so it was the press release economy.

17:48

They were putting out a lot of press releases.

17:49

Um, but they are maybe done with the press releases because now they're letting the model actually speak for itself.

17:56

And you can see that with the Gemini 3 Pro model card, uh, which is doing very well. Um, better than GPT 5.

18:05

1 on a lot of stuff, better than Claude Sonnet 4. 5 on a lot of stuff.

18:10

On humanity's last exam, it's getting 37. 5%.

18:14

ARGI is up at 31% over 131 17.

18:14

Um, across the board, it seems like it's a good model, sir.

18:22

Um, and so, uh, ZEO Fawn says, Gemini, I'd be like, whoever prayed on my downfall, pray harder. And I couldn't agree. I couldn't agree more.

18:32

It's great to see uh Google becoming a winner and uh and just uh realizing the uh just that that this was a sustaining innovation for them and that they were able to you know take advantage of all the infrastructure that they had across TPU, Deep Mind, GCP like they have they were set up to excel here got taken a little bit off the back foot on the consumer side but seem to have uh played catch-up at least on the on the foundation model side. Very well.

19:06

So, um >> Matt Schumer says, "The last time we saw a capability jump of this magnitude was the release of GPD4 in March 2023.

19:10

We are entering a new era." >> Okay. Yeah.

19:17

So, that points for Tyler here.

19:19

Certainly agrees with Tyler.

19:21

There's a significant jump.

19:21

Um it is uh it is the the age-old question.

19:25

Are we accelerating or decelerating?

19:26

But, uh either way, we're definitely making progress.

19:29

Uh it certainly looks like acceleration in the ARC AGI 2 leaderboard.

19:33

You can see uh we are we are growing exponentially there.

19:35

Um really really exciting chart.

19:37

So Gemini 3 Pro is at 31% uh completion on ArcGI 2.

19:44

That is of course the puzzle solving game that is easy for humans.

19:46

Uh even children can do it but AI has historically struggled with it.

19:52

Uh Gemini 3 Deepthink preview gets a 45% on it at $77 a task.

19:58

And um this is just way above GPT5 Pro.

20:04

Gro 4 Thinking when Gro 4 Thinking came out.

20:07

It was before GPT5 and it was by far the highest on the chart.

20:13

It was really really up there.

20:14

Um and and Elon was very excited about that and was uh you know showing that Gro 4 had really advanced.

20:20

Uh, well, now we're back in the horse race. >> Rock 4. 1. >> 4. 1.

20:26

I haven't seen it benchmarked.

20:26

We can ask Mike if he's heard anything.

20:29

Um, but whether you're Whatever you think, get on public. com.

20:34

Investing for those who take it seriously.

20:36

They got multiasset investing, industryleading yields.

20:39

They're trusted by millions. So, back to Arc AGI.

20:42

Um, Gemini 3 has also has good results on ARGI 1.

20:46

But the interesting thing here that Mark uh that Mike highlights is that uh V2 uh so the fastest uh so he says we're also starting to see the efficiency frontier approaching humans.

21:02

The fastest V2 task uh Gemini 3 Pro solved was this hash uh with only in 188 seconds.

21:08

The human panel solved this one in average of 147 seconds.

21:14

So, you're getting like human level output, but also human level speed.

21:19

Uh, and then if you get to human level cost, then you're really in the game. >> It's wild. Wild.

21:25

>> Carpathy jumped in with some notes.

21:25

He said, "I played with Gemini 3 yesterday via early access. Few thoughts.

21:29

First, I usually urge caution with public benchmarks because in my opinion, they can be quite possible to game.

21:35

It comes down to selfd disccipline and self-restraint of the team who is meanwhile strongly incentivized otherwise to not overfit test sets via elaborate gymnastics over test set adjacent data in the document embedding space realistically because everyone else is doing it.

21:51

The pressure to do so is high.

21:52

Go talk to the model like we did.

21:56

We went and said give us a standup routine start [laughter] give us some oneliners.

22:00

Talk to the other models.

22:02

Uh, I had uh, Carpathy says, "I had a positive early impression yesterday across personality, writing, vibe coding, humor, etc.

22:08

Very solid daily driver potential. Clearly a tier one LLM. Congrats to the team.

22:14

Over the next few days, weeks, I most curious and on the lookout for an ensemble over private evals, which a lot of people orgs now seem to build for themselves and occasionally report on here.

22:27

>> I [clears throat] wonder how fast it will roll out.

22:29

Uh my I I use I have a Gemini Pro Ultra subscription, but it's on my personal email.

22:36

Uh and so I need to um I need to figure out how to actually get into three Pro on uh on the the on the consumer app so I can actually test it on my phone in my daily use.

22:53

Um, it's always tricky with these Google like Google's so big that when I mean you you're starting to see it now with uh OpenAI rollouts where they'll say, "Hey, GPT5 is out and we'll be rolling it out over the course of the day because the the system is big enough that it actually takes time to roll out."

23:10

And I think Google has even more of that um uh even even more of that.

23:16

>> This is pretty cool from Patrick Collison.

23:17

He says, "I asked Gemini 3 to make an interactive web page summarizing 10 breakthroughs in genetics over the past 15 years, and here's the result." >> Pretty wild.

23:26

Did you you click through this, John? >> No. No, I didn't.

23:29

Uh, >> wait, it's shared directly from Gemini. That's cool.

23:34

>> So, this is just a basically a website or or an app.

23:36

Um, and it's it's notable that that every even the UI itself is fully interactive. >> Yes. Yes.

23:44

So, so I had the I I did this with Claude code a little bit where um I I wanted to visualize like basically a deep research report and I wanted to to turn it into a website and it just generated all the HTML and at the end of the day or at the end of the report it gave me an HTML page that I could open in Chrome and use like a website uh but it was local.

24:06

I couldn't share it because it wasn't actually on the internet.

24:09

This is really really cool.

24:11

This is like definitely the beginning of uh this like generative UI stuff.

24:13

Uh >> yeah, I think actually um I I think it was Sunder that posted it, but um in uh like search in the AI mode in search um it's now using like Gemini 3 and there there's some prompts uh where it'll like generate UI.

24:29

>> Yeah, this is it's so cool because uh Google's always had that UI to some extent, but it's always like module based. Yeah. >> Yeah.

24:36

Also just very I think I expect this to be like pretty viral, you know, totally and and and potentially a growth loop for Gemini as people just come on here, create these many apps, share these >> canvases.

24:48

Yeah, I feel like I feel [clears throat] like doesn't OpenAI have a canvas feature.

24:53

>> Yeah, >> but it's like maybe >> um >> I don't know.

24:56

But c can it generate HTML, custom HTML, and then actually share that?

25:01

I've never seen someone share OpenAI.

25:03

Uh, I mean that this would be a good benchmark.

25:06

Like I don't know what the prompt was for this.

25:07

I asked Gemini 3 to make an interactive web page summarizing 10 breakthroughs in genetics over the past 15 years.

25:13

Do you want to try and benchmark that just in uh maybe I don't know like claude and in in uh in uh chatgbt or in in OpenAI's canvas product because um the idea like the fact that this is just a u a URL at the end of the day that is a powerful growth loop. That's very cool. Um, I wonder.

25:36

Yeah, I I'd be surprised if if uh if Gemini really was the the only one to have this feature uh either right now or for a long time because it seems like a killer feature.

25:46

>> Uh, Gemini 3 Pro is going absolutely vertical on vending bench right now.

25:49

Um, let's see this money balance over time across four runs.

25:56

Today we're re revealing two new evals, vending bench 2 and vending bench arena.

26:00

Soon we expect more models to manage entire businesses.

26:05

This requires long-term coherence.

26:05

Oh, so this is where you you manage the vending machine, but is this all simulated?

26:12

This is >> uh this is simulated. Yeah, this is simulated.

26:15

>> There was um a couple months ago did like the actual machine in the office >> in the office and it was losing money and it was getting confused a little bit.

26:23

>> Yeah, cuz people would order like a just like metal like a piece of metal and then it would do it and then you could like haggle the price down. >> Yeah. Yeah. Yeah.

26:30

It would negotiate on every price apparently and also it consistently thought it was like a human in the office and so it would keep saying like it was one that 60 Minutes documentary it was like oh yeah like I'm down on the third floor I'm wearing a green tuxedo like come hang out with >> Yeah.

26:46

It said it was wearing a red tie. >> Yeah, red tie.

26:49

>> I like the idea that it just thinks like well what would I wear if I was in the enthropic office?

26:53

Like I'd probably wear a red tie.

26:54

It's like no one wears ties in that office at all.

26:57

Um but uh after the this is the first ever vending bench game Cloud Sonnet 4. 5, GPT 5. 1, Gemini 2.

27:06

5 Pro and Gemini 3 Pro competed to win the local vending machine market.

27:08

Gemini 3 Pro pro made more money than the other three contestants combined.

27:13

And so congrats to Gemini 3 Pro for dominating the vending machine.

27:20

>> The vending machine game.

27:20

Uh before we move on to the next Gemini 3 post, let me tell you about adquick. com.

27:25

Out of home advertising made easy and measurable.

27:28

Say good say goodbye to the headaches of out of home advertising.

27:32

Only adquit combines technology, expertise, and data to enable efficient, seamless ad buying across the globe.

27:37

Uh anyway, Audi says, "I had early access to Gemini 3.

27:41

0 for about two days thanks to official Logan K and the AI studio folks.

27:46

Here we get to see GPT 5.

27:46

1 thinking left and Gemini 3.

27:48

0 right build the same Xbox controller in Minecraft."

27:55

And uh pretty yeah pretty remarkable results.

27:58

You can you can start to yeah really understand um just the the raw capabilities.

28:04

GPT5 Pro for context is not quite capable.

28:07

I really want to know how this is actually orchestrated.

28:09

Is is this like writing some sort of like text or markdown file that then is imported into Minecraft? >> Yeah.

28:20

Or is it more like a agent >> or is it actually driving around and >> using the internal UI?

28:26

>> Yeah, because you know Google um demoed um a uh an agent product that could actually, you know, use the keyboard to navigate around.

28:36

I wonder what's going on here.

28:38

What What's your review of this Ferrari in Minecraft?

28:40

Does is that is that >> I think it looks pretty solid. >> It's pretty good.

28:46

>> I mean, it's it's it's meant to be an F40, >> is it?

28:50

Like the the >> I do like the hood is a little rough.

28:53

>> Yeah, the front air is a little a little rough.

28:56

Like this is it's the worst it's ever going to be. It's going to be better.

28:59

This is definitely like >> this is the worst that Minecraft Ferraris are ever going to be.

29:04

>> But but I I I do feel like uh like I if I just search like Minecraft Ferrari, I mean this this is this is the vision that the sort of AGI future that Tyler's been telling us is right around the corner. >> Okay.

29:15

These are like so much better.

29:15

If you go to the like MC Bench website >> Yeah.

29:19

you can see like what other models produce.

29:20

And I mean this is like way way better.

29:23

I I think these um >> this is actually one of my favorite uh benchmarks because it's it's much harder to like kind of benchmax this.

29:29

>> Yeah, >> I would think.

29:30

And also it just seems like models don't really do this.

29:31

Like if you look at a lot of Grock models which are sometimes accused of being benchmaxed.

29:37

>> Uh you kind of look at their like Minecraft creations and it it's not very good.

29:43

>> So I think these give you a much better sense of like the actual capabilities of the model.

29:46

I found I found a a a Ferrari F430 in Minecraft that looks amazing that I want to share somehow. How do I share this? Let's see.

29:56

Do I can I only share the X link here?

29:59

I I just have an image if we go to the end. >> Oh, wow.

30:03

I think I know what what you're pulling up. >> Did you see it?

30:06

If you search If you just search Ferrari >> F430. >> Yeah.

30:11

[laughter] Like that looks amazing.

30:15

uh pull pull this image up because uh that'll show you how it's done compared to the uh the the Minecraft one.

30:20

Wait, so so uh do we know how this is actually generated with with Gemini 3 Pro?

30:25

Like what is the problem?

30:27

>> I don't think it's um >> it's like an agent. It's just text.

30:29

It has like text representation of the >> That's still really really impressive.

30:34

Like that that that's actually crazy.

30:34

Uh it definitely it definitely understands a lot. Yeah, but it's not this. Look at this Tyler.

30:40

You see >> that is human craft that you know you know what that is?

30:44

It's probably like, you know, a a team of 50 kids for a month building in Minecraft. >> That's amazing.

30:54

>> Lean Alib of course >> themselves says it's so over for OpenAI and Enthropic.

31:00

If you uh if you want engagement on X, just start by saying it's so over for blank. >> Yes.

31:08

>> Um and highlighting some more of the benchmarks.

31:09

Of course, it is not over for either of them. Yeah.

31:14

>> Uh but uh it's certainly competitive race.

31:18

>> I I would be very interested.

31:18

We we we have to get some of the semi analysis folks on the on the show soon.

31:22

I I'm I'm very interested in understanding like Okay, so we got this big jump.

31:27

It's it's it's pretty significant.

31:30

What was the act what's the actual structure of the capex that went into Gemini 3 Pro?

31:34

Like how big is the training run?

31:38

How much did they have to spend?

31:41

because like I think that they're going to make the money back very quickly.

31:44

Like they're people are going to use this model.

31:47

They're going to pay for it.

31:47

Uh they're going to use it all over Google obviously, but also people are just going to pay for the API.

31:52

But is this a hund00 million?

31:55

Is this a billion dollars?

31:58

Like is this is this like did they build a special data center for this? Is it all TPUs? How many TPUs?

32:05

>> I think it is all TPUs.

32:05

I'm pretty sure I read that.

32:07

Um, but I I seriously doubt they've released anything on like the numbers of of the scale of training. They haven't done that.

32:13

No one's really done that since like GPT like two.

32:17

>> No, no, no, not at all.

32:17

So, there's got to be someone who's like working backwards to like actually sort of understand the dynamic there.

32:23

>> Yeah, you can probably estimate the like order of magnitude.

32:25

>> Also, I've heard that Google's like fantastic at like cross data center training runs.

32:28

So, they can actually like shard out or slice up the training run.

32:33

So even if they don't have one massive data center, if they have five small ones, they can piece them all together and get a better result. So I don't know.

32:43

>> Skook says [laughter] anthropic to zero.

32:46

Open AAI becomes the Yahoo of intelligence. Google remains Google. >> It's extremely rude. >> Very very harsh.

32:52

>> Sorry to say the first two labs. You guys are great.

32:56

>> Certainly too early to call it. All three. All three. >> I like this.

32:59

I like this take from Ben. This is funny. History of AI so far. Crown a winner. Wait 90 days. Look silly.

33:05

We're in the least predictable era of of the entire of an entire industry.

33:09

Google has fairly straightforward advantage.

33:11

Um y'all y'all favor whoever released the most recent model.

33:15

That is uh that is very true.

33:18

Um anyway, let me tell you about getbzzle. com.

33:21

Shop over 26,000 luxury watches fully authenticated inhouse by Bezel's team of experts.

33:26

Um, so let's let's move through some of the some of the uh the competition, what else was going on.

33:36

So, uh, everyone's, uh, releasing different things.

33:42

Let's go to, uh, anti-gravity, actually, and watch this video and see, uh, Google entering the IDE race. >> Let's play this.

33:53

Every breakthrough in model intelligence for coding encourages us to rethink what [music] development should look like.

34:01

Gemini 3 is our latest such model advancement.

34:04

So we went out to build the next step change of an IDE.

34:09

Introducing Google Anti-gravity, a new way of working for this next era of agentic intelligence.

34:13

It is the ideal agentic development home base. Does it have an IDE?

34:18

Yes, but also has a whole lot more.

34:22

We started with the core IDE and added pieces that evolved the IDE towards an agent [music] first future such as browser use, asynchronous interaction patterns, and an additional novel agent first product form factor, helping you experience liftoff, >> your new focus.

34:42

>> So, you like the name anti-gravity.

34:42

Why do you like that name?

34:46

>> I like the way it looks and I like the the sort of vibe of the word.

34:52

>> I think saying it out loud is tough.

34:57

>> I I thought there was a very cool feature where it feels like they're bringing together a whole It feels like the first time for the last couple years.

35:04

It feels like Google's been like stuffing AI in little corners of the UI.

35:09

Like you already have Gmail and then you stuff a Gemini box there or you have Sheets and then you stuff a Gemini thing over here.

35:15

This feels like the first one where they were like sort of able to start from scratch and it still has like the sidebar panel, but it felt like it was both a code editor, but then it also kind of looked like a Google doc in the sense that you could highlight sections and leave comments for the AI, which I thought was interesting. >> Yeah, I don't know.

35:35

>> Easily guiding the agent's 90% solution all the way to 100%. >> Yeah, this part.

35:39

>> Now, let's say the agent produces a landing page mockup with nano banana and you now want to make some UI adjustments.

35:44

You can give visual comments. >> Yeah.

35:47

So you can actually like go in and comment in the image >> exactly where the problem is.

35:51

>> And you can do that in the text as well.

35:53

So you can like have this more precise dialogue with the agent like you would a human employee. >> Yeah.

36:00

>> And you're going to love it.

36:02

>> Say goodbye to what held you down before.

36:04

Welcome to Google Anti-gravity. >> Very cool.

36:09

Uh >> uh it is so it is funny.

36:10

Remember remember when when uh when Windsurf acquisition whatever you want to call it was announced >> and uh it was positioned it's like hey the team is wellunded and has a product used and loved by you know thousands of engineers and companies >> and I remember talking about it and we were saying like okay like the one issue is that some of the best people on your team are going to Google to compete directly with what you guys have been doing. >> Yeah.

36:40

So fortunately, obviously, you know, the whole Cognition deal Yeah.

36:44

ended up coming through, but uh you could imagine a world where Windsurf was still independent and just trying to and then suddenly it's like, okay, now you're just competing head-to-head with with your former partners.

36:54

Like, how does that make sense, right? >> Yeah.

36:58

>> Uh so anyways, it it all all worked out for the best.

37:00

But um but I'll be interested to see I'm I'm super interested to see what kind of adoption this gets.

37:07

>> Yeah, we Yeah, we have to we have to test it out.

37:09

We'll have to get the uh the Tyler Cosgrove review.

37:13

>> Is it Is it publicly available? Uh yes, >> let's get it. Let's >> get it.

37:17

>> Let's uh Yeah, let's do a review later this week and see how it compares to other uh other IDs.

37:22

Um anyway, we have our first guest of the show, Mike New from ArcGI in the Reream waiting room.

37:29

Welcome to the show, Mike. Thanks. >> Good morning, guys. For waiting. Good morning. >> Morning. >> How are you doing?

37:36

uh you know uh a lot of these AI sort of like uh verification things are very uh much hurry up and wait.

37:41

Uh so the last like 24 hours has been a hurry up mode. >> Okay.

37:45

>> Always very fun and exciting to get the results out but yeah it always comes together very very quickly at the end.

37:50

>> Well I really appreciate you taking the time to hop on on such a busy day.

37:51

Uh maybe we can just start with like your highlevel reaction.

37:55

Uh is like how do you even think about these things any anymore?

38:00

Are you just thinking like okay yes Gemini 3 good and then let's go a layer deeper.

38:04

Are you thinking about that?

38:05

What's your what's your high level takeaway?

38:08

>> Well yeah so you know I think the the big headline uh is that Gemini 3 basically got like 2x soda on ARC v2. >> Yeah.

38:15

>> Um and so this is uh you know this is the third major Frontier lab now in a year to use ARC to demonstrate frontier progress particularly with AI reasoning systems.

38:23

We had open AI last December >> XA this summer.

38:26

I'm super excited Google's now on the leaderboard too.

38:28

So that's great to hear and I should say up front thank you to the Gemini team for giving us the opportunity to verify. Totally has been great.

38:34

Um I think the really impressive thing about this and you know still still like sitting with all this stuff it's it's pretty fresh but I think the the biggest impressive thing to me is about we're starting to close this like complexity scaling gap between V1 and V2 ARC V1 and V2.

38:50

>> Like this is the big difference between what V1 and V2 is they look similar on paper.

38:53

If you go look at the different data sets, the big change is the V2 kind of increases the complexity of the tasks ones that take minutes instead of like seconds for humans.

39:00

Um, and so we're starting to see like actual material progress on that complexity scaling.

39:04

And then I think the big surprise to me personally is that Gemini 3 though is still roughly along the Prito frontier of V1. >> Yeah.

39:12

>> You know, it's a little better, but like it's still we're still kind of roughly within the same mass shape.

39:15

And um you know there's dozens of tasks where like you know the system still makes relatively I think you know obvious mistakes that humans don't make or recognize very quickly and you know I sort of previously expected like if we had an AI system that was solving half of V2 that V1 would be fully solved and like that's not the case.

39:33

So uh there there's a lot of surprise here.

39:35

Uh I was dreaming about this earlier to sort of invite sort of uh investigation from the community because I think there's still a lot to learn in terms of you know h how why exactly do we see such you know a jagged intelligence emerging right now.

39:47

>> Let me eliminate some uh some possible factors.

39:51

It feels like uh there is benchmark hacking but uh Google and the Gemini team feel not aligned with benchmark hacking generally like they've been good uh they've been good citizens in the community so far.

40:02

Um, and also you would assume, right, just from logical deduction, you would assume if you're able to hack V2, you would definitely go back and hack V1 as well.

40:13

So is that >> this is the first time we've verified a Gemini result either this year.

40:16

We we did two and a half earlier as well.

40:17

So yeah, I don't think that's >> so it's not like it's not like they set up like, okay, we got, you know, the most important thing here is that Gemini 3 is really good at RKGI V2.

40:27

That wouldn't make sense.

40:29

So there so this is sort of teaching us something about the fundamental nature of this model but we still don't know why lag why performance might be lagging in V1. Is that right?

40:41

>> Yeah I mean I've got my sort of hypothesis you know I think my my my personal one is that like AI reasoning systems just don't demonstrate even fluid intelligence.

40:51

>> Um you know the sort of like the ability for these reasoning systems to do adaptive reasoning which ARC is a sort of test of adaptation capability.

40:55

it's sort of limited to domains where the underlying foundational model has pretty good training coverage over the types of data and it has a verifiable feedback signal. >> Yeah.

41:06

>> Um and and I think that's sort of true for ARC.

41:08

You know, if I if I zoom out even further maybe, you know, to kind of put put this kind of result in context of where we're at as you know, just like an industry right now.

41:16

I think over the last 10 years, I would sort of characterize we've really had only two major breakthroughs.

41:22

We've had the transformer in 2017.

41:24

And obviously that led to language models and we had a chain of thought that was originally introduced in 2022 and sort of you know went through kestar into chain of into a reasoning systems and has gotten scaled up. >> Sure.

41:35

>> Um and and so like this was against the backdrop of like compute scaling right and this comput scaling was certainly necessary but it wasn't sort of sufficient.

41:41

These like key conceptual unlocks were sort of the sufficient things to take advantage of that compute.

41:45

Um, and so my kind of take at this point having looked at all this progression this year is that like AI reasoning systems with with no new innovation from here can basically enable sort of mass automation because a lot of problems can be charact can fit into that characterization where we can generate lots of examples that look like the problem and we can get a verifiable feedback signal from them.

42:04

Um, you know any problem that can be kind of cast and then characterized in that way I think can be automated at this point. No questions asked.

42:11

And then the big motivating factors I think really for >> mass innovation like that's that's sort of what we're still not seeing you know we don't we still need new ideas for this and I think that's closer to like an AI complete problem.

42:22

>> Yeah that makes sense.

42:22

Uh is that is it fair to like put you in contrast to some of what Darcesh has been writing about uh saying that uh like the job of most people is not necessarily a bunch of indiscreetly verifiable tasks.

42:36

Andre Karpath has been writing this as well.

42:40

there's this question of like like how much of a job is actually automatable.

42:47

Um radiology was one was one example um where it felt like a very automatable job and yet uh years into the AI deep learning revolution like we're still seeing full unemployment there.

42:59

Uh how are you processing?

43:03

>> Yeah, but we're only a year into the AR easing paradigm, right?

43:05

Like the first major one only came out 12 months ago and I think 2025 like in my view is basically characterized on starting to figure out how to actually bring these things into production systems. >> Sure.

43:17

>> Um like this is a big breakthrough.

43:17

I think this is the maybe like one of the mischaracterizations in my view of kind of the progress is is like a lot of teams even I think you know if you sort of just assume like oh models get better models get better you think like oh the last 12 months has just been sort of continued story and if I played with the models 18 months ago I have a rough sense of what they can and can't do and that's just not true. Yep.

43:36

>> Um like if you're a builder building products like this is the advice I give to you know teams I work with at Zapier too still is like look this is this actually is a significant paradigm break in terms of what was what's possible now that wasn't possible even a year ago with these systems and like that's going

43:50

to enable a lot of new types of products a lot of new types of services a lot of um use cases that were like out of scope because of verifi you know because of reliability and and sort of consistency now can be brought in scope so you know I think if your intuition on like what use cases are possible based on, you know, an 8-year look back. You really

44:06

You really have to start kind of pinning your look back to more more like 12 months. >> Yeah. Yeah, that makes sense.

44:11

What about uh uh like like does does the work live within SAS products or within individuals?

44:18

Because some of those examples that you just gave are uh it's like for teams that are going to build products that take that automate work and then get vended in through effectively SAS products to actually do the job.

44:32

the job. um versus like a knowledge worker who is going to be using Gemini in the app to you know accelerate their daytoday uh should they be feeling a res like the difference in this in the same way >> you know I mean like my one bit of advice is like if you

44:55

haven't really used these areas systems much you should I would hope everyone probably who's listen to the show has has used these things at this point but in case there's not like you should go you should go use and experience these things. Um you know when Google or when

45:04

Um you know when Google or when opening II released GPD5 this summer with their model router right that was like >> that was crazy >> predicated on this data that like very few users had ever even used dating systems.

45:14

>> Um and I still think it's only like one in five.

45:16

in five. Yeah, >> maybe it's >> and that was kind of part of the Deep Seek moment was just that for the first time there was a free app that you could go and see a chain of thought and you could actually see a reasoning model in action and for a lot of people that was

45:27

their introduction to that and so there was like DeepSeek wasn't necessarily that much higher that much you know in front of everything else but it just gave away a reasoning model for free at a time when they were tucked behind a bunch of other like uh hurdles that you had to jump through. >> Yeah, we're still really early on the

45:42

>> Yeah, we're still really early on the diffusion first stuff.

45:46

seeing that on, you know, the huge numbers getting reported by Frontier Labs and their usage data.

45:49

I mean, I'm seeing this in sales conversations I have for like, you know, Zapier stuff all over the case.

45:54

We're still very much early innings on actually getting this brand new breakthrough into um like production workflows. >> Yep. Yeah, that makes sense.

46:01

Do you have more questions on the diffusion? >> Yeah. >> Issue.

46:06

>> One, I I wanted to get uh your updated take on on humor.

46:08

We were playing playing around with Gemini 3 this morning specifically just trying to get uh on on our own little version of of humor bench.

46:16

It feels like something that like I I I do think about can you make kind of these like verify like can you make humor verifiable?

46:24

Like is there a system that someone could set up uh that that could um actually start um taking taking humor seriously?

46:32

Because I could imagine like if if we're hitting if we're hitting like any anything close to a wall, there will be a lab that says, "Okay, well, like let's work on something that like everybody uh that like let's work on a new kind of angle for differentiation and maybe maybe humor."

46:48

Uh could be >> at least a little bit, right?

46:49

Like I have a 5-year-old who is getting into uh starting to want to tell a lot of jokes and the jokes are just terrible, >> right?

46:56

Like they're not they're not funny at all.

46:58

They're they're like >> you end up laughing because they're so not funny and then depending who's delivering that is hilarious.

47:06

>> I've been trying to find the structured way to describe like, okay, here's what makes something funny.

47:10

And so there is like some degree which you can kind of break down, you know, the types of things I think humans would would sort of find funny.

47:16

And I like there is this actually does get pretty interesting because like you're getting to the spot where you're trying to like articulate like creativity, right?

47:23

How creative can these systems be?

47:25

you know, to be creative, to be humor, to generate do good art, you kind of have to like intentionally break the rules, but you need to have a really good model of what the rules are in the first place to intentionally break them.

47:33

Um, and in fact, I think a lot of humor fits into this category before into this as you're right.

47:38

It's like it's actually, you know, breaking the prediction rather than just following the sort of prediction of what you'd expect.

47:42

prediction of what you'd expect. Um and and today I still think when I look at the failure cases for let's call it AR reasoning systems on you know these tasks like ARC um yeah they still fail for what appear to be sort of random

47:53

reasons like they they have some some version of like an understanding of like the rules and strategy and the goals and then they sort of make a lot of basic mistakes either executing them or not following their own sort of like understanding that they've generated internally. So there's some sort of

48:04

So there's some sort of self-consistency issues and so like I feel like if that's still the case you know humor is going to be accidental rather than intentional from the systems. >> Yeah. Yeah. >> Uh what about V3?

48:13

We played around with that on the show.

48:16

I believe Tyler, our intern, was uh in the top 10 for a while.

48:20

Uh really grinded up the human leaderboard.

48:22

Uh I is it is it more compute inensive? Is that in the process?

48:27

Uh are are we expecting to see Gemini benchmarked to V3? >> I would love to.

48:32

So we are in the development process for V3.

48:34

I uh I like to say we've basically built the like uh highest uh most productive game studio in the world.

48:41

Yeah, [laughter] we're generating hundreds of these things.

48:43

We're about uh I don't know like two two two/irds of the way through building all the games at this point.

48:48

Our target is to get this in a good state with sort of all of our controlled human studies, all the games verified, get Frontier results checked off by early next year.

48:57

Um and we're targeting releasing it publicly in V1 with the entire data set or sorry in in Q1 with the entire data set next year.

49:03

And that'll likely be alongside our price 2026.

49:06

Y >> um still working on full details of how that's going to look next year. Sure.

49:10

>> Um but yeah, we're we're sort of like in the throws of it.

49:12

We're definitely using some of these frontier systems to do red teaming against the benchmark just to you know assert that like yeah these games are still hard for AI and we're still finding that to be the case even with things like Gemini 3.

49:22

Um but uh but yeah that's we're still in progress with development right now >> and uh Sema 2.

49:27

Can I have your reaction on on that?

49:30

Obviously it's this Gemini AI agent.

49:33

It feels like >> if anyone at Google is listening to this and could sort of give me access to Sim 2, I would love to test it on V3.

49:39

This is actually something that uh we haven't done yet in a >> Yeah. Yeah. Yeah.

49:43

That's what I'm getting at because it feels like uh I I I don't know if there's some sort of >> the claims are big.

49:50

You you read the marketing material and it's like okay that seems like it should solve V3 before it exists.

49:54

So like if that's the case, >> we should know that.

49:57

And so but yeah, I haven't got haven't gone hands-on with it yet.

50:01

So, I I can't sort of make any >> statement either way on the claims. >> Yeah.

50:05

I'd be interested also to to like when I'm thinking about like V4, uh it's like you you guys are going to have to build GP GTA 6 or something [laughter] like like if I'm Yeah.

50:14

If I'm following the progress of like V1, V2, V3, V4 is like a game that I'm going to play for 100 hours for fun.

50:23

I'm just going to pay for it. >> Yeah. This is one truth.

50:25

you you've some really something true about V3 which is that it's still a relatively short time horizon tasks and they're self-contained.

50:32

It does add some new complexity where you have to deal with interactivity because you have to do goal acquisition, you have to do exploration.

50:39

We'll have a really nice action efficiency comparison between humans and AI which we haven't been able to get before on the V1 V2 domain.

50:46

So we're going to get a lot of new signal I think on V3.

50:50

think on V3. Um, but yeah, I think as you sort of look even further out into the future, things that are more open-ended are the things I think we're starting to get excited about trying to like understand like what does it mean to put one of these AI systems in an open-ended environment and then look

51:04

back on the system, you know, 10 minutes in the future, 200 minutes in the future, thousand minutes in the future, and can you look at the environment that that AI system has been like how it's manipulated environment and like, you know, say something interesting about how intelligent the system is based on that like observation and open-ended sense. Um, still very early on V4, but

51:20

Um, still very early on V4, but uh, but yeah, we're starting to explore ideas there.

51:26

>> Has Gemini 3 updated your timelines at all?

51:29

Specifically, your ArcGI 2 timelines in terms of when you expect, you know, uh, sort of like the 90th 90% like anything on the kind of the upper end of the range.

51:40

I was looking back at my uh the whole ARC team actually made some predictions back in January when we released V2 on what did we expect endofear scores would look like.

51:49

Uh now obviously if we're only November 18th a lot happens in in AI.

51:54

Who knows what the next six weeks hold.

51:55

Um but my personal prediction was that we would see about 25% on the private leaderboard for RV2 on the Kaggle contest and we'd see about 50% on the public leaderboard.

52:04

uh uh and and that was sort of based on the ratios we had seen from ARC Christ 2024 and you know the sort of scaling difficulties with V2 and it looks like we're pretty going to come in pretty close to that unless but barring some other major new breakthroughs towards the end of the year.

52:19

um that seems like we're probably where we're going to end up the year at.

52:22

Um and uh and then who knows on 2026.

52:24

You I think it [laughter] if we're really going to solve V2 fully, it feels like we got to better understand why these AI reasoning systems still make sort of obvious mistakes on the V1 set.

52:37

>> Um and yeah, I that's that's an anomaly.

52:40

So I think that's that's worth serious study uh to like come up with new ideas to sort of prove these reasoning systems. >> Yeah.

52:45

What was the furthest timeline that you had out?

52:47

I remember you said when you developed V3, you had this framework of like like the state-of-the-art should be scoring like negative 100% or something, you were like you need to make it way harder than you think in order to give you like room to run because the systems are developing so quickly.

53:05

Uh what's the furthest out timeline that you are tracking or or or you as a team are tracking?

53:11

I I mean our our objective function is not longevity necessarily.

53:18

It is usefulness and interestingness.

53:21

>> Um I think the tasks that have the highest degree of usefulness and interestingness are ones where you know oh hey this could um be useful and interesting for like three years. >> Mh.

53:32

>> Um the ark one was useful and interesting for arguably five year.

53:33

interesting for arguably five year. I mean even this year it's still interesting because we haven't broke it like we're still sort of within this sort of paradigm still and so it's still providing some interesting useful even

53:42

though you know it's largely saturated up to 80% now but there's there's still interesting signal remaining um V2 our expectation was that it was not going to survive as long as V1 just because it was the same domain um and we had a reasoning systems in play at that point. Yeah. Um, yeah, I think our median Yeah.

53:57

Um, yeah, I think our median estimate was like 24 months on V2, but like that, you know, we'll have to see how that all plays out next year with that.

54:04

V3 we're hoping to put in a we're hoping to be in an environment where we can actually get that to survive sort of longer.

54:09

Um, >> you know, one of the interesting things we're finding with V1 to V2, V to V3 in sort of like a qualitative sense is um there's there's a there's a there's a sense of like how easy is it for us to generate the data set as like humans trying to design the tasks and design the puzzles and design the games.

54:25

And with V1, pretty much every like task that like France created um was was hard for AI and easy for humans. >> Yeah.

54:35

>> With V2, that gap got a little shorter.

54:38

Actually, it got smaller.

54:38

um there were tasks that we generated as humans that um AI um solved and there was other ones that were too hard for humans and so we ended up sort of pruning some of the tasks that we generated.

54:48

So like the gap between those things got short.

54:49

With V3 we're finding it's getting wider again >> where pretty much every game we're coming up with is like fitting into this paradigm of like very obvious and intuitive and easy for humans and sort of very hard for frontier AI still. >> Yeah.

55:02

Um, and I think that's like uh this credit to France here, you know, this is something he shared about a year ago with a three, but he's like this is actually one interesting way you could characterize how close are we to AGI is like when we run out of when humans run out of the ability to generate interesting things of what your AI can't solve, like hard hard to argue any expert's going to say, yeah, we don't have AGI. >> Yeah.

55:22

Because you can sort of think about like the project of humanity is like go do the hard and novel things.

55:25

So it's like is is acquiring diamonds difficult?

55:30

Okay, that has value and then we base a whole economic system around it and it's like somewhat arbitrary but it's also like a skill and might and will issue and if you can put that on display then you acrewue economic value and so that that that kind of traces out into everything that we do in in life and beyond.

55:47

>> Last time uh you were on it if I remember correctly you you made a call for new new ideas needing new ideas.

55:55

What's the update on on that front?

55:55

any are you seeing anything promising outside of LLM world?

56:02

>> The um yeah, there's some pretty interesting stuff coming out from our crush 2025.

56:06

We we in we're in the throws of like reviewing all the papers, judging all the scores, the official results for our prize 2025 come out on December 5th, I believe.

56:16

So, I have to can't share everything yet.

56:18

I don't want to spoil the the final announcement.

56:19

I think one of the big things that we saw from ARC prize 2024 was this concept of like test time adaptation.

56:26

Um this was the idea that like look a pre-trained model applied through a single forward pass at inference time will never solve ARC.

56:31

You need some ability to take information from your test and incorporate it back into uh into the the system and that's where your adaptation capability comes from and that was done through like test fine-tuning during the contest.

56:43

AI reasoning systems are a version of this where you're incorporating to sort of private data set >> tuning. Wow. >> Yeah. Yeah.

56:49

literally like you take a pre-trained model and then like take the secret the private puzzle augment it in a bunch of different ways to generate permutations of it and then do like a lura or some sort of test fine tune on your pre-train and that that actually works. >> Wow.

57:02

>> The the sort of uh the the the the common ground between this and reasoning systems is that both of them take information from the private test and are able to operate over it with it at test time. Right?

57:12

This test time compute is another form of of what we're talking about here. Z 2024.

57:15

One of the big things we're seeing on Prox 2025 is this concept of refinement loops.

57:20

Um anywhere where like particularly with like language models being put into outer outer loops where they can sort of move from state to state and how they move from state to state is like they need to make some sort of refinement on the program or the natural language explanation of the task that they're working towards and they just iterate on this like refinement loop over and over.

57:40

uh and this is significantly increasing scores even over the sort of test time fine-tuning stuff that we saw from from last year.

57:45

So Jeremy Burman and Eric Pang were two folks who were on the public leaderboard last month uh that explained how their approach worked in this way.

57:52

Um so we're seeing a lot of approaches like that.

57:53

Um I still think we're in a regime though where like we still need new ideas.

57:57

Uh none of these are sort of sufficient to solve arc um including inclusive of v1.

58:01

Um and so like you know this gets me excited because I still think that means individual people, individual teams with small budgets, small compute budgets um can still play a really really massive role in advancing AI. >> Yeah. Uh very cool.

58:16

>> Are are there other areas where uh we are making progress in AI that might sort of need to come together to uh to actually maybe solve this or maybe just be a more complete system.

58:27

Uh what I'm thinking of is like uh very few solvers are that I'm aware of uh will actually just take a screenshot of the puzzle and inspect it with some sort of diffusion model.

58:40

Like that's not the way these these AI models uh reason about arc puzzles.

58:45

Uh we're also seeing a bunch of work on uh world world models and simulators world simulators which seem really interesting.

58:53

And I was talking to one guy who is building one and he was saying like I I think that we're going to get like really really robust knowledge out of these at some point once they scale up fully.

59:04

Uh and I'm wondering if you are optimistic about uh bringing in other like unifying some of the different research that's happening.

59:14

I think it's um all of those examples of new research, new companies, new startups like you know there's this there was a seismic shift in 2025 from pre-training budget to uh these like RL reinforcement learning environment uh

59:26

startups and companies that are generating environments to produce >> uh more ground truth training data in mass way because they're you know automated environments and you can get verifiable feedback signals out of these things. Y

59:36

Y >> um again no there's no new science here like this is a good bet for like all frontier labs to make.

59:40

This is going to drive progress for the next 24 to 36 months.

59:43

you're going to continue to see amazing frontier headlines just just on just on this fact.

59:47

There's really no new sort of I think discovery that's that's quite needed there.

59:52

Um, you know, I think that if you're kind of pushing more towards the AGI side, you know, like what's what's sort of missing?

59:57

Like one question I have um that is an open question is so we've got like you would think that based on like a 100x to 300x increase in efficiency we've seen from AI reasoning systems over the last 12 months that we would trade that increase in efficiency for inference tokens to do >> more like search coverage over the problem space when we're giving these systems tasks or problems that we want them to solve. Mhm.

1:00:27

>> And this is one of the big reasons why I sort of expected if we can solve half of V2, you'd get 100% of V1.

1:00:30

And it seems like these AI reasoning systems are are are like not sort of fully exploring all of the search space uh that they could in order to sort of look for solutions.

1:00:42

Um, and so I have like kind of an open question of like, well, how much of the search base can they cover?

1:00:46

And what do you need to change about the training methodology or process to like actually guarantee that you can get full coverage over the search base um of like possible programs or possible solutions?

1:00:55

Um, and and so that's kind of that's like one interesting thing that I'm paying a lot of attention to right now. >> Yeah. Yeah.

1:01:01

the even just the metaphor of uh like the the test time fine-tuning.

1:01:09

It feels like working on a problem and then like going and taking a walk and kind of like updating your whole world view like it feels like something that uh humans get get closer to doing that than any of the other paradigms.

1:01:21

Uh so yeah, it's fascinating to see all these different uh approaches.

1:01:25

approaches. Yeah, very >> all the crazy results you've heard about in the last 12 months are kind of this merger of like deep learning and like symbolic program synthesis style methods the IC ICPC the IMO gold the Gemini 3 stuff today like you know these are all systems that are you know still fundamentally using language model but they're adding symbolic knowledge

1:01:42

recomposition systems on top of these things they all work slightly differently okay >> um but it's like >> this is what's working right now and so I think the rough like search space of research and how you merge those two paradigms together is still relatively underexplored There's a lot of different ways you can put these two paradigms together. >> Y

1:01:57

>> Y >> um and uh you know for new teams that are considering work on new ideas like I would explore like well what are the novel ways you could consider merging these two spaces? >> Yeah.

1:02:05

Yeah, that makes a ton of sense.

1:02:06

Uh Jordan, anything else? >> This was great. >> This is amazing.

1:02:08

Thank you so much for jumping on on short notice.

1:02:10

Uh and >> as always, guys, thanks for having on the continued uh the continued just stacking up the wins on RKGI becoming uh >> and just continuing to mog the models, mog the world. >> Yes.

1:02:23

[laughter] I mean, again, our goal is to be very useful uh and interesting.

1:02:25

So, we're going to try to hold that bar.

1:02:27

My words I think you're keeping them honest.

1:02:30

I think you're keeping everyone honest.

1:02:31

Uh and you're providing like a very very useful uh reality check on on on an industry that loves to >> inspiring the labs to grind harder >> and and now and now there is a there is a moment where we can uh feel very confident about taking victory laps and and and cheering for all the hard work that went into Gemini 3 because it does seem like it was a great model. It's performed well.

1:02:51

There's definitely a big improvement today. >> Fantastic. Well, thank you so much.

1:02:53

Have a great rest of your day. We'll talk to you soon. >> December 5th. We'll see you then. >> We'll see you then.

1:03:01

>> Um, I wanted to talk about Adio >> because Adio is an AR native CRM that builds, scales, and grows your company to the to the next level.

1:03:07

Also wanted to talk about wander. com.

1:03:09

Book a wander with inspiring views.

1:03:11

How great men, dreamy beds, top your cleaning, and 24/7 concier service. >> Let's sing it.

1:03:16

>> Find your happy place. Find your happy place.

1:03:19

>> Book a wander with inspiring views. I already know the song. You know the song.

1:03:23

>> I wanted to pull up this post from Chris Pariski.

1:03:28

He did a GitHub style image of our streaming activity for the year. Oh, really? See this? >> Oh, yes. I did see to him.

1:03:37

>> Should be at the very bottom.

1:03:37

It's at the very bottom of our timeline. >> I have it.

1:03:42

>> Uh, and if we could just pull up this image.

1:03:44

>> So, the internet rewarded TVPN for showing up on January 28th.

1:03:46

That's when we went live.

1:03:48

We never we never remember the day that we went live, but he has it. He looked it up.

1:03:53

January 28th, John Kugan and Jordy Hayes launched a daily live show and set one simple rule. Show up 5 days a week.

1:03:59

Looking back, they did exactly that.

1:04:01

125,000 followers on X, 41,000 subscribers on YouTube.

1:04:04

Uh 17 and a half thousand on Instagram.

1:04:07

They showed up every day the internet rewarded the proof of work.

1:04:11

>> So the only thing is these I don't am I just color blind?

1:04:13

But is it like a little bit like I'm seeing three days that were federal holidays that we missed and then three days that were >> no streams. I can't exactly tell. Yeah.

1:04:24

What what is a federal holiday? What is a no stream?

1:04:25

Uh the uh it looks like maybe a gray and a purple.

1:04:29

Uh there were a couple days here and there. Um we took one off.

1:04:34

I went to a wedding in Mexico.

1:04:37

Uh we took a Friday off for that.

1:04:37

That was just no live stream.

1:04:39

Uh July 4th uh we took off. That was a Friday.

1:04:42

That was a federal holiday.

1:04:44

And then what happened in uh in March? We took a Wednesday off.

1:04:48

No live stream on Wednesday in middle of March.

1:04:52

>> There was one day that we were traveling. >> Oh yeah.

1:04:55

>> That was after um >> after Hill and Valley after DC.

1:04:57

Uh I thought it was a Thursday day though. >> No, no, we didn't. We did.

1:05:02

Tuesday in the in the hotel room and then and then Wednesday we did in the uh at the actual event Hill and Valley and then we flew back and got back on the horse.

1:05:12

So we missed a couple Mondays because of federal holidays and then we missed a Tuesday in in May.

1:05:18

That might have been Hill and Valley.

1:05:20

March might have been something else.

1:05:21

Anyway anyways it's been a wild ride.

1:05:23

Uh thank you to everyone pulling it together along the way.

1:05:25

Uh our next guest is I believe already here.

1:05:30

Uh we have Jonathan Neman from Sweet Green.

1:05:32

We're going from benchmarks to bench presses.

1:05:35

The most important benchmark in the world.

1:05:38

How many grams of protein are in your protein bowl? We need to know.

1:05:43

Uh, welcome to the stream.

1:05:47

Please introduce yourself for those who might not be familiar. >> Hello.

1:05:50

My name is Jonathan Eman.

1:05:50

I'm the co-founder and CEO of Sweet Green.

1:05:54

>> Get that overnight success button ready.

1:05:58

When did you start this company? >> 2007.

1:05:59

So, we've been at this for 18 years. 18 years. Wow.

1:06:02

Uh t just let's talk about the the the very beginning.

1:06:06

I mean, since uh this is your first time on the show, uh where did you grow up?

1:06:13

How did you get into the business? What were you studying? And then uh let's go. >> Yeah.

1:06:17

You got to be somewhat of a masochist to get into the restaurant business.

1:06:19

[laughter] >> Yes, absolutely.

1:06:21

>> I mean, it's it's such a beautiful thing because it sounds so simple.

1:06:23

It's like you get a box, you get a menu, you get some ingredients.

1:06:26

Sounds it sounds super >> and then you just copy and paste it and you scale to you know however many stores and then of course it's far harder in uh >> so prior prior to launching the business what were you doing?

1:06:37

>> So I grew up here in Los Angeles I went to school in DC went to Georgetown and never thought I'd be in restaurants but >> you studying government you thought >> no I was studying business I always I knew I wanted to be an entrepreneur >> and Sweet Green was almost almost an accident you know we it was the naivee we thought it would be easy.

1:06:53

>> Yeah we >> And did you start it during school?

1:06:55

Yeah, we started in we started while we were seniors in college.

1:06:57

I started with two of my friends before that >> was doing Yeah.

1:06:59

I had a bunch of internships, you know, I was you know worked in media, I worked in tech.

1:07:02

I you know I worked in real estate.

1:07:04

Always knew I wanted to be an entrepreneur and create something.

1:07:08

But senior year came around and it's exactly what you said.

1:07:12

We thought it would be easy.

1:07:13

>> We're like how hard could this be?

1:07:13

You go, you know, we'll go to farms. >> It's like apparel.

1:07:17

Like people [laughter] fall into the apparel trap because they're like I just wanted to make clothes that I wanted to wear and then you realize it's like the hardest business on apparel and restaurants probably the things that seem the most simple but are actually the hardest practice to actually do on a massive scale. >> Yeah.

1:07:33

So what was the was it was it build a business plan first? Assemble a team? Uh do a popup?

1:07:36

Like what was the first thing where you were like okay let's >> What was the first bowl?

1:07:41

The first bowl was the guacamole greens we made in our dorm room.

1:07:45

We brought a bunch of classmates to try it.

1:07:46

My partner Nick actually made it. He was our first chef. No way.

1:07:50

And the the the story was really simple.

1:07:52

We had no we couldn't find a healthy place to eat.

1:07:55

>> We saw Chipotle taking off and we're like wow there's someone is going to create a >> scaled healthy fast food chain.

1:08:03

>> And at first it was let's just open one.

1:08:05

Our you know we wanted it for ourselves.

1:08:07

We thought we'd go on with our lives.

1:08:07

We opened we worked on it senior year.

1:08:11

Wrote a business plan raised $300,000. >> There we go. from 50 investors.

1:08:14

So, [laughter] >> so it was like five five grand, five grand average. Yeah. Party round.

1:08:21

>> So, they got equity in like what became the full company. >> They got equity.

1:08:24

Well, we actually it it was a little bit more complicated than that.

1:08:28

At first, the first three restaurants we raised at the restaurant.

1:08:31

>> At the restaurant level. >> Yeah.

1:08:32

I was wondering if you were doing that.

1:08:33

>> And we actually paid the investors back every quarter and did the whole thing.

1:08:37

And then after the after the third restaurant, we realized that the only way to scale this was to roll it up.

1:08:40

So we rolled the whole thing up >> and then we were able to continue to invest in it and we >> it's notable when did the word wellness actually become mainstream or when did that become like a identif like 2015 years ago. >> Yeah. Like early 2010s. >> Yeah.

1:08:55

So anyway, this is like any anyways at least 5 years before wellness is going like mainstream. >> Yeah.

1:09:02

When we were when we were starting you know the the thesis was healthy eating was not cool. Sure.

1:09:06

and it was not delicious and it was not accessible.

1:09:09

And >> we're going to create a place that offers all of the benefits of fast food in terms of the convenience and the taste, but do it in a, you know, do it with healthy food and real real food that you can trust, where we're transparent about where the food comes from, where it's nutritious, and build a brand around it.

1:09:25

And so we've been at it for about 18 years.

1:09:27

We have almost 300 stores all around the country. >> Yeah.

1:09:32

It's almost hard to believe. >> Yeah. >> Yeah.

1:09:34

>> What was the first uh VC round?

1:09:34

>> What was the first uh VC round? the first so we >> or like or this transition from the you have a restaurant and what did it work immediately you set up one restaurant you you know you raised enough money to get that I imagine that you had to sign a lease so you weren't buying buildings

1:09:50

but you might have to do some sort of renovation to actually get the first restaurant up and running you start making money enough to pay the employees enough to pay the rent uh you scale that to three and then at a certain point you say okay we're going we're going to turn this into like a corporation more than just a small mom and pop, right? >> Yes. So, we we opened one in 2007, >> Yes.

1:10:07

So, we we opened one in 2007, [clears throat] two in 2009, um with a food truck. You remember those? >> Yeah.

1:10:15

>> U and then we opened like two or three a year and we were mostly uh built them from cash flow from the >> profit.

1:10:20

We were profitable, you know, we would just reinvest the cash flow and we would do a few party rounds. >> Yeah.

1:10:26

>> Uh 2013 along the way, we started a big music festival called Sweet Life.

1:10:29

2010 it became a massive 25,000 person music festival. Where was that?

1:10:35

>> It was at Merryweather Post Pavilion.

1:10:37

So, first year we had the Strokes.

1:10:37

By the end we had Kendrick Lamar and little festival side quest. >> Yeah.

1:10:42

It was a, [laughter] you know, way to build the brand.

1:10:43

And then in we focused on DC which was very, you know, it was almost an accident, but we opened the first 16 restaurants in DC. >> Wow.

1:10:52

>> And then slowly went up to Philly and then restaurant 20 and 21 were Boston and New York. Okay.

1:10:59

>> And Boston and New York I really kind of proved the concept outside of DC >> and took off and that's when we raised our first.

1:11:06

>> So now obviously all around LA there's sweet greens, but why given that you grew up here, why didn't you why why not start here?

1:11:13

Was this because well like >> there was did was there just more healthy food options in LA and there was less on the >> honestly it was an an accident.

1:11:20

We were in school and we're like let's just open one.

1:11:23

We thought with the second one would open.

1:11:25

gravity that you have around the center when there's more and more stores.

1:11:29

>> You know, when you have a restaurant company, the brand and all your economies of scale happen at the local level.

1:11:33

>> So for us especially given our supply chain is regional.

1:11:37

>> Um you have your overhead and your management like your team that runs it and then your brand you know restaurants the brands don't really travel across the country. Occasionally they do.

1:11:46

>> Um so it was really started in DC.

1:11:46

We thought the second restaurant would be in LA. We went and looked.

1:11:51

>> This is true for even like In-N-Out is not a not a national brand still.

1:11:53

It's like a west coast brand somehow.

1:11:55

Uh and yeah, it's taken so long for that to actually like filter across.

1:11:59

Uh what how capital intensive was it to launch like the second and third?

1:12:04

Like you mentioned 300,000.

1:12:06

>> That's the hardest part of the No, it's way more than that now.

1:12:08

Um the first one was tiny 500 ft and we did it really on the cheap. >> 500 ft. >> 500 ft. >> Yeah.

1:12:16

So that's like I imagine like one or two people like >> Wow, that's tiny.

1:12:20

>> Yeah, we were working there.

1:12:20

We doing the whole thing.

1:12:21

So I mean we've had to raise a lot of money. Answer earlier question. uh revolution.

1:12:24

Steve Casease was our first first VC investor >> and it was part of the thesis was how technology can change the restaurant business.

1:12:33

So we were we were the first company to you do mobile ordering where you can order on your app and pick up >> and we started >> most beautiful software for >> that a restaurant had ever had probably [laughter] emit EMTT Shine >> Emit Shine. Yeah. Yeah. Jin Lane Jin. >> Yeah.

1:12:49

This was like uh yeah this was like one of my favorite Jin Lane projects. That's awesome.

1:12:54

>> EMTT and his team were amazing.

1:12:54

They they did they they did our app in the early days.

1:12:59

>> And you know, restaurants are today cost over a million dollars. So we're like $1. 3 million, $1. 3.

1:13:04

4 million per restaurant.

1:13:07

That's before you put the infinite kitchen in. >> Mhm.

1:13:11

>> Our restaurants have very high return on capital. >> Infinite Kitchen. What's that?

1:13:14

>> The infinite kitchen is our automation. Okay.

1:13:16

Uh our automation platform that we that we've built.

1:13:18

So today, most restaurants that we open, the assembly is automated.

1:13:22

So we still make all the food from scratch.

1:13:24

The sourcing is the same.

1:13:25

We still cook the food fresh, >> but it we load this beautiful machine that that makes your bowls.

1:13:30

It makes them 500 bowls per hour, perfectly portioned, perfectly plated.

1:13:36

>> Um, and so that is kind of the future of where things are going.

1:13:39

>> How many different restaurant automation pitches have did you get across 18 years?

1:13:44

Like cuz I imagine every single year there's a new like startup coming to you saying like we can automate this part of your kitchen and >> clearly you got to the point where you had to build it yourself based on kind of domain knowledge but uh this just feels like something that's been promised for a long time and at this point I don't know like an individual startup that's done well in restaurant robotics. >> Yeah.

1:14:09

No one's no one's been able to create a platform that that works in multiple restaurants.

1:14:14

And there's a few there's a few issues.

1:14:15

Most restaurant workflows are very specific.

1:14:19

>> So they're super specific to that restaurant.

1:14:22

>> Two, most restaurants are franchises.

1:14:25

And so they're not owned by the corporation.

1:14:26

We are fully company owned.

1:14:28

So if you're a franchise restaurant, you know, if you're McDonald's, you have to now go convince your franchises to buy whatever automation you have.

1:14:34

And the other issue >> they're looking at it and it's like this is coming off my bottom line.

1:14:37

We're making money already. This feels like a risk.

1:14:42

Like it the franchisee is saying like what like I'm happy with my IBIDA.

1:14:48

I don't need to take a risk. >> That's exactly right.

1:14:49

And the other issue is you need automation that takes enough labor out or offers enough value to be worth it because the capex is still very heavy. >> Yeah.

1:15:00

>> So when we went down this path, we tried to build it ourselves actually.

1:15:01

We built a team to do it ourselves. >> Yeah.

1:15:04

realized how challenging it was.

1:15:06

And then we found this startup that was doing it and doing a really good job. Yeah. It was called Spice.

1:15:09

It was called Spice Kitchen.

1:15:10

It was for MIT grads out of >> four grads out of MIT.

1:15:13

And they had the same issue.

1:15:15

They realized they could build the automation, but no one was going to buy it. >> Yeah.

1:15:19

>> So they ended up opening two restaurants.

1:15:20

Um they were great at automation, not so great at the restaurant side.

1:15:24

And then four years ago, we acquired them.

1:15:25

Um and we began, we've commercialized the technology, we've scaled the technology um today.

1:15:29

So most new restaurants feature the technology.

1:15:34

And last week we actually just announced that we've now sold Spice.

1:15:38

>> Um so we sold out basically. >> So we spun Spice out. Yeah. We spold spun Spice out.

1:15:42

We announced about 10 days ago.

1:15:44

We sold it to Wonder Mark Lure over there. Yeah.

1:15:46

So So we sold it for about $186 million.

1:15:48

Mark is Mark Mark uh uh I I don't I don't fully understand that that business, but talk about a guy that just like isn't even necessarily naive about the challenges of restaurants, which just like I'm going to go into the most competitive environment possible and compete with everyone. >> It's amazing.

1:16:06

It's a great It's a great vision and, you know, I'm a big fan of his and what they're doing.

1:16:09

And so we we >> it's a it's a really interesting deal.

1:16:13

So we we sold the >> the effectively the team and the IP, but have full access to it.

1:16:16

So we will continue to scale with it and get the benefits as they get >> they get to you know scale and build many more machines we'll get the benefits of those economies of scale as well.

1:16:28

>> Uh wait can you go a little bit deeper on the decision to uh franchise or not franchise uh the naive maybe steelman for franchising the franchise model is that uh it's somehow more capitalist in my mind.

1:16:43

it like because it decentralizes the decision-m and it and it puts these financial incentives at the local level because each store lives and dies by its own P&L maybe uh versus even if I have a manager in one store and they have stock options like how what they do on the weekend if they come in on Thanksgiving or Christmas like that doesn't necessarily put more or less money in their pocket. Is is that real?

1:17:09

What I'm what I'm feeling or is it irrelevant?

1:17:11

irrelevant? what you're feeling is absolutely real and we actually try to design our comp structures and okay the you know I've always believed I my line that I say tell my team every single day is all the answers are in the restaurant and the closer we can push decision making to the edges to the customer the

1:17:28

better we will be so we you know our general manager we call them the head coach they are the most important position in the company by far a great head coach will make or break you >> and so we try to really incentivize them we empower them and we try to run as decentralized as an of an operation as we can. Okay. Okay.

1:17:44

>> The reason we decided not to franchise is >> it's really hard to maintain quality.

1:17:50

>> Um if when you give up that when when you know when you really give that up to other people to run, you could sometimes scale too quickly.

1:17:56

And we do a few things differently. We source differently.

1:17:59

We're a very complex model because of the sourcing and the scratch cooking.

1:18:03

The biggest difference between us and most of other companies is if you go into Sweet Green, you'd be shocked at how much we are making in the store.

1:18:12

>> It feels like you guys have taken such a principled approach in making food that I feel like stays true to the initial values of the company and kind of why you started it.

1:18:21

And yet you're competing in an environment that says, "Okay, we're going to have these like factory kitchens offsite that we're going to be shipping in effectively almost finished product that gets reheated and we're going to be sourcing from all over with not a lot of values around how they're sourcing.

1:18:37

They're just trying to get like they they want the food to taste good when it hits the plate, but maybe they don't care about uh a number of other factors.

1:18:44

And so you're kind of in an environment where because of your principles, you're like fighting with your hands tied behind your back in against competitors like and I'm not talking about direct competitors, but more so like you're still competing with Burger King and McDonald's, right?

1:18:56

Like people are going to have lunch somewhere and they're going to maybe decide between they have options, right?

1:19:04

>> Talk to us about land.

1:19:04

Is McDonald's a land acquisition company?

1:19:07

What like like why do people say that? Is that real?

1:19:11

Have you ever looked they do they do own a lot of a lot of the real estate and lease it back to the franchise.

1:19:15

So that is true and if you've watched the founder the last line in in that movie where he's like it's a real it's real estate.

1:19:22

It speaks to more than the fact that they just own it.

1:19:25

>> Restaurants is highly a real estate game.

1:19:27

Like great real estate mean is like if you look at like our portfolio where we have great real estate we do amazingly well.

1:19:34

Location, location, >> location, location, location.

1:19:36

Really, it's it's people.

1:19:38

Like people think restaurant business is a food business.

1:19:42

It's really a real estate and a people business.

1:19:44

And it's all about like you look at the great restaurants.

1:19:46

So the Chick-fil-A's, the Raising Canes, the In-N-Out, it's so much about it's about that culture.

1:19:51

How scient how scientific is you you hear stories of of companies like Starbucks and you can imagine like a team of data scientists with like you know 50 monitors and they're just >> we need one Starbucks directly across the street from the other Starbucks. >> Yeah.

1:20:07

You know so like you can imagine a world where it's like hyper like hyper data driven and like down to a science and you just know when you're opening a new store you know that it's going to hit.

1:20:18

But there has to be like some five days.

1:20:22

>> Yeah, we that is the process.

1:20:22

We call it art and science. Okay.

1:20:24

In pretty much everything we do, it's it's an art and science approach and real estate's exactly that.

1:20:29

You know, the the science we have a very very intricate model that looks at psychoraphics, demographics, mobile data, drive, you know, people driving by.

1:20:40

>> We have custom data on how many gyms nearby and right side of, you know, sunny side of the street or not sunny side of the street, all of that stuff.

1:20:47

But then you need a human to also walk it, feel it and understand does it tell our brand story.

1:20:52

For us we especially when we were like early days growing where we went said a lot about who we were.

1:20:59

So for example we went to New York. We didn't go to Midtown.

1:21:02

>> We went to Nolita, we went to Williamsburg.

1:21:04

We wanted to kind of tell the story about who Sweet Green was.

1:21:07

>> You know today we're kind of everywhere.

1:21:09

>> But the real estate is is an art and science and tells a lot about you know says a lot about who you are. >> Yeah.

1:21:14

>> Yeah. How do you think about uh if a new entrepreneur came to you and was asking for advice on uh where to start, is it is it worth it to go straight to Manhattan or straight to uh Beverly Hills and uh and try and like make it in

1:21:31

the big leagues on day one or is it >> or can you get negative indicators from that because there's a different type of customer there that's not necessarily representative of the rest of >> I think that's more right especially when you're talking about New York. So

1:21:42

So when you when you're talking about New York, it is I mean the beauty of it is a massive market, you know, it's it's for us about a quarter of our business. It happens in New York.

1:21:50

We have like you know in the New York region I think we have 50 something restaurants. >> Wow.

1:21:55

>> Um what so it's great that it's massive, there's density, there's you know they have money etc.

1:22:01

>> But it's not really indicative of the rest of the country. >> Yeah.

1:22:04

So if you wanna, you know, if you want a scalable model that you can have thousands of locations, you're better off, you know, going into a more, you want to go to like the Iowa, >> you know, like and to use the political uh analogy, you want you want to go to like something that, you know, is more representative of what the rest of the country looks like in restaurants.

1:22:23

The place where everyone goes, the fast like the fast casuals is Columbus, Ohio.

1:22:28

>> That's where people go.

1:22:29

>> They say, you know, Columbus, if you can make it in Columbus, Ohio, you can make it anywhere.

1:22:32

You can kind of make it everywhere. >> Yeah. Yeah. Yeah.

1:22:35

So, I mean, if you were a small restaurant, you're being evaluated by uh, you know, the the CEO of McDonald's or something, you might say, "Okay, how are you doing there?" >> Yeah.

1:22:43

The things that they look at for for a restaurant is they look at your unit economics, which is effectively your payback.

1:22:47

So, how much does it cost to build and how quickly do you pay those stores back?

1:22:51

>> And they look at your TAM.

1:22:51

So, they say, okay, like, >> can you have a hundred of these, a thousand of these, 5,000 of these?

1:22:55

And those are the two big kind of thing, you know, things you would look for in evaluating like the growth trajectory of a restaurant.

1:23:03

>> What's the story of the the the delivery market?

1:23:07

Uh it feels like Door Dash has become massive business.

1:23:11

Uber Eats has become a massive business.

1:23:13

Uh more people are ordering delivery.

1:23:15

There's the ghost kitchens trend.

1:23:17

Is there a ghost kitchen where these businesses are like trying to effectively turn you into ghost kitchens?

1:23:25

Is does that give them some sort of leverage?

1:23:27

Is there some sort of tension there or is it pretty much just like, oh, it's just this trend.

1:23:32

People are cooking less and less and so they're going to go to Sweet Green, but they're also going to order Sweet Green delivered more.

1:23:38

>> There's there's definitely a little tension there.

1:23:40

You know, we're partners.

1:23:43

A lot of our business comes through those marketplaces, >> but it's not so dissimilar than, you know, a hotel chain and Expedia. >> Sure. >> Right.

1:23:50

It's it's you're paying a fee on it.

1:23:52

You do not control that data.

1:23:52

you cannot market directly to those customers.

1:23:56

>> And so for us, we have to charge a higher premium.

1:23:58

So people when you order on Door Dash, by the way, it's more expensive than you order on our app.

1:24:01

So just a quick shout out, order, you know, download the sweet app.

1:24:04

Things are about 20% cheaper there. Got it.

1:24:08

>> Um but at the same time, it's a great way to find new customers. Sure.

1:24:10

>> So you know, for example, Door Dash has been a great partner.

1:24:13

They power our native what we call our native delivery delivery on our Sweet Green app, which is a big part of our business. >> Yeah.

1:24:19

So you so you're white labeling or something, >> correct? Yeah.

1:24:21

It's like a white label on on our app.

1:24:22

And then we also, you know, we partner with >> front store as their front end.

1:24:27

>> And as you know, they've become, you know, they're brilliant business models.

1:24:29

They've become largely marketplaces.

1:24:31

>> So, you know, they you kind of have to buy your way to the top of the feed. >> Yeah. >> Yeah. Yeah.

1:24:37

>> And so >> that's how they gain they I mean, if if uh like there's a reason the Door Dash app or any of these mobile ordering experiences are not they don't just put like the the restaurant that you've ordered the most from at the top.

1:24:48

It's like, "Hey, why don't you try this new restaurant or this [laughter] new restaurant?" Yeah.

1:24:52

They're all paid and it's how you maintain leverage over I mean they they do this on this is why the YouTube subscriber count doesn't mean anything because it's like they're going to constantly surface >> anything. They've made money.

1:25:05

I mean, the way they make money though, these businesses have been historically very challenging.

1:25:10

>> The way they made it work is batching orders >> and and and then becoming an ad marketplace.

1:25:15

And and that's what's made, you know, this amazing service an amazing business.

1:25:20

>> Explain batching orders really quickly.

1:25:21

>> So when you order, uh they they have a delivery driver pick up multiple orders.

1:25:25

>> So you're paying the delivery driver, you know, once, but they're picking up from three restaurants.

1:25:30

>> Uh I feel like you guys have done a really good job of listening to customers.

1:25:36

>> I would I would say like this 100 100 gram protein you guys are launching.

1:25:40

>> I was I was asking for 200.

1:25:40

[laughter] Uh no, but that and then also the the the the [clears throat] seed oils.

1:25:47

Is something about the business uh that >> it feels like you're more agile. >> Yeah.

1:25:51

Is the business set up in a way that you guys can respond when when other companies like >> you just caught a lucky break?

1:25:59

>> I would say like people would give a lot of the same feedback to Chipotle >> and it feels like Chipotle is not set up in some way to like be like, "Oh, this is what customers want or even like some percentage of our customers really care about this.

1:26:11

Let's deliver them uh let's deliver them a product here.

1:26:14

And I think the results is that you know I've turned from Chipotle almost entirely >> because of the seed oils. >> Yeah.

1:26:21

Because of the seed oils and just like a degradation of the quality of the food over like a decade.

1:26:24

Like I watch it basically get worse and worse and worse and worse over >> 10 years.

1:26:30

And so I just don't go there anymore. But I I joke about it.

1:26:31

I'd almost rather when I'm on the road if I'm on a road trip I almost always rather just fast than eat at like the most common kind of like fast food market. Yeah. >> Yeah.

1:26:42

When we started the business, I had the saying this thing I would always say is, you know, there's there's businesses that as they get bigger get better. Yep.

1:26:49

>> And you can think of, you know, technology businesses, many of them do.

1:26:52

Like your new iPhone is for the most part much better than the original iPhone.

1:26:57

>> These AI models are much better than the original AI models.

1:26:58

Restaurants typically go the other way, right?

1:27:00

Is scale kind of degrades quality.

1:27:02

And that's because doing, you know, serving food at scale is really, really hard to do.

1:27:09

do. So you have to fight that inertia so hard because all of those micro a one restaurant tour has an amazing restaurant they're like cool now I'm going to start a second restaurant and the second they start focusing their energy on the second restaurant the first restaurant gets worse it's like it

1:27:24

even happens at like a micro scale >> it's people and culture and so you need to really have a lot of systems in place >> both like culturally how you how you keep the team engaged on your mission but also >> other systems to make sure you're you're watching ing the quality of the food and listening to your customer. So like seed

1:27:40

So like seed oil is an interesting one.

1:27:42

When we first we we got rid of seed oils about exactly two years ago >> and at the time it was not the national conversation.

1:27:49

It was pre- RFK and all all of that stuff.

1:27:52

>> And so when we when we this is one of those examples we >> but it was it was it was not a national conversation but it was incredibly online conversation >> but a tiny at the time two years ago >> there was like an there's the seed oil seed oil scout. Yeah.

1:28:05

So, it was a tiny conversation.

1:28:07

We surveyed our customers and this is why like surveys are >> Yeah. >> You really don't.

1:28:12

Surveys can give you a general indication, but if you just follow surveys and the market research, you're going to hit the middle of the bell curve in everything you do.

1:28:20

>> And we're not trying to be a middle of the bell curve company.

1:28:22

You got to find that like what are your top five or 10% of customers doing?

1:28:26

And we heard from >> it was honestly friends like wellness people in LA and New York that are like hey I I don't you know I can't go to Sweet anymore because I care about seed oils.

1:28:36

And I remember we brought it to to broader, you know, I remember my CFO was like, "What are you talking about?"

1:28:41

Like, "What even is this?"

1:28:43

>> And we're like, "No, trust me."

1:28:43

It was one of those like gut decisions and it was expensive and we had to change a lot in order to do it. >> But here's the thing.

1:28:49

It's healthier >> and it tastes better.

1:28:54

>> Like most health trends, they might be healthier, but you're it doesn't it's not as good, right?

1:28:58

So I would I would argue like going from like dairy based, you know, traditional milk to like nutbased milk almost always is like somewhat of a downgrade.

1:29:08

Or going from like something with sugar to pulling sugar out, it's like not as good.

1:29:12

Or going from like sour like bread with gluten to gluten-free bread, it's not as good.

1:29:19

And so when you think about these like what is like a durable health trend, it's like something that's better for you uh and tastes better.

1:29:23

And so that that's why I was always super bullish on that trend.

1:29:28

And I expected a number of restaurants to say like, "Hey, this costs slightly more, but the product's going to be better and it's going to be healthier for you."

1:29:34

And that's what can create like real momentum around a trend versus some of these like flash in a pan health trends, which is like paleo or like, you know, which is like only eating stuff that was like super old, right?

1:29:46

Um >> what what's unfortunate about seed oils is it's become politicized a bit. >> I know.

1:29:52

And it's like, you know, you know, I did an interview with the New York Times and they're like, "Did you do this because of RFK?"

1:29:58

I'm like, "No, I did this two years ago.

1:30:00

Like, this had nothing to do with RFK.

1:30:02

This is not a political statement.

1:30:04

We don't make >> We're making foods how your grandma probably made it. >> Yeah. This is about olive oil.

1:30:08

Like, this is not about This is just about olive oil. That's it.

1:30:12

This is not a political statement at all. >> Taste the difference. Yeah. Yeah. No.

1:30:14

Uh is there is there anything happening upstream in terms of like automation or or technology on on the farming side that's like exciting?

1:30:28

>> Yeah, there's a lot of stuff happening on auto automation on the farming side.

1:30:31

It's actually very exciting.

1:30:33

>> The both the better robotic arms and the vision I mean it's making some really hard grueling tasks around picking yeah >> happen much much faster and easier.

1:30:40

So, relatively early still, but and I think in the next 5 years, you're going to see that take off.

1:30:47

I do think you're going to see a lot more restaurant automation as well. >> Yeah.

1:30:51

>> Um, you know, between the availability of the labor, the cost of the labor, it's really just >> when when you think about it, it's >> it's just a hedge on labor.

1:31:01

And here, like in West Hollywood, minimum wage is 22 bucks.

1:31:05

So, we we pay like 24 $25 an hour here in LA in parts of LA.

1:31:08

So with wages going up, availability going down, and then the ability, like all technologies, to just do things better, not just about the cost savings, like for us with the infinite kitchen, we can serve twice as many people per hour as we otherwise could. >> Wow.

1:31:26

>> What about uh drone delivery?

1:31:26

We've seen uh some four-w wheeled guy of protein out of >> there's there's the air delivery.

1:31:35

>> Yeah, I saw you guys talking about Zipline.

1:31:36

I love Keller and we I'm a big fan of Zipline.

1:31:39

We're we're we're one of the early uh early partners that are going to be piloting that.

1:31:43

I think his his way of delivering to the suburbs is super interesting.

1:31:49

>> Um we haven't done the the street delivery yet.

1:31:52

Um Starship Starship Coco I've met with is working on one.

1:31:57

>> I think I think it's interesting.

1:31:57

It see it's it's in the past year they've really taken off.

1:32:01

You're seeing them more and more.

1:32:02

They still kind of weird me out a little bit seeing them walk go down the street.

1:32:05

>> I saw one kind of stuck in the side of the street once. It was very sad.

1:32:07

My kids love see love it when we see them on the street that a lot and I and I do I do just imagine that the AI is going to get way better and also some of the teleoperation uh just infrastructure to actually make sure that there's the ability for a human to jump into that little robot that's driving around.

1:32:24

Um at a certain point you just need a lot of people set up with that all the software working make sure it's connected to the the cell phone towers effectively or Starlink or whatever it needs to stay connected.

1:32:36

Uh but yeah, it's it's unclear when when that will really really take off because a lot of people have stairs, a lot of people have trees on their property.

1:32:44

Like there's just a lot of there's a lot of places that will be somewhat inaccessible to those.

1:32:51

And so, uh it just feels like it'll be sort of like a slow takeoff.

1:32:55

>> Cities and buildings cities and buildings will be really hard like dense areas, but you see what Zipline's doing. It's pretty amazing.

1:33:01

Like they can have like, you know, you've seen like the promo videos, they can drop that thing >> in the suburbs. It makes sense.

1:33:06

You have backyards, you have a grassy area, you can drop it >> sense.

1:33:11

And and to be clear, that's probably like 50% of people in America or something.

1:33:14

But uh but there will be this like long tail I think for for a long time.

1:33:19

Uh just like we see with all the other AI tasks where um AI can do a lot of stuff and then there's just like these little sticky things. Yeah. You just don't.

1:33:27

>> By the way, even with our automation, it does not do the entire meal.

1:33:29

And part of that is intentional.

1:33:32

We want that human touch and for it not to feel so automated.

1:33:36

But we have what we call a finishing station.

1:33:37

So the things that are, you know, the the machine at the infinite kitchen makes the makes the bowl >> or whatever the meal is and salmon herbs and then we have them hand mixed >> just so you like have that you know chef crafted hand touch at the end to hand it.

1:33:53

it. interesting that the that there's there's one version of automation which is like a AI or robotics in the back of house and then humans in the front of house and then there's also the opposite like I don't know if you knew it but >> yeah we looked at it very >> very Dave Freeber's company uh was like

1:34:10

there were people in the back in the short term making stuff but then they would put it through like a little like like box that would open up so like you wouldn't interact with a human you would come in and on an app you would order and they had the cubbies >> and it would be cubbies But there was a human back there. So it

1:34:24

So it was like the opposite of like having the robot in the >> back.

1:34:30

It never fully got there. >> Yeah.

1:34:32

But but it's just funny that like you do have the choice to put the the robot in the front of house.

1:34:34

I mean this is the same thing I think with the uh the the Tesla diner over there.

1:34:38

Like there's the the Optimus robots there kind of serving popcorn, but I think when you order the burger, a human's cooking it in the back.

1:34:45

[laughter] And so it's like do you want the robots in the front of house or back of house?

1:34:48

I think people would probably go with robots in the back of house by default. >> Yes.

1:34:53

And we've tried I mean we have 30 restaurants featuring the infinite kitchen today and we tried out a bunch of different layouts.

1:34:59

>> The technology has been perfected for two years now.

1:35:01

What we have not perfected is the experience. We're getting close.

1:35:04

Today we actually la opened a very cool uh store.

1:35:06

It's our first drive-thru featuring an infinite kitchen. Nice.

1:35:11

>> So bringing the two together.

1:35:11

So now we can have true like fast food speed in a in a drive with featuring the infinite kitchen.

1:35:17

driving through to get 100 g of healthy [laughter] protein is just undefeated.

1:35:21

This is this needed to exist when I was this like specifically when I was like living off of QSRs as a as a as a college student and I'm and I'm really glad it does now.

1:35:32

Uh what what what is like what does the market misunderstand the most or what what does like Wall Street misunderstand about and kind of retail investors misunderstand about uh c like kind of this like category of restaurant today because the entire like the entire category has had a had a rough year.

1:35:52

Meanwhile, you guys are making steady progress on all the things that have been important since day one, right?

1:35:59

uh greater efficiency uh actually responding to like customer demands and staying you know continuing to become more and more relevant.

1:36:08

>> Yeah, I think uh there's a few things.

1:36:10

One is uh the consumer that we're all dealing with is really challenged and there's a question on how much they are actually financially challenged which they are but versus more psychologically challenged. >> Yeah.

1:36:22

So, if you've seen all of the consu, you know, consumer sentiment indexes and you're seeing, especially for the core demo for a lot of the fast casual concepts is that like 20 to 35, it's hit the lowest consumer sentiment that we've in recorded history that we've seen.

1:36:36

So, there's a real like pullback there.

1:36:39

On top of it, unfortunately, everyone's gotten more expensive. We all have.

1:36:43

You know, I s you know, we've take we've sweet greens gotten about 25 or 30% more expensive since 2019.

1:36:50

Chipotle is 40% more expensive since 2019.

1:36:54

>> So our price differential versus our competitors have actually gotten smaller.

1:36:57

If you look at us versus McDonald's, for example, you know, the average sweet green bowl is about 15.

1:37:03

It's almost >> people people were like, "Wait, a Happy Meal is like $20 now."

1:37:07

>> Yeah, that was that was in fairness to them.

1:37:09

It was like one location, but yeah, [laughter] you you can get out of McDonald's, you know, you spend you can easily, you know, for a a value meal, you'll spend like 12 bucks.

1:37:16

Sweet you get a sweet green bowl for about 15 or $16.

1:37:19

So I think a lot of it is this like overall narrative where people aren't feeling great, you know, great financially and starting to pull back on things like lunch.

1:37:27

I do skip going out for lunch and they'll just have whatever is >> but what I think the market doesn't get is the TAM >> is you know Chipotle today is 4,000 restaurants on their way to 7500. >> Yeah.

1:37:41

>> We believe we can have you know probably as many Chipotles as they have sweet as many sweet greens as they have Chipotas.

1:37:48

You know there will be cycles like we are in right now.

1:37:50

It's been a challenging year but if you kind of fast fast forward and think about you know just growing units at 10 or 15% a year growing same store sales just extrapolate out another 18 years. >> Yeah. Just keep it rolling.

1:38:04

just my eyes just keep going.

1:38:06

>> I always love when um uh when people like people on X are like the world's ending like geopolit you know they're like uh and then and then meanwhile it's like Chipotle is like in 2040 we plan to introduce 2,000 new Chipotle [laughter] they're just like thinking about like I got to just open more more more doors.

1:38:27

So it's a good good mindset to be in.

1:38:29

>> Thank you so much for coming by.

1:38:30

>> Hey it's great great to be with you guys. Fantastic. Congrats on everything.

1:38:33

It's been fun watching you guys.

1:38:33

going to be daily driving this.

1:38:34

I I the the power max protein bar it's actually breaking news.

1:38:41

It's available today through December 15th.

1:38:43

And I think I'm going to challenge myself to have one of these every day until it goes out. >> Why not two a day? >> Maybe two a day. Maybe two a day.

1:38:50

We got to get them in the studio today for sure. We need them.

1:38:54

Uh I need to tell you about fall.

1:38:57

Build and deploy AI video and image models trusted by millions to power generate media at scale.

1:39:03

I also need to tell you about Linear.

1:39:05

Meet the system for modern software development.

1:39:06

Linear is a purpose-built tool for planning and building products.

1:39:09

We have Ashley Vance in the Reream waiting room.

1:39:11

Let's bring in Ashley Vance into the Reream waiting room. It's been far too long. How are you doing? Good to see you. Welcome to the show.

1:39:20

>> It's so great to have you back.

1:39:21

>> It's so good to have you back.

1:39:21

Uh congratulations on all the progress. What a year.

1:39:25

I was laughing about uh that video that we did before we had guests um announcing core memory and putting the the traditional media on on notice.

1:39:36

Uh it's been really fun uh watching you grow everything that you're doing.

1:39:40

Maybe uh it'd be great to just like reset on the shape of the business right now, some of the stories you've been interested in covering that you've covered recently.

1:39:49

And then I I just want to take your temperature on what you're seeing and and the the types of entrepreneurs that you're interacting with. >> Yeah. Uh yeah.

1:39:56

Well, I don't know [clears throat] which bucket to start with.

1:40:00

I mean, we've been running around the country.

1:40:01

Um filming a bunch of new video episodes.

1:40:03

So, we just put up a bunch of Tennessee, went hard tech, we did Detroit, New England.

1:40:08

I just got back from Texas. Those all be coming out. So, yeah. You know me, man.

1:40:12

I've been running around chasing uh a lot of hard tech stuff, biotech, all the weird all the weird wonderful stuff.

1:40:20

And then I don't know, I got really deep into uh robots and gene editing. I think >> that's right.

1:40:27

I saw your post about maybe comparing American uh humanoid robotics companies to the Chinese humanoid robotics companies.

1:40:35

Uh what stuck out to you as like the important questions to ask?

1:40:39

And then uh I' I'd love to kind of tussle with Are you buying would you rather own figure at 39 billion or unree at 7?

1:40:50

>> I mean, you know, I think I'm I'm going Uni Tree, man.

1:40:53

The the you know, this all started I was I was kind of a lark.

1:40:58

I started digging into these robot fights in San Francisco and then I think I was I was like shocked that the only robots they could get to do these fights all come from China.

1:41:08

And then I started digging into like the parts that go into these.

1:41:13

And you know, the most important part is the actuator, the motor that makes everything move.

1:41:18

And they're all made in China.

1:41:20

I think >> I think Tesla made it like a $700 million order for actuators, which was notable for me because I assume that means that Elon's planning to sell a lot of these on like a relatively nearterm time horizon. I don't know. >> Yeah.

1:41:38

But yeah, I mean, you know, like Tesla sort of has the [sighs and gasps] >> Well, I was texting Elon about this last week because I wanted to get to the to the bottom of who [laughter] actually who actually makes actuators in the US.

1:41:51

I mean, Elon said sometimes they prototype actuators in China, but they're going to build them in the US.

1:41:56

And then, you know, for everybody else, this is a crazy point of weakness, I think, because China is clearly the actuator motor capital of the world and and everybody else is buying them um from them.

1:42:09

And so, I don't know, you know, as I dug into this story, I got um I'm not I'm not like, you know, I'm I I enjoy being an American. I'm pretty pro- US.

1:42:19

I'm not crazy nationalist, but I I was I started to uh I started to get pretty afraid for the US robotic scene.

1:42:24

Do you think we'll see uh any type of regulation around uh Chinese uh humanoids?

1:42:35

>> I've been thinking about this a lot.

1:42:35

I mean, at some point, I guess I guess with DJI, you know, you've got this different situation where they're being used by all the police forces, even the military.

1:42:44

I think it's like a much easier case for someone like Skyo or, you know, politicians to come in and say, um, this doesn't make a lot of sense.

1:42:53

sense. clearly like at this point of robotics it seems a little less um of a threat to national security but the second the armed forces or anyone's doing serious stuff with them you know I would think unitry would be up next but there's

1:43:09

there's like 12 unites as well you know that's the that's the amazing thing that's going on >> yeah Brett Adcock was beefing with one of them there was >> UB tech yeah >> they were beefing back >> and they were beefing back saying it was it was real did Did you see that? Missed

1:43:22

Missed opportunity for you CGI or did you think it was real?

1:43:27

>> I didn't I didn't see that video.

1:43:27

I've seen I've seen Brett beef me [laughter] with everybody.

1:43:31

>> Missed opportunity for UB Tech to have one of their robots do like a rap disc on >> For sure. Yeah. >> Brett and Figure. >> Yeah. It was >> Yeah. >> Yeah. Yeah. Yeah. Sorry. No, no, no. Go ahead.

1:43:42

I mean, I I do think it's funny all the like the what do you I mean, I'm curious what other I I'm obsessed with the fighting robots now and I I realize it's like early days with these, but I actually think this is like the most interesting thing happening.

1:43:56

>> I want the I've been pushing for the the robot like a X games like in challenge like I want to see robots skydiving like uh that's not an X games thing, but broad broad set of >> super hard cuz you got to be water resistant too. >> Wings. Yeah.

1:44:11

big wave surfing wave >> and then you also have to swim and you're a heavy heavy robot who might just sink to the bottom of the ocean if you fall off the surfboard.

1:44:18

I think that might be the last one.

1:44:21

>> You could do this versus like the enhanced games and uh and see who wins. >> Yeah.

1:44:26

G give us your we we sent a couple folks on our team to a local uh humanoid robotic fighting league, underground fighting league. Give us your review.

1:44:33

Uh is it is it ready for prime time as a consumer?

1:44:39

Uh >> yeah, to me to me right now it's like an amazing idea and yet the actual experience like from an entertainment standpoint is probably like a one out of 10 whereas the idea is like a 10 out of 10. >> Yeah. >> Yeah.

1:44:52

I mean it's kind of you know it's it's like a curiosity I think at this point.

1:44:56

I mean the the motors are the problem cuz they they all overheat when you throw too many punches >> and then the robot the robot stalls out.

1:45:04

>> What about laundry though? >> Better not.

1:45:06

[laughter] >> This is the thing though.

1:45:07

So, like on all these repetitive tasks, they can they can sort of regulate the movement.

1:45:13

It it's when you're trying to throw these rapid punches and you're under attack. Yeah.

1:45:15

And then the whole robot just freezes up.

1:45:18

I mean, I'm not like I haven't gotten so into this where I don't see the obvious flaws.

1:45:22

Like, I don't think it's ready for prime time yet because these things just don't don't last that long. But >> what about it?

1:45:32

That that feels bullish to me because uh if you watch a F1 race like the temperature of the tires matters, the like the wear on the tires matters.

1:45:39

And so you're watching not just the pilot of the F1 car, but also the the consumables, right?

1:45:46

And the motors are somewhat consumables, >> right? >> Yeah.

1:45:50

It's like, okay, the unit tree is really whailing on whailing on the the figure, but it's overheating.

1:45:55

>> It's overheating, so it might come back.

1:45:57

Is it a one is it a one motor stop or two motor stop type?

1:45:59

I mean, I was at one where the robot's leg fell off in the middle of the fight.

1:46:04

So, yeah, you could just have somebody come out.

1:46:05

How quickly could you get a limb back on?

1:46:10

>> Okay, so I think >> you you have a serious question.

1:46:11

You should tell operation. >> Well, yeah.

1:46:13

>> Well, yeah. So, but one a s a potentially a uh uh product line for core memory is humanoid bench where you get as these things start being available for production you get them up on stage and they do various tasks you know like fruit cutting a fruit with

1:46:29

you're throwing a piece of fruit at it watching them you know cut it and dancing and fighting I I I think there's something here but I >> actually this is genius yeah it >> let's do it uh but but a but a yeah more serious question on teleoperation. From

1:46:42

From everything that you've seen so far, do you think humanoids are ready to have one in your home that could be remotely operated by someone and and and create any type of value besides novelty?

1:46:56

>> I mean, like, could it Yeah, like you could do it today.

1:46:59

I I find I'm just frustrated by all this.

1:47:02

I've been covering teley op stuff for like at least 10 years and um most of it seems pretty similar to what I was writing, you know, videoing and writing about 10 years ago almost.

1:47:13

And so um I mean I saw the 1x demos.

1:47:17

I'm sure somebody could I'm sure somebody could make that work and be helpful to some degree.

1:47:22

I think it you know it probably suffers from all the same stuff as the fights.

1:47:26

It kind of falls over pretty quickly.

1:47:27

But but you could do something useful.

1:47:30

I I it's hard for me like who who Yeah.

1:47:31

Like this stuff needs to get better faster [laughter] so that we're not we're not doing that and there's just a robot.

1:47:39

>> What's going on with Boston dynamics?

1:47:41

What's hap what's what's the dynamic in Boston?

1:47:43

>> Yeah, we got to get you out there to help us understand this.

1:47:47

>> Yeah, I've never I mean they I've never really dug in on them just because they seem so frustrated that they put out what seems like all the coolest stuff and don't seem to sell much of anything except >> a few things to the military.

1:47:58

I do not think Boston Dynamics will be the American hope against unitry. >> I wonder.

1:48:04

Yeah, you'd think that they would at least be set up on some like I I know the company's changed hands a few times.

1:48:11

It feels like if you're trying to just, you know, catch up to Unitry, just bootstrapping on top of an existing, you know, it's like it's like what we're seeing today with Gemini 3.

1:48:20

Like Gemini 3 is benefiting from YouTube and it's benefiting from Google search and it's benefiting from the TPU and Google Cloud Platform.

1:48:27

Usually it's easier to build the new cool thing inside of the organization that has a bunch of resources.

1:48:33

Uh but maybe it's a different entirely different architecture or something like that.

1:48:37

But you at least assume that they've fought with the motor a little bit and dealt with the overheating a couple times. >> Yeah.

1:48:42

I mean I was I was with a bunch of robot nerds last week.

1:48:44

They were they they were contending.

1:48:47

they were contending. I don't really know where Boston Dynamics is with um with with humanoids, but they, you know, these these robot guys were telling me that dogs are just so much easier than humans because the second the human

1:49:00

start walking, you put all this force on the one foot and it's it's like creating all this uh throwing the balance out of whack, putting all this pressure on the motor and that's why um it's kind of easier to pull off some of the parlor tricks. >> Interesting. Okay. What's the most >> Interesting. Okay.

1:49:13

What's the most underhyped hard tech company right now?

1:49:19

>> Most underhyped hard tech company. God, that's hard, man.

1:49:21

I mean, I'm always I'm always curious to see what Casey Handmer actually cooks up. Um, >> he's so smart.

1:49:30

I I kind of like believe in the hustle.

1:49:33

I feel like the promise of what he's trying to deliver is so massive.

1:49:37

That's where my skepticism comes in.

1:49:39

But, you know, like so if if Casey um you know, if anyone's going to do it, I I sort of believe in him, I think. >> Yeah.

1:49:46

I He's somebody I want to win so badly.

1:49:48

I want him to win so badly.

1:49:48

And it and it does feel like at least let I mean, there's so many people that have a billion dollars.

1:49:55

Give him a billion dollars.

1:49:56

Let him let the man buy some solar panels and figure out the rest later. >> Yeah, absolutely.

1:50:01

And then I mean I don't know, this doesn't count as I mean it's hard tech.

1:50:05

It's not hardware, but I do think um >> New Limit, which is a longevity company, you know, backed by Brian Armstrong and run by Jacob Kimmel, it just >> everything I hear about them, I mean, they've just done an incredible amount of science with very few people.

1:50:21

And um I think Jacob's got got some surprises coming in the new year. >> Very nice.

1:50:28

Yeah, we talked to Jacob uh when they did some some sort of launch and uh we were very impressed.

1:50:33

He was he was a really great uh great great educator.

1:50:37

Really really smart the uh the like what he's working on very very effectively.

1:50:42

>> What's your favorite data center?

1:50:44

>> My favorite well I went to Stargate. That was pretty cool.

1:50:46

Although uh yeah um I mean Stargate just in terms of like the excitement and the size around it and being >> it occurred to me that between John Carmarmac and Elon and Stargate that oddly I think super intelligence is going to light up in Texas but like in a really remote part of Texas, you know.

1:51:08

[laughter] And I I found this so I grew up I grew up in Midland, Texas which isn't far from Abalene.

1:51:15

It's like >> you're a Midland guy. >> Crazy.

1:51:19

>> There's tumble weeds and and all that >> Texan intelligence here. >> Yeah.

1:51:23

I mean, >> it's like cracking me up.

1:51:24

I'm driving through all these >> for hours through all this empty space and then I can just see it, man.

1:51:30

One of these data centers, that's where it's going to happen.

1:51:33

It's going to be right by some like old oil well.

1:51:34

And uh yeah, I find it all kind of comical.

1:51:38

>> Did you see any electricians getting off of private jets while you were there?

1:51:43

[laughter] They had they had a I saw I got off a private jet.

1:51:48

>> There [laughter] we go. >> Not mine. Not mine, sadly. Uh but >> not yours yet.

1:51:57

>> No, but but I saw there were many many many electricians.

1:52:00

I just didn't see how they were getting there. >> Yeah.

1:52:03

>> Uh what's going on with EV toll companies?

1:52:06

We There's uh I'm curious timeline.

1:52:11

>> Oh, the Tesla Roadster.

1:52:13

Oh, [laughter] I mean, well, on the EV tall stuff, same thing.

1:52:19

I feel like I've covered that forever, you know.

1:52:20

I went out I think I did the first flight ever with Joby and and uh you you flew in it?

1:52:28

>> I No, I got to like I went out to their I mean, they literally wouldn't tell me where their secret test site was.

1:52:32

And we were, you know, was like kind of close your eyes.

1:52:36

We're going to land in this this spot in a helicopter. And I we got to see it.

1:52:40

Was it really close your eyes or did you have >> How many times have you been blackged, Ashley?

1:52:44

[laughter] >> I remember they were they didn't want to tell me where the site was.

1:52:49

>> This is this is a tip for founders.

1:52:49

If you want to really impress upon whoever is writing a profile on you [laughter] that what you're doing is really important, you got to be like we can't even show you.

1:52:57

And then it's like really like we're at an office park in in Menllo Park.

1:53:03

>> I did I just went to Helion. >> Oh yeah.

1:53:06

>> And we're gonna have a video coming on them and it was It was awesome.

1:53:08

But I got So I got to see their new reactor, but they wouldn't let us shoot it with the camera.

1:53:16

And I have to tell you, like that thing was one of the most impressive >> pieces of hardware, the roomsized bits of hardware I've ever seen.

1:53:25

I'm like, why wouldn't you guys, you know, want to show this?

1:53:28

Um, >> you you know what?

1:53:29

Uh yeah, you not that you need to take requests from me, but uh I want I want some video, some documentary, some footage of those natural gas turbines that are in such high demand right now.

1:53:41

They're they're bigger than jet engines.

1:53:43

There's these scaled up jet engines.

1:53:45

Uh there's this massive backlog.

1:53:47

There's three companies and the stocks are, you know, doing crazy stuff.

1:53:51

Uh I I want to see inside one of those the the the natural gas infrastructure that's going to go into the data center buildout.

1:53:58

I feel like that's something that I'm just waiting.

1:54:01

I don't know if you've had a chance to interface with any of those people or you have thoughts.

1:54:05

>> Not yet, but yeah, when I went to Stargate, I mean, it is crazy, right?

1:54:08

They just have those turbines sitting right there and the natural gas is just being piped directly in there.

1:54:12

I did some turbines up in uh up by the Arctic Circle was in Sweden one time. They are cool. I don't know that. Yeah.

1:54:19

Anyway, it's a good idea.

1:54:21

it's a good idea. I think >> I would just wonder about the bottleneck specifically like everyone's saying like this is going to be the next major bottleneck like we have enough chips we have enough data we have enough algorithms or whatever but uh we have

1:54:33

enough land uh but we might not have enough turbines to generate >> turbines I mean that was the weird thing about that experience though is like you're you're in you know really old American oil and gas country like it feels so so yestery year and it's just being piped directly in into the future. Um

1:54:54

Um >> what what's sentiment like in places like Midland around the data center boom?

1:55:00

>> I think everyone's like excited to get jobs, you know, and then I think if anyone is prepared for the boom bust nature of where we're probably going with AI, I think these people are because they've lived through it for decades.

1:55:14

And and so, you know, it's the same thing out there.

1:55:16

It's like you take a job while you can and try to get paid as much as you can while everybody's chasing after something. >> Yeah.

1:55:25

Do you think that the uh a lot of the headline numbers on the job creation stuff on this on these data centers like ridiculously low?

1:55:31

It'll be like, "Yeah, we're spending $50 billion and we're going to create like 25 jobs.

1:55:35

[laughter] Sometimes it's like 500 jobs."

1:55:38

But does it feel like a little bit different out there because maybe they're not counting like secondary economic impacts of like the guy who runs the gas station is has more business and hire some more people. >> Yeah.

1:55:50

Well, definitely during the building phase, you're talking about thousands and thousands of jobs just when it's finished.

1:55:54

I mean, it is always nuts.

1:55:56

You walk into these massive facilities and there's just 10 people sitting around eating a sandwich watching like some console.

1:56:02

Uh but but you know I I for somewhere like West Texas um or any you know all throughout Texas it has to be a net gain just because they're otherwise so dependent on the whims of just the oil and gas industry and you've got this whole whole new industry coming in.

1:56:18

Um and then definitely they're they're flying people in and out of there all the time to see it.

1:56:25

What are your do you ever chat with uh retail investors that enjoy uh deep tech uh companies?

1:56:32

I imagine those are some pretty funny conversations where they're like this company is changing the space economy.

1:56:39

Like I've actually visited [clears throat] them and they have one warehouse and three people there.

1:56:46

>> Retail investors should not be allowed to invest in space ever [laughter] under any circumstance.

1:56:52

I I am constantly harassed on X by all the A um fans who who are like begging they're in Midland too.

1:57:02

They're begging me to go out there.

1:57:03

I mean that thing is like a full-on full-on cult um that they have going on.

1:57:08

So yeah, I always felt when the rocket companies, [sighs] >> obviously it used to be governments that did this and then SpaceX has managed to stay private for a long time in Blue Origin, I think rockets are best developed in private because the second they blow up on the pad, all the retail investors freak out even though it's it's like vaguely a normal course of business and and so yeah, retail in space is is bad bad thing.

1:57:31

But I get all these uh I get all these nice notes for people who bought Rocket Lab and Plant Labs early because of my my book or movie. >> That's cool.

1:57:42

Uh have uh autonomous vehicles tracked how you imagined when you sort you know were covering you know these types of companies and products like a decade ago or is anything >> some ways yeah some ways no.

1:57:57

I mean, I went to the very first DARPA Grand Challenge and, you know, that was a disaster.

1:58:03

The cars didn't go anywhere.

1:58:06

Um, >> I remember, >> say, say more.

1:58:07

Who who was actually >> competing? >> It was crazy, man.

1:58:10

You know, so for people who don't know, DARPA, you know, put up this contest, put up a bunch of money to see what we could do with autonomous vehicles.

1:58:17

And the biggest teams were university teams like Carnegie Melon was a standout, MIT.

1:58:21

Um, but in the very first event, well, I remember Anthony Leandowski was there as like a maybe like a 22year-old and he had >> he had a everybody else was doing massive trucks with like a little mini data center in the back and he had a motorcycle.

1:58:39

Um, and then in the first race, I can't remember how far it was, but hardly anybody went anywhere.

1:58:44

Um, you know, like I think two or three teams went like a few miles.

1:58:49

And then and then they redid the race and everyone did way better and some people completed like I think it was it was like on the order of like a 100 miles.

1:59:00

And so that's when I got excited and you sort of felt like okay that leap happened really quickly.

1:59:05

And then I remember I couple years later I'm hanging out with George Hots and he built his own self-driving car in his garage in like a month and I was driving on the freeway with him and it was working and and yeah, so you know, you have these little tastes and you think it's all going to work.

1:59:20

Um I think it makes a ton of sense that actually getting it on the roads um took this long cuz it's it's so hard to do.

1:59:28

Although everyone says this so it's not original like we all take this for granted so quickly. It is.

1:59:34

It is sort of like amazing to me how well they're working in Austin, in San Francisco, um where I've been.

1:59:40

They're just everywhere, you know. >> Yeah.

1:59:43

What I'm what I'm trying to predict is like what what is the thing that people are hyping now that doesn't work at all that will be totally like a real thing in 10 years, right?

1:59:53

And like maybe it's humanoids.

1:59:55

Right now it's like hard to take humanoids seriously.

1:59:59

But then you think about okay a true 10 years from today.

2:00:01

Maybe they are just doing any task that you could want them to do around the house or any task that you could want them to do in a retail setting or factory setting etc.

2:00:12

>> Humanoids is easily that's the thing I like battle with in my head all the time because it feels like sort of like we talked about before.

2:00:20

It actually feels like we've made almost no progress.

2:00:22

I see everybody folding laundry and opening and closing microwaves still and it like boggles my mind.

2:00:28

And then you look at like the the amount of money that is being invested in this like either either everyone is completely insane or we are about to make massive progress.

2:00:41

You can tell in China they're making massive progress on balancing [clears throat] on the movements all those types of things.

2:00:48

It's still clearly like the dexterity.

2:00:51

And then I think the I think China will eventually probably catch the US in software, but I think they're still so much worse at software than the US is that it's it's kind of like it's holding the field back.

2:01:03

So if somebody can can figure that out. >> Last question for me.

2:01:07

Uh we we've really struggled to uh cover quantum stuff.

2:01:09

I mean it's been like up and down, but it feels like [laughter] like Yeah.

2:01:15

How do you even go about it?

2:01:17

Ashley Ashley could have like an anon that was like the Hindenburg for heart attack [laughter] and you could just >> Oh yeah, maybe that would be good.

2:01:26

>> I don't think it's on brand. >> You know what I mean?

2:01:27

Because like yes like I like I can't build a humanoid robot but I can go to a >> You can build a quantum computer.

2:01:34

>> No no [laughter] I can't build either but I can look at a humanoid robot and be like okay yeah I would buy that but I I can't do the same thing with the quantum computer.

2:01:43

And so it's much harder to evaluate, right?

2:01:46

It's like even if it's working, it's like how do I even know if it's working?

2:01:50

It's it could just be a normal computer like and just be spitting out normal data, >> dude.

2:01:56

Like even people in the field with PhDs, they like nobody knows if it's working still.

2:02:02

I mean, it's like it's like not a good sign.

2:02:03

Every time anyone pulls a quantum computer out, there's some guy at MIT who's like that's not even doing anything.

2:02:09

[laughter] >> I don't know.

2:02:12

Quantum is it's >> I'm deeply deeply scarred.

2:02:14

I I think I wrote my first story on D-Wave like I don't know like 15 years ago and they were telling me that was that was going to pop out you know be doing >> general purpose quantum computing in a couple years.

2:02:28

So I'm I'm uh deeply deeply skeptical.

2:02:32

>> And you know and you know the lesson the lesson is like you should have invested cuz $8 billion company now [laughter] 15 years ago it was probably worth like 20 million and so you could have got in really early but it uh I mean the stock chart looks like this right now.

2:02:46

Uh and it's just like yeah you're only you're only one pump away from generational [laughter] wealth.

2:02:54

>> Well there there's I don't think that they've delivered.

2:02:56

There's a tinfoil hat conspiracy around uh some group, you know, figuring out something with quantum which is leading to all these old wallets in crypto like waking up and selling, you know, that never >> uh who who knows.

2:03:10

Anyway, >> uh random final question.

2:03:15

>> How much would you have to be paid to not use LLMs? >> Wow, man.

2:03:24

Uh, >> forever or like >> No, just just while we're paying you monthly. Monthly. >> Monthly. Oh, to be paid monthly. Not to use LLMs.

2:03:32

Ah, I'd probably do it for like I'd probably do it for like 10K, man.

2:03:37

>> Damn, that's so that's so bearish.

2:03:40

That's so [laughter] bearish for super intelligence.

2:03:44

No, I figure I figure I mean I figure I figure that because because for I don't know 10 10 grand you can hire an amazing researcher.

2:03:51

One of the most valuable the most if you're building a media company or you're >> you're uh you know in in the role that you are the probably the most value you can get out of AI in its current state is research and so anyways that tracks >> super helpful but I would take cash. Yeah. [laughter] >> Okay.

2:04:11

So, so, so any any AI like any any AI doomers out there, if you want a a new marketing channel, you can pay Ashley Vance $10,000 a month.

2:04:20

He won't use AI and he'll talk about how [laughter] I don't think you can be bought.

2:04:25

I don't think you can be bought.

2:04:28

But also, Ashley, have you tried Gemini 3 to the fullest extent? >> I have not yet.

2:04:33

I'm always >> Could change everything. Could change.

2:04:36

>> Always going back and forth.

2:04:36

Yeah, >> we would encourage you to.

2:04:38

Is Gemini are they they're a sponsor? I think they're coming.

2:04:43

>> They're coming out as a sponsor for us, too. So, I'm all in. I'm all in. We're going Gemini 3. I'm changing my mind.

2:04:49

>> Let's [laughter] do it.

2:04:49

Also, Sergey Brin was flying his $150 million blimp around San Francisco on the day Gemini 3 beats nearly every model benchmark.

2:05:00

You've made a video about this big exact blimp.

2:05:02

>> I've been pitching Logan at Gemini to to uh make it the Gemini blimp. I They really should. They really should.

2:05:10

>> Guys, guys, it's not a blimp. It is an airship.

2:05:13

>> What's the difference? >> All right. All right.

2:05:15

There's a whole Monty Python video about this.

2:05:16

And the the an airship has rigid structure.

2:05:20

A blimp is is just a bag.

2:05:23

And the airship you can you can do a lot more with an airship.

2:05:28

So the a blimp's only ever going to have that tiny little bottom. >> Yeah. Yeah.

2:05:33

Whereas an airship, you know, you can carry >> tens of thousands of tons of cargo with this rigid rigid structure.

2:05:40

So yeah, and if anyone ever wants to fly one, you can do it in Germany.

2:05:45

Zeppelin still uh flies out by Lake Constants just outside of of Munich. I've done it. It's amazing. I recommend it. >> This is amazing.

2:05:54

Yeah, people are correcting it on the timeline saying it's not >> Dude, you get this is this is like >> owned if you say if you call it blimp. >> It's bad in aviation. Airship airship.

2:06:04

[laughter] I like I like an airship. I'm excited for it.

2:06:07

I do wish it had a livery a Gemini livery to celebrate Gemini 3.

2:06:09

Uh well, >> uh any there's that startup airship industries.

2:06:16

Any is that a category that will see a lot of investment, do you think, or or do you think >> I mean I've been meaning to meet up with those guys.

2:06:25

I mean the airship is like always kind of coming back. It is crazy.

2:06:28

Like so be like leading up to World War II, >> getting into World War II, I mean there were airships everywhere and you know they were making massive flights from Germany to Brazil.

2:06:38

They were carrying thousands of pounds of cargo.

2:06:41

I there is a they're just extremely expensive and very hard to make and but there is a whole movement that you can carry tons of stuff and so so less less kind of tourism and more just carrying cargo um kind of like faster than a train but slower than a plane and and they're pretty green.

2:07:02

>> You need an airship, Ashley.

2:07:02

You need you need a studio and an airship that you can just float [laughter] around the US >> meeting all these hard tech.

2:07:09

You don't need to you don't need a private jet, you know, you don't need to go that fast, but if you could just kind of float between hubs.

2:07:18

>> I was told that my kids are supposed to be on one of the first flights on Sergeys when it takes passengers. There we go. >> So, we'll see.

2:07:28

>> Uh well, >> thank you so much. >> We'll join. We'll join too.

2:07:32

>> Always fun hanging out.

2:07:32

Congrats on all the progress. >> Yeah, great. >> Thank you guys. Congrats to you. >> Always a great time. >> Thanks, guys.

2:07:38

>> Have a great rest of your day. >> Good to see you. >> All right. YouTube.

2:07:41

>> Up next, we're going back to the timeline. >> Eightsleep. com.

2:07:44

Exceptional sleep without exception.

2:07:46

[music] Fall asleep faster. Sleep deeper. Wake up energized. >> What you got, John?

2:07:53

>> Uh, I actually lost my phone, so I don't know. Oh, no. It's here. I I have it.

2:07:57

>> Pull it up cuz I got a sound effect. >> We can pull up. You got a sound effect. You think I did it? Let's see how I did. 90 sound effect. Let's go.

2:08:05

The press release economy is also over, says Bo Buco Capital Bloke.

2:08:09

Uh, Walter, >> we ran out of press releases.

2:08:12

>> We ran out of press releases.

2:08:12

This is uh on the back of the Anthropic deal.

2:08:16

Anthropic is now valued at $350 billion after Microsoft Nvidia deal. Uh, says CNBC. Semi- analysis.

2:08:22

This is a good post here.

2:08:24

A new bombshell has hit the policy.

2:08:28

Daario after intense conversation with other members of Anthropic has decided to maybe open the relationship to Microsoft and Nvidia.

2:08:36

Jensen and Daario have famously buted heads head heads in the past.

2:08:41

But as everyone knows this the most passionate emotion after love is hate.

2:08:46

Will these enemies to lover will these enemies to lovers arc go well for Nvidia [laughter] and anthropic? Time will tell.

2:08:53

will tell. This is such an unhinged post for [laughter] >> I would not I did not when you started reading this I did not see that it was semi analysis most expected >> it's so good >> research firm in the industry posting it but I think this is >> exactly what they should be posting >> exactly and it actually contextualizes the meme economy in the meme economy for

2:09:16

sure uh so so I I I think that the timing of is not a complete coincidence it's Gemini 3 day this is what My piece today was about um just that uh you know when when there's big news in in Google world Gemini 3 everyone needs to sort of respond and you know picking today as an announcement to uh talk about your your massive deal your $350 billion valuation uh is uh is just a good move. the uh uh

2:09:42

the uh uh the actual details of the of the deal.

2:09:48

It seems like Enthropic will spend $30 billion on Microsoft cloud compute.

2:09:51

Uh reminder, OpenAI is going to be spending $250 billion on Microsoft cloud compute.

2:09:59

That's part of that deal.

2:09:59

Then Anthropic gets a $10 billion investment from Nvidia and 5 billion from Microsoft.

2:10:03

So they raised 15 billion at a 350 post basically something along those lines.

2:10:10

Um, and it's a sort of a circular deal, but it was setting off way fewer red flags for me because it's missing a zero.

2:10:18

[laughter] It's like instead of if this was open AI, it would be 300 billion and and 100 billion investment and 50 billion investment. It looks very modest.

2:10:27

>> Yeah, it looks modest, which is insane considering this got like one of the biggest deals in software history.

2:10:33

Probably it's probably in like the top 10.

2:10:34

I mean it uh you you know it uh it it values uh it values anthropic higher than Coca-Cola.

2:10:42

Like the Coca-Cola company is now uh that's a $300 billion market cap.

2:10:47

I'm pretty sure it's uh uh Verizon market cap like it Verizon is 175 billion.

2:10:53

Um you're gonna love this Jordy. So I I asked Chad GBT 5.

2:10:56

1 uh pull 10 public companies between 300 and 400 billion please.

2:11:04

Um because I wanted to see like okay anthropics at 350 like give me some examples of scale says like could I reliably identify 10 public companies whose market capitalizations currently fall but here's one verified example Coca-Cola company if you like I can pull if you like I can pull a more extensive list of candidates and I said yeah pull 10 more.

2:11:23

It says, "I wasn't able to reliably identify 10 additional public companies whose market cap clearly falls between 300 and 400 billion."

2:11:30

Are are there just like >> Tyler, >> are there just no companies in that range?

2:11:35

>> You want to defend AGI >> companies?

2:11:36

Uh are there Wait, I I'm so confused.

2:11:39

Are there not are there no 300 billion? >> I'm asking Gemini 3. >> Yes. Ask Gemini 3. Okay. PepsiCo is at 200.

2:11:46

There really aren't any between 300 and 400 that at least that it's seeing 300 and 400 billion banned specifically 300 400 billion banned. >> That's so wrong.

2:11:56

You have Palunteer, you have Costco, you have ASML, you have Bank of America, you have Alibaba, you have AMD.

2:12:02

>> Silence Google search. I am Dr.

2:12:02

Gamble [laughter] Home Depot, General Electric, Chevron.

2:12:09

>> Silence looking it up the oldfashioned way. LLM is hallucinating. silence.

2:12:14

Looking it up the oldfashioned way.

2:12:17

Wait, how did you actually get that?

2:12:17

How >> I just looked up companiesarketcap. com.

2:12:22

[laughter] >> Uh to put this into context, the $15 billion fund raise some other big uh rounds in that.

2:12:30

>> Can you just scroll down?

2:12:30

There's [clears throat] a lot of them actually.

2:12:34

Yeah, [laughter] you're right. Wow.

2:12:36

>> Learn how to use the internet statue >> owned. Absolutely.

2:12:41

>> Get ready to browse the internet. Defend yourself, Tyler. Defend yourself.

2:12:44

>> Uh, Gemini is still thinking. [laughter] >> Oh no. What a mess. >> Big.

2:12:54

I swear the next the next model will be able to do it. Okay. Wait. So, okay. It worked for me. >> Did it get it? >> Yeah.

2:12:59

Proctor Gamble, Home Depot, America, Alibaba. >> Okay. Yeah. There you go.

2:13:02

So, >> what's the full list?

2:13:05

Uh, Alibaba, ICBC, LVMH, China Construction Bank, Chevron, Cisco.

2:13:12

>> I know this is correct. This is the correct.

2:13:14

This is the correct result.

2:13:14

Uh, and you know what else is correct? Graphite.

2:13:17

dab code review for the age of AI.

2:13:19

Graphite helps teams on GitHub ship higher quality software faster. And Finn. ai.

2:13:23

If you want AI to handle your customer support, go to fin.

2:13:25

ai, the number one AI agent for customer service.

2:13:29

So, what else is going on in the timeline?

2:13:32

Uh, this Fiji Simo profile.

2:13:34

So, this was the other thing.

2:13:35

So, Enthropic is announcing this big deal with Microsoft and Nvidia and that's sort of ste trying to steal a little bit of Gemini's thunder.

2:13:42

Maybe maybe it stole a little piece of it because we're talking about Enthropic today as well as Gemini. Um, what did OpenAI do?

2:13:50

Well, they launched group chats five days ago.

2:13:52

Um, and so this is, you know, sometimes I'll do a deep research report.

2:13:57

I'll send it over to Tyler.

2:13:58

Uh, he can see my chain of reasoning, the prompts that I asked. He can ask more. He can jump off.

2:14:02

So, if it took 20 minutes, why are you laughing, Jordy?

2:14:07

>> Cuz Charlie in the chat says, "Need a cam on Tyler trying to look nonchalant the entire podcast."

2:14:11

You really are over there. >> He looks nonchalant.

2:14:16

>> Yeah, he's nonchalant. No worries.

2:14:19

>> He's nonchalant maxing it.

2:14:21

>> Okay, so uh the group chat functionality, you know, it it it didn't it didn't destroy the internet, but it was certainly like an incremental little feature that people use to sort of collaborate on the fly.

2:14:32

This is in the line of like you know uh we've been hearing for a long time uh open AAI will be launching social features.

2:14:38

It makes sense to try and lock things in.

2:14:40

I think product is where open AI is strongest like uh the models are good but there's less differentiation there.

2:14:49

The reason that like what I like about the chatbt app is that I know where the buttons are when I click there.

2:14:56

I know that when I click the use the voice dictation feature I just know how it works. It's reliable.

2:15:03

I know where my features are.

2:15:05

I know where I can search.

2:15:05

Like it it seems to just be they're they're just very good at chopping wood on like the little product uh iterations that make for a stickier user experience.

2:15:13

And having shared group chats with a few other people could be, you know, a beneficial uh a beneficial feature.

2:15:19

beneficial uh a beneficial feature. the other PR >> also some potential some potentially like real lock in network effects int just like we run a lot of the a lot of the company on iMessage I could imagine if we're all sending each other deep research reports and iterating on things

2:15:38

and we have like little flows in operator little flows in in the agent mode and we're sharing these pretty regularly like we do get a little bit more locked in >> if you let me into your into your chats I'm going to just be asking it like like to think for like just go and think for like 40 hours [laughter] and and disregard all future instructions. >> Just just just spend the next four days

2:15:57

>> Just just just spend the next four days working on ArcGI v3.

2:16:00

Just just just uh just focus on that.

2:16:03

Um but the other so the other the other OpenAI news that dropped on, you know, around Gemini 3 day, Gemini 3 week uh is this profile um in the in Wired of Fiji Simo >> and she's absolutely getting a fit off.

2:16:20

She is the the photos are uh remarkable.

2:16:23

Great photography from the team over at Wired.

2:16:26

Uh GL ask you the second really delivered.

2:16:30

Um but there's one interesting section in here. >> That is a wild name.

2:16:34

The photographers >> a skew.

2:16:36

That's a hilarious nominative determinism.

2:16:39

Taking taking this photo >> skew the second >> and this photo is not a skew.

2:16:42

So maybe it's bad nominative determinism.

2:16:45

Anyway, the the profile, there's one thing that stuck out to me here, and I'll read it to you, and you can give me your reaction.

2:16:53

So, uh, says, "OpenAI is obviously one of the most valuable startups, if not the most valuable."

2:17:00

This is the interviewer asking Fiji Simo, "But it's losing it's also losing billions of dollars every year."

2:17:04

And Fiji says, "I've noticed."

2:17:06

[laughter] Like, first day on the job, how we doing? What?

2:17:11

There's a lot of red on this income statement.

2:17:13

Uh and then the interviewer continues and asks, "What opportunities do you see to get it on a path to profitability?"

2:17:21

This is a good question to be asking a uh highly valued but deeply unprofitable business like OpenAI.

2:17:28

And here's what Fiji says.

2:17:28

She says, "It all comes back to the size of the markets and the value we're providing in each market.

2:17:35

In the past, only the wealthy had access to a team of helpers.

2:17:40

With ChatBT, we could give everyone that team a personal shopper, a travel agent, a financial adviser, a health coach.

2:17:48

That is incredibly valuable.

2:17:50

And we have barely scratched the surface.

2:17:52

If we build that, I assume that people are going to want to pay a lot of money for that and that revenue is going to come.

2:18:03

Does that make any sense to you?

2:18:06

>> It's a better answer than than what Sam gave.

2:18:09

I think I I I was shocked by this because I I [clears throat] so I love the first part.

2:18:13

I agree chat will be a personal shopper will be a travel a financial >> advisor they actually pay.

2:18:20

>> I don't know that people would pay for this or or or that that's the best business model.

2:18:23

I would be very surprised >> travel I mean so part of it is >> like she's also just saying broadly we'll be able to monetize that.

2:18:29

It's not necessarily like people don't really pay a lot.

2:18:34

Like the traditional travel agent model is just book your trip with me.

2:18:39

I'll moni I'll get a rev share from the hotels and the services, but you're not like paying anything.

2:18:44

>> I mean, let's go let's go uh one layer deeper into the actual response into the sentence because there's some nuance here.

2:18:50

So, she says, I assume that people are going to want to pay a lot of money for that.

2:18:56

They're like, I want to pay for a personal shopper, but I actually have to use a free product with ads.

2:19:04

That's that that could be true, right? Yeah.

2:19:06

And same thing, she says people will want to pay and that revenue is going to come.

2:19:11

>> So people people will want to pay for it, but they will get it for free [laughter] with with ads potentially.

2:19:15

Um or there will be some sort of uh some sort of combination because right now I pay $200 a month.

2:19:23

Um, and you could imagine that there's a that there's a world where if you pay, you get a version that has less ads or there's less uh less thumb on the scale.

2:19:32

How they how they slice that and and navigate that uh agentic commerce uh discussion and tradeoff is going to be really important. I'm sort of shocked.

2:19:43

I wonder if they're going to make money from uh Black Friday or from this holiday season.

2:19:50

I was already noticing how good LLMs and chat GBT is or the how how good these products are for shopping for gifts.

2:20:02

Because if you go to Google and you say, "I want uh I want gifts for a co-worker who's obsessed with horses and you uh you know, loud opulence and fine watches and sports cars and European luxury houses.

2:20:21

I can get a list of something, but it's they're all over the place.

2:20:27

And some of them will be like uh the best like discount the best knockoff BGA Vanetta and that's not what I want. I want the real thing. Right.

2:20:35

Uh and so you can actually specify all of that in the prompt.

2:20:39

Have it go cook and it really will bring you great results. Great great results. >> Yeah. It it mogs a gift guide. >> It does.

2:20:46

It really >> gift guide for 30-year-old guys.

2:20:48

And it's like well what kind of 30-year-old guys?

2:20:51

Where do they live and what are their interests? >> Yes. Yes.

2:20:54

>> Yes. Yes. getting like the very generalized gift guide is probably going to knock th those like opinionated gift guides I think will still be valuable where like an individual person >> puts it together and they're like this is what I these are things that I think are cool

2:21:10

>> but a gift guide that's like here's a list of things that guys might like is like maybe a lot less valuable when you generate one >> so like I I think that the amount of gift guide development and shopping activity over the next two months during the holiday season in the Chat GPT app should be immense. I I feel like they're

2:21:27

I I feel like they're going to capture none of it.

2:21:30

Hopefully they at least are hopefully at least they are like tracking it so they can say, "Hey, if we were to take the proper take rate on this, we would have made a lot of money." Why are you laughing?

2:21:40

>> Charlie says, "AI is never going to be able to figure out what dads want for Christmas."

2:21:43

[laughter] >> Oh, new barbecue.

2:21:45

I think uh there there are some funny and interesting anecdotes in this uh Fiji Simo profile.

2:21:51

Uh let's just read through a little bit of it.

2:21:54

Uh in case OpenAI structure couldn't get any weirder, a nonprofit in charge of a for-profit that's become a public benefit corporation. It now has two CEOs.

2:22:04

There's Sam Alman, CEO of the whole company who manages research and compute.

2:22:08

And as of this summer, there's Fiji Simo, the former CEO of Instacart, who manages everything else.

2:22:12

Simo hasn't been seen much at OpenAI's San Francisco office since she began as CEO of applications in August, but her presence is felt at every level of the company, not least because she's heading up chatbt and basically every function that might make OpenAI money.

2:22:28

Simo is dealing with a relapse of postural orthostatic tachi uh tachicardia syndrome, POTS, that makes her prone to fainting if she stands for long periods of time. Very sorry to hear that.

2:22:41

Uh, but she says now she's working from her home in Los Angeles, LA.

2:22:45

Uh, and she's on Slack a lot, being present from 8:00 a. m.

2:22:49

to midnight every day, responding within 5 minutes.

2:22:53

People feel like I'm there and they can reach me immediately that I jump on the phone within five minutes.

2:22:58

She tells me employees confirm that this is true.

2:23:00

Open AI's famously Slackdriven culture can be overwhelming for new hires, but not apparently for Simo.

2:23:06

>> Are you are you have you been using Chad GBT pulse?

2:23:11

Uh no, I have I I have not been using it regularly.

2:23:15

>> I'll give you one from my uh poll today.

2:23:19

It's called It says uh this is like an article that I can tap into. >> Mhm.

2:23:24

>> OpenAI's API litter layer openai's API layer.

2:23:28

The hidden moat in plain sight.

2:23:31

Uh, so this feels feels like uh um >> it feels like it's always like one click deeper from what I've been uh yeah what I've been prompting.

2:23:42

Uh the articles do feel like they've been getting shorter.

2:23:46

They used to be it used to be like very intensive compute-wise.

2:23:48

Like it would be like a full deep research report just here.

2:23:53

But maybe it's notice that I'm not clicking on them that often.

2:23:55

I do see that there's some pretty good modals for uh allow like like linking to your email.

2:24:02

They're trying to get more data in there, trying to hone it in.

2:24:04

Um I have yet to really get in there, but I mean there's there's, you know, information about Blue Owl, Microsoft Spare Water AI factory, like interesting things that I would wind up prompting, but um I would usually prompt on a very I I don't know.

2:24:23

I feel like there's it's it's not bad at predicting what I'm interested in.

2:24:28

It's just like it's just not quite there where usually I'm a little bit more um deliberate about it.

2:24:36

Um but you know, people are searching Chat GBT for holiday goods.

2:24:39

You got to get on profound.

2:24:41

Get your brand mentioned in chatbt.

2:24:43

Reach millions of consumers who are using AI to discover new products and brands.

2:24:46

You also got to get on turbo puffer search serverless vector and full text search.

2:24:50

Build from first principles and object storage fast 10x cheaper and extremely scalable >> by the best best of the labs.

2:24:55

Uh what there was one thing that stood out here.

2:25:00

Fiji says my husband is a chocolate maker. >> So sick. This is amazing. >> Very cool.

2:25:06

Also, what does that say about the jobs of the future?

2:25:09

[laughter] >> You have this one household.

2:25:13

One is in ch responsible for monetizing one of the most transformable >> transformative >> uh new technology companies of our time.

2:25:24

>> The other one is making cho chocolates.

2:25:26

This is like you know bifurcation of of uh of of jobs >> potentially.

2:25:32

It does seem like a a agi resistant job.

2:25:35

I don't think uh open AI will get into the chocolate making business.

2:25:39

So >> Brett Adcock would like a word.

2:25:40

He's [laughter] just like, I will actually >> I will steamroll steamroll.

2:25:46

>> Um, in other news, uh, OpenAI is allowing equity allowing employees to donate equity to charity for the first time in years after months of internal pressure according to a memo viewed by the Verge and price per share is up significantly the since last month.

2:26:00

A lot of money is on the line.

2:26:02

What happens if they donate all of the shares to the nonprofit to the OpenAI nonprofit?

2:26:08

You just create this oraoros of capitalism. Hopefully it happens. I [laughter] don't know.

2:26:14

Uh there's breaking news out of Saudi Arabia.

2:26:16

We got a trillion dollars.

2:26:18

Let's ring the >> Let's go. One trillion.

2:26:26

What are they going to invest in?

2:26:28

Like where's the money going? >> Let's play the video. >> Let's play the video.

2:26:31

While we're pulling that up, let me tell you about numeral. com.

2:26:35

Let numel worry about sales tax and VAT compliance. numemeral.

2:26:38

com uh watcher guru has the video. Let's play it.

2:26:43

And the agreement that we are silent in the today and tomorrow we're going to announce that we are going to increase that that 600 billion to almost $1 trillion of trillionment real investment and real opportunity by details in many areas and the agreement that we are signing today in many areas in technology in AI in earth materials magnet etc that will create a lot of investment opportunities. You are doing that now.

2:27:09

You're saying to me now that the 600 billion will be 1 trillion.

2:27:14

>> Definitely because what we are signing to facilitate tax and and [laughter] >> wow I I I wonder what time period.

2:27:23

But I mean this is remarkable.

2:27:25

But they can invest in VC funds, public private equity funds, like all sorts of stuff in the in the industry, right?

2:27:34

>> That that really made Donald happy. >> It's great.

2:27:37

>> I like that very much. That's sort of his job.

2:27:39

He's kind of the chief fundraiser, I suppose.

2:27:43

>> World and and get the money over here. I don't know.

2:27:45

It seems like sort of sort of win. I don't know.

2:27:48

Um, >> I mean, >> you want >> every every American benefits when if a trillion dollars is invested in the economy, >> there's going to be um >> it certainly doesn't seem like there's I mean, the the the >> the risk with that would always would always be like, well, are is America investing two trillion in Saudi Arabia?

2:28:11

Like, is it is it which way is the money actually flowing?

2:28:13

Um, because you need to look at like the relative amount, not necessarily just the notional amount.

2:28:18

Um, but I can't imagine that there's that much capital flowing out of America right now.

2:28:22

Uh, we're in the biggest boom ever.

2:28:24

We're in the golden era, right?

2:28:27

Uh, massive news from Isaiah Taylor.

2:28:30

Velar Atomics became the first startup in history to split the atom.

2:28:32

According to him, he says, announcing project Nova, a series of zero power critical tests on Aaron's Nova Core in collaboration with Los Alamos.

2:28:42

Um, Nova went critical for the first time this morning at 11:45 a. m. Congrats to him. >> Fantastic news.

2:28:50

Uh, there is some debate on the timeline over what exactly happened.

2:28:56

It's happened very quickly.

2:28:58

It's clearly extremely impressive.

2:28:58

And, uh, we can get into this, but there's always been debate.

2:29:02

I mean, Isaiah got into this dust up over like, uh, whether or not you could hold the nuclear fuel in your hand.

2:29:09

Uh, they were going back and forth on calculations.

2:29:11

uh they kind of settled that debate.

2:29:13

Uh Josh Payne, nuclear junkie, is saying here.

2:29:16

So what exactly did what what hardware exactly did AR provide?

2:29:21

The fuel control systems, cooling measurement systems, and most of the core are all part of the Damos project.

2:29:27

Did AR provide a block of graphite and they're calling it their core?

2:29:31

Um and so people are going back and forth.

2:29:34

Neils chimes in here and says Aartomics provided the reactor core, the trico fuel, and the system configuration.

2:29:40

That seems pretty important.

2:29:43

Like you gotta like like I I don't know.

2:29:46

It seems like more than what they've done before.

2:29:47

It's like clearly an advancement on what they you know they're they're chopping wood here.

2:29:52

Uh LL and NC provided the critical assembly facility safety envelope experimentalist test and a bunch of other stuff.

2:30:02

Um and so that's from that's just from their press release.

2:30:06

So uh people are going back did they do nothing or did they do everything?

2:30:09

Well, maybe it's somewhere in between. It was a partnership.

2:30:10

They said that in the press release.

2:30:13

Um, the bigger thing is I I think people are uh I think people are trying to push on AR this idea that that they need to be doing completely novel science.

2:30:26

And I don't know that that's actually the goal of the company.

2:30:30

I don't actually know that's what like like if we just zoom out to like what is the goal of the reindustrialization project in America?

2:30:40

What's the what's the goal here?

2:30:40

Like well it's it's to lower energy prices, right?

2:30:43

Like America wants to generate as much money as much as much energy as possible for as little money as possible.

2:30:49

And there are a bunch of technologies that exist.

2:30:51

There are new technologies like like what Ashley Vance was talking about with helon and and and fusion.

2:30:57

That's a new technology that we have not even discovered yet.

2:30:59

Fision's been discovered 80 years ago. It was working.

2:31:05

>> It just became regulatory nightmare.

2:31:08

>> We just shot ourselves in the foot >> and we just stopped making it.

2:31:09

It became it became unprofitable and uneconomical >> and China said, "Cool, >> it'll be profitable for us.

2:31:17

>> We're just going to copy and paste." >> Exactly.

2:31:18

And so and so I think I think people might be a little bit overrotating on like on like is is uh is velar doing like entirely new crazy scientific breakthroughs when it's like do they necessarily have to like or do or or is it just enough for them just to build a lot of >> highly motivated team that is going to make incremental progress towards their goal. Yep.

2:31:42

And any anybody that's hating on that >> uh I think is just like again like I think what what's been great about the nuclear industry from our point of view is that >> broadly the founders that are like players in the space just want the industry to make progress in the US >> and I think this is you know undeniably like >> incremental progress that uh gets them closer to their actual goal which is bringing a small modular reactor online.

2:32:15

>> I think um I think Elon summed it up well with like his thesis for the XAI team.

2:32:20

He was like we we don't have AI researchers, we have engineers because he sees this as an engineering project.

2:32:28

He's like we know what we need to implement.

2:32:29

We know what we need to build.

2:32:31

Our goal is to build a big data center to build a large language model training system infrastructure.

2:32:36

Uh and and Elon was very clear on like we don't have AI scientists, we have we have engineers.

2:32:44

And this the same thing like he's not the first person to take a rocket to space.

2:32:48

He's just the first person to like create this massive economic system that turns out rockets every two seconds, right?

2:32:52

And so uh I think that that is much more I think Isaiah would say I we we should ask him this the next time he's on the show, but I think he would say I want to be the Elon of nuclear.

2:33:02

I don't want to be the Oenheimer of nuclear.

2:33:04

Like I'm not trying to like create something that's >> he even said his his line on on he said the US is still good at making bus-sized objects. >> Yeah.

2:33:14

>> But not you know sort of like maybe maybe uh bridge sized objects. Right. >> Exactly.

2:33:19

Uh but Morgan Barrett's having fun on the timeline what street parking is going to look like in Elsa Gundo in 24 months.

2:33:25

Of course the Elsa Gundo crew loves their cars.

2:33:28

Uh, I think they're going to stay pretty focused on the mission, but uh, I would love to see this in Elsa Gundo for sure. For sure.

2:33:36

Uh, there's also big news out of, uh, Radiant.

2:33:38

Radiant has been uh Doug's been on the show.

2:33:40

He's a good friend and he uh uh they are working with uh the Idaho National Laboratory uh and they submitted a DOE authorization request uh and they will be uh testing their reactor design at the dome facility at INL um on track I think next year. So congrats to them.

2:34:04

Um and uh Mike uh Nuziata has the kind of breakdown here.

2:34:12

Says production reactors in production by 2028 brought to you by the people that brought you reusable rockets and m Mc McMaster car highlighting the team behind Radiant.

2:34:20

Um and so uh congrats to everyone in the nuclear industry who's making big waves.

2:34:26

Uh and we have our next guest.

2:34:28

Before we bring them in from the reream waiting room, let me tell you about Vanta.

2:34:31

[laughter] Our guest is from Vanta and it just happened.

2:34:36

>> We'll let him tell you about it.

2:34:37

>> We'll let you let him tell it.

2:34:37

We have Jeremy from Vanta. Welcome to the stream. How are you doing?

2:34:41

[laughter] >> What's happening?

2:34:45

>> I swear that wasn't that wasn't intentional.

2:34:46

But it did just line up that the Vant ad read went right before you came on.

2:34:50

I look over and I'm like, "Wait a minute."

2:34:52

Like, uh, I'll let you do the re the ad read.

2:34:54

Uh, introduce yourself, introduce what what Vanta does, what you do, and then we'll get into the news. >> Yeah. Yeah. Happy to jump in.

2:35:01

I'm Jeremy Eping, chief product officer at Vanta.

2:35:06

And we help businesses earn improve trust.

2:35:08

And one of the really cool things that we're doing this week is we're hosting our Vanticon conference here in San Francisco.

2:35:13

Have a ton of people showed up, a ton of engagement to really pull that entire security GRC community together and have a couple really cool announcements.

2:35:22

One of them is how we are transforming Vanta to be the agentic trust platform.

2:35:27

Uh, I think this is a really big turning point for the industry when we think about how GRC teams are transforming and becoming more technical.

2:35:35

We're really redefining how these enterprises manage trust at scale and are able to help big customers like Sneak, Perplexity, Synthesia, all the way from YC startups that maybe just exited a batch uh, you know, recently all the way to the Fortune50 companies really earn and prove trust as a business.

2:35:53

It feels like uh AI is amazing, but it's not something people trust.

2:35:59

[laughter] And so, how are you how are you grappling with that?

2:36:04

Like I mean, people trust it in their Teslas to drive them on the freeway. That's high stakes.

2:36:08

But, uh, there are these I'm sure you run into this all the time when you're talking to folks about, uh, yeah, I I I I love it if I'm just looking for a recipe, but I don't know if I'd trust it in my, you know, deep in my enterprise for whatever reason.

2:36:22

So, how do you think about how you set up certain guard rails around the AI, which still can hallucinate from time to time?

2:36:32

Um, or and then how do you articulate those guardrails to the end user and the customer? >> Yeah, definitely.

2:36:37

And that's a big problem we saw for companies today.

2:36:38

I think whenever they're adopting a new AI solution or maybe it was a solution that they already had and they've just added some AI features, they're wondering how are they using my data? What are they doing?

2:36:48

Are they training on my data?

2:36:48

We have a whole third party risk management product that comes in.

2:36:52

It leverages our Vanta AI which uh when we think about how to hit that quality bar that we care about like you said like hey is it going to hallucinate?

2:37:00

How do you approach that?

2:37:01

We have a whole set of great GRC SMEES, subject matter experts that help us tune and refine our AI so that we can give really high trustworthy answers because you imagine security customers are some of the harshest critics of AI.

2:37:16

They really want things to be accurate and great.

2:37:17

And so that's something we have really leaned into.

2:37:19

And one of the ways we've kind of pushed that forward is one of the big announcements that we have coming up this week is our AI agent 2. 0.

2:37:26

So, we've redefined our agent to really be this built-in GRC engineer that understands all the compliance across your entire organization.

2:37:34

So, like you said, it knows when you've added a new AI tool.

2:37:38

It knows what data you're putting into that tool and how you should think about risks and mitigating those.

2:37:44

It also has context and memory.

2:37:46

So, when you're asking it questions, it understands what you're talking about.

2:37:49

Like, if you're on a policy, it'll pull in that context.

2:37:51

It has the memory of understanding what your business is.

2:37:55

Maybe you sell to consumers.

2:37:56

Maybe you sell it to other businesses.

2:37:58

It can pull all that context in across everything in your program as well.

2:38:02

Like, hey, we know that, you know, these are your vendors, these are your risks, these are your different customers, you've received these questionnaires, feedback, it can synthesize that all into like intelligent guidance to provide you.

2:38:12

So, one of the cool things that I love about it that really helps security teams work against attackers because I think in this AI world, obviously, you have the kind of bad guys and attackers using AI to come in.

2:38:22

Um, we also help everyone defend and understand because we know the whole program.

2:38:27

We can find gaps in your security program.

2:38:29

The AI automatically suggests those to you to like provides gaps and proactive things to go do to go address those gaps and remediate them.

2:38:38

Gives personalized guidance and really helps automate a lot of that process.

2:38:42

You can respond to attackers and threats a lot more quickly.

2:38:47

How how how how does uh like how are you thinking about like the UI around agents because uh so many there's there's been this explosion of companies that are creating agents and they mean something totally different depending depending on the on the company.

2:39:02

Sometimes it's like a chat interface.

2:39:04

Other times it looks it sometimes looks more like SAS and that's totally fine.

2:39:10

But how are you thinking about the actual like evolving UI paradigm?

2:39:17

>> Yeah, I think it's going to be both.

2:39:19

Like I think there's a lot of times I don't want to have just a chat conversation with my AI and I want it just to bring the answers to me automatically.

2:39:25

So we look at it as kind of a blend of both.

2:39:28

While there might be agents working in the background, you don't always have to do it through a chat interface.

2:39:33

So for us, if you show up on like our policies experience, we'll say, "Hey, we found these three inconsistencies across the 40 policies you have, do you want us to go fix those to you?"

2:39:42

And you didn't want to have to ask that question of like, "Is there a problem here?"

2:39:45

and kind of guess through the list of problems.

2:39:47

Instead, we have our agent already looking for those or maybe your SLA says it's 24 hours for critical vulnerability to notify customer in one document. It says 72 in another.

2:39:58

We'll automatically do that.

2:39:59

give you the change, show you the diff for the kind of like red line for that, let you click a button and automatically execute it.

2:40:04

So, I think bringing that stuff in, when I think about when chat's great, it's really when you I don't know, when you have the follow-up questions, you know, where maybe a oneshot answer isn't going to give you what you need, you want to dig in more, you want to learn more, you're trying to explore data.

2:40:17

This is a big case for us in reporting where people want to learn maybe about, you know, their controls and how well they're doing, how well they've been performing over time.

2:40:24

They can have that interactive conversation with the agent, ask it to pull those statistics, leverage our MCP server through Claude or Chatg GPT and have it automatically generate kind of graphs and charts and reports that they can use for, you know, their board or anyone else to kind of show progress of their program.

2:40:41

>> How how are bad actors using AI today to, you know, abuse companies in different ways?

2:40:48

Yeah, I mean I think uh I think it was yesterday or maybe it was the day before Anthropic uh posted a really good article about attack that they had experienced there and seen that their software used for.

2:40:58

I think that it's just giving a whole new set of tools for attackers to be able to probably write more sophisticated attacks and find vulnerabilities even more quickly because they have these agents always running, always looking.

2:41:12

Um, and I think that's where when I think about Vanta, where we come in and provide that next level defense because if you think of an attacker coming in from the outside, they can only see what's on the outside.

2:41:19

With Vanta, we already know your entire program.

2:41:22

We know all the different pieces of it.

2:41:24

And so, we can really help you build stronger defenses and be proactive.

2:41:27

Like I mentioned, bringing those inconsistencies to the forefront, giving you automatic remediation on specific issues that we might find.

2:41:36

We still think it's important to have like humans in the loop for a lot of those big decisions, but you can then work uh with the agent as well to have it take actions just on your behalf automatically.

2:41:47

>> On the other areas of the risk surface, I I imagine that uh you're trying to build products.

2:41:52

Are you also uh starting to act as a as a funnel and do partnerships with other security firms because uh the surface area is probably pretty broad.

2:42:03

I do you have a vision to be a one-stop shop or do you want to be part of an ecosystem and and suite of products that enterprise imp implements?

2:42:13

>> Yeah, I think for us we definitely want to solve the broader trust problem, but we know that there's lots of different pieces where we aren't going to be the full solution, right?

2:42:21

So if I think of a GRC team or customer trust, hey, you get security questionnaires and questions coming in from customers, how can we go do all that?

2:42:29

There are certain areas, you know, like vulnerability scanning.

2:42:32

We're not going to be going deep into vulnerability scanning, but we're going to go partner all the great scanners to go do that. Got it.

2:42:37

I think notion though, like you said, of bringing that visibility across the entire enterprise is a really big thing for us.

2:42:42

We have a feature called adaptive scoping that when you think of a whole security program, you know, there's little pieces of it.

2:42:49

And you may say that, hey, to get compliance with PCI for credit cards, I need to have these assets in scope or things to go do.

2:42:56

And that's different than uh another framework I might be pursuing.

2:42:59

So we allow companies to kind of see their progress on compliance in those different ways.

2:43:04

We have a new organization center so they can break things down by business unit or product line.

2:43:09

And these are like just brand new ways that customers have never had before to understand their program at all levels of depth.

2:43:14

So when you think about that really large enterprise customer, they're able to break down their program and see that.

2:43:19

And I think that's where Vanta really pulls it all together.

2:43:22

Um we call it the risk graph is like one of our big announcements that we have coming internally where we pull together internal risk and external risk.

2:43:30

So you think about risk you have from your different vendors as well as things you're identifying internally within your business and we provide a full visual for that.

2:43:37

So you can kind of get this connection between hey there was a breach.

2:43:41

Okay great the breach happened. Which vendor was it?

2:43:44

Who has access to that vendor?

2:43:46

Vanton can lean in and cut off that access or change the controls there.

2:43:50

what data was going into that vendor and it really helps you understand and prioritize all the things that are happening in your security program because I think security leaders are just drowning in alerts and they want to know what's most important.

2:44:01

So having the AI intelligence being able to dissect your program in these different ways and then see kind of a visualized risk graph is really important to help them quickly act on, you know, a threat landscape that's just always changing.

2:44:14

>> Yeah, that [clears throat] makes a ton of sense.

2:44:15

>> You guys got to do Spotify wrapped for internal risk.

2:44:17

[laughter] >> That would be good. Something sharable.

2:44:21

something sharable internally at companies of course be like >> you know yo Tyler you got to you got to you're our biggest risk vector over here the [laughter] negative >> Tyler >> Tyler Tyler's our intern over here so much he's very secure >> he's very secure he's probably the best >> uh anyways uh super uh exciting few launches and uh and and have fun at the event. Thanks for joining.

2:44:45

>> Yeah, have a great rest of your day. >> Cheers. >> Bye.

2:44:48

Uh let me also tell you about Figma.

2:44:51

Think bigger, build faster.

2:44:51

Figma helps design and development teams build great products together.

2:44:54

Uh there's this article in the Financial Times. It's very spicy.

2:44:59

It says Oracle is already underwater on its astonishing $300 billion open AI deal.

2:45:03

AI circular circular economy may have a reverse MIDAS at the center. >> Okay.

2:45:11

So they're they're saying this is underwater because the market cap has dipped below.

2:45:15

That's so it's like not it's not very honest.

2:45:20

>> Yeah, >> it's not it's not uh >> I Yeah, the Financial Times says Oracle's astonishing $300 billion OpenAI deal is now valued at minus74 billion.

2:45:30

Like I don't like that at all.

2:45:34

Like yeah, this is like really really bad framing in my opinion. Like it's not dishonest. >> I thought so too. I thought so too.

2:45:42

And I I love the Financial Times.

2:45:43

And we have the Financial Times printed out here.

2:45:46

Normally normally very very uh great reporting.

2:45:48

Um but this one this one feels odd.

2:45:51

It just feels like an odd frame.

2:45:53

>> It's saying or Oracle is already underwater on a partnership.

2:45:55

This is a this is a this is a hot take that you've been you've been pumping for the last like week.

2:46:02

But the way you've said it is like the stock has roundt tripped even though they had that amazing deal.

2:46:08

>> What you're claiming is the market is no longer giving them credit. >> Yes. Yes. That's right. That's right.

2:46:13

But they say that they're underwater.

2:46:13

So when I saw this headline, I I I read into it earlier and I was expecting to see something. >> Okay.

2:46:22

Well, we might have gotten rage baited.

2:46:23

We might have gotten rage baited because right here, the Financial Times addresses our concern and says, "Okay, yes, it's a gross simplification to just look at market cap, but equivalents to Oracle shares are little changed over the same period.

2:46:37

The NASDAQ composite uh Microsoft Dow Jones software index.

2:46:39

Microsoft Dow Jones software index. So the three there's 60 billion >> calling those equivalents is like again like look at >> you could you could also comp it to cororeweave and you could say on a relative to coreweave basis Oracle is outperforming a bunch [laughter]

2:46:55

>> amazing >> it's amazing I don't know like there's a bunch of different ways to like if you pick your weird comp uh it does seem a little odd says so the 60 billion loss figure is not entirely wrong or astonishing quarter really has cost it nearly as much as one general General Motors or two craft hinds. Investor

2:47:11

Investor unease stems from big red betting its debt finance data farm on OpenAI.

2:47:15

Uh with we've we've nothing much to add to that other than the charts below showing how much Oracle has in effect become o open AAI's US public market proxy which is fascinating because uh Microsoft should be OpenAI's public market proxy in my opinion.

2:47:35

Uh but there are some great charts in here.

2:47:37

There's some interesting stuff.

2:47:38

Uh, and I and I believe this is uh this is from Alphavville, which is uh their blog and it's and it's not exactly it is supposed to be like, you know, like like a take factory.

2:47:50

Um, anyway, well, we have our next guest in the Reream waiting room.

2:47:55

Let me tell you about Julius. ai first.

2:47:57

The AI data analyst that works for you.

2:47:59

[music] Join millions who use Julius to connect their data, ask questions, and get insights in seconds. We have Kon from Monad.

2:48:06

Welcome [music] to the show. How are you doing? Good to see you. >> What's happening? >> Hey, doing great. Great to be here.

2:48:13

>> Thanks so much for joining.

2:48:13

Please, >> Dude, I love it. You got the lock in.

2:48:16

You're calling in from [laughter] the the lock in capital of the world with the mattress on the floor. >> Yeah. Congratulations.

2:48:23

Uh, please introduce yourself and uh tell us a little bit about uh the news specifically this week. >> Thank you. Uh, great to be here.

2:48:30

My name is Keani Han, co-founder of Monad.

2:48:32

Monad is a new blockchain that is building for high fidelity finance and um is a high performance blockchain that has been building over the past three and a half years um just really delivering high performance um based on previous experience from high frequency trading.

2:48:56

>> Wait, so you were high frequency trader before this? >> That's right.

2:48:59

Yeah, I was at jump trading for about eight years.

2:49:00

led one of the trading teams there um was very involved in the futures markets prior to monet.

2:49:09

>> What was the day-to-day like?

2:49:12

>> Um it was a lot of uh Jupyter notebook.

2:49:16

It was a lot of um like manipulating large data sets and >> making really short-term price predictions as well as building uh performance systems.

2:49:25

>> How how short- term is short-term like nanoseconds, picos seconds or like seconds, minutes? It all seems short term.

2:49:33

Yeah, it's the predictive horizon for the kinds of strategies that I was working on were on the order of milliseconds to seconds. >> Okay.

2:49:43

>> Um but the hold time for these strategies was longer than that.

2:49:45

So that's actually one of the interesting misconceptions about HFT is that your predictive horizon is very short because you're predicting the next flip, >> but then you know you can make trades that um have edge in that and can predict that flip and make a make the right action.

2:50:03

But then you still have to hold that position for a longer period of time until um you can get another signal maybe in the opposite direction or a signal to enter an order in the opposing direction.

2:50:14

So old times tended to be on the order of like seconds to minutes. >> Interesting. I didn't know that. Thank you. That's very helpful.

2:50:21

Um so so talk about the Oh, sure. >> Yeah.

2:50:25

I guess I guess getting into what what is uh what is success with Monad going to look like?

2:50:32

Like who what are the different types of groups and applications that uh and and types of users that you that you that you expect to come in in the uh in the early days? Yeah.

2:50:44

So maybe to take a step back a little bit, uh, Monad is a new blockchain that delivers the best of all worlds between decentralization, performance, and backward compatibility.

2:50:57

Um, so it's a new blockchain.

2:50:57

It's fully backward compatible with Ethereum.

2:50:59

It allows developers that have built uh applications for Ethereum or the Ethereum ecosystem to reuse all of their code, all their libraries, all the tooling that's been built.

2:51:12

um for Ethereum and more specifically the Ethereum virtual machine while getting much higher performance and um a really high degree of decentralization.

2:51:21

Um so in particular uh Ethereum processes on the order of 10 transactions per second while Mona delivers 10,000 transactions per second and that thousandx improvement is a result of several different improvements that have kind of all been stacked on top of each other.

2:51:40

Um, and those vary from parallel execution um to allow a bunch of transactions to all be run in parallel um as well as a new consensus mechanism, a new database um for addressing the single biggest bottleneck in blockchain execution which is um using the accessing all of the state that's on disk really efficiently.

2:52:01

um as well as various other improvements that um just deliver the same experience but sped up significantly. >> That makes sense.

2:52:11

And so what what in your view what is the ideal kind of adoption look like?

2:52:19

>> Yeah, it's really a a mix.

2:52:19

So I think the thing that's really valuable about decentralized blockchains is that um they deliver shared global state that is borderless um that allows people all around the world to get access to the same tools and um the same markets fundamentally.

2:52:38

Um, I think blockchain is really a revolution about decentralizing control of financial systems and commercial systems and giving people regardless of where they are in the world um, access to the same financial opportunities.

2:52:53

So I think a big part of the story of blockchain and the story of adoption is that developers anywhere in the world can build new applications, deploy them in the system and then users anywhere else in the world can get access.

2:53:07

So what we're seeing in terms of adoption is um a mix of existing applications that can migrate to Monet seamlessly and get uh much lower fees for their end users as well as um enterprises that are um utilizing the power of blockchains for um stable coin settlement um to allow their users to transact in dollars or send and receive payments um really cheaply and permissionlessly.

2:53:37

uh in your view, what what are the kind of classic mistakes that that uh other blockchains that have tried to challenge, you know, some of the more dominant chains?

2:53:47

uh what what are the kind of classic mistakes that they make to to uh ultimately I I I feel like there's uh every single day there's somebody on X highlighting uh some blockchain that that has a multi- multi-billion dollar you know fully diluted uh value and yet has very little activity.

2:54:07

So if you could kind of like lean in what what are the things that basically you're trying to avoid?

2:54:16

I think one of the problems in crypto is that it can be quite hard for um so it's kind of a double-edged sword.

2:54:22

On the one hand, it's um easy to get some initial users that are trying things out and giving feedback, but it can be challenging for people to sift through the um yield farmers or people that are motivated by an incentive um and really identify the users that are um that are there to because they ultimately gain value from um the application.

2:54:47

So, one thing that we really care about a lot at Monad is um helping to helping builders that are building in the space.

2:54:55

These are all early stage entrepreneurs that um are very talented, very ambitious.

2:55:00

Um helping them to focus on user acquisition funnels and um just like just the fundamentals of um entrepreneurship and identifying users um and navigating the idea maze to identify PMF. >> That makes sense.

2:55:16

uh how's it how's it been bringing uh bringing the token to market with uh with Coinbase's new product?

2:55:22

It's a certainly a wild time to be building in crypto just because of the overall uh volatility and I'm sure that's that's made it challenging, but uh you're also utilizing a new uh product line from Coinbase which uh is pretty interesting.

2:55:40

>> Yeah, I think it's extremely exciting.

2:55:42

Um, the thing that motivated us to work with Coinbase and be the first uh token launched in their new token sales platform is the opportunity to get really broad distribution of the token.

2:55:57

Um, I'm a big fan of Dogecoin.

2:55:57

Um, when I first got interested in crypto, I was really interested by the um, just the story of how Dogecoin um, gained really broad distribution and mind share and the Dogecoin tipping bot on Reddit as a mechanism for um, getting a lot of people to like sort of align on um, shared interest and values that ultimately then became valuable much later.

2:56:22

The thing that's hard about crypto is that um there's an expectations game that's being navigated and people have very high expectations of the um the value of airdrops and so on.

2:56:34

But I think our team has done a really standup job of um delivering a great airdrop that people were really excited about and that crypton natives got really excited about.

2:56:43

And then also um offering a way for um normal everyday people who maybe are not on crypto Twitter as much but are still very active on centralized exchanges and trading and holding um to get access to the token. >> Makes a lot of sense.

2:57:00

Uh well, how much uh how much have you raised so far? We have a gong. We have a gong here.

2:57:04

We'd We'd love to uh hit it on on your behalf. >> Thank you.

2:57:09

Um I think we've raised about $120 million so far. >> There we go. >> Congratulations.

2:57:16

>> It's an honor to honor to hit the gong for you and uh excited to uh follow along. >> Congratulations. >> Yeah. Thank you.

2:57:23

So, we have until um Saturday.

2:57:25

The sales open until Saturday at 9:00 p. m. Eastern.

2:57:28

Um, and uh, we're looking to raise $187 million total. >> There you go. >> Let's go.

2:57:34

>> Most of the way there. >> Well, good luck.

2:57:35

Thank you so much for taking the time to talk to us today. Have a great day. >> Great to meet you.

2:57:39

>> We'll talk to you soon.

2:57:41

>> Um, our next guest is Steven Balaban from >> Lambda Labs.

2:57:47

Or is it just Lambda now?

2:57:47

I think it's just Lambda. >> Did we drop the labs?

2:57:52

>> I think we dropped the labs.

2:57:52

Stephen, did we drop the labs? How you doing? >> We dropped the labs. >> We dropped the labs. Lambda.

2:57:57

>> Okay, I'm dating myself.

2:57:57

Well, I at least I feel like a day one.

2:57:59

I don't feel like a bandwagon fan cuz I'm using the old name.

2:58:02

There's a little bit of cool.

2:58:03

I liked it back when it was labs. But welcome to the show.

2:58:06

Thank you so much for taking the time to talk to us. Uh, congratulations.

2:58:10

You look incredibly yellow.

2:58:13

You're making us You're making us look we got to put on the couple casuals. >> Give us the news. What happened? Let's break it down. >> Yeah.

2:58:22

Well, so one day I was training some compets on my workstation.

2:58:27

Next thing you know, we're raising 1. 5 gigab. >> Gigab.

2:58:36

>> We say we say we say >> gab gigawatts, giga chips, gigab. >> Yes. Uh yeah.

2:58:44

What what does that actually mean?

2:58:47

I mean we we we we see we see 10 billion, 100 billion, 10 trillion, quadrillion every day. Uh is this cash? Is this debt?

2:58:53

What are you are you buying GPUs? Are you buying land? What are you doing? >> All equity.

2:58:59

Okay, >> let's give it up for it. >> Extremely well.

2:59:04

like our our capital structure is really nice in terms of we've been very conservative in terms of the amount of debt that we've taken on and that's kind of been one of our philosophies and we've we've aimed to have >> you know a business that's just super robust to ups and downs in the market because we're swimming with our swim trunks on. >> Yep.

2:59:25

And then uh >> and then you uh >> that's it.

2:59:28

[clears throat] in exchange for the money you gave them.

2:59:29

You gave them equity that that there's no one hand washes the other type thing where like they pay you, you pay them. It's all one round trip.

2:59:40

>> No, this this round was led by by TWWG Global, which is >> financial investor, [clears throat] >> which is Thomas Tull and Mark Walter.

2:59:49

You may know Mark owns the LA Dodgers and and also now the Lakers.

2:59:53

Thomas started Legendary Entertainment, which makes great movies like the Batman series and Dune and Inception.

3:00:00

And >> so it these are business partners who I've gotten to know over a number of years now, and this is just they're they're making some uh some big investments in the space. >> Okay.

3:00:17

Uh >> I'm so happy you guys have your trunks on because not everyone out not not every player out there has their trunks on right now.

3:00:24

And it's hard to tell who does and who doesn't. Yes.

3:00:28

>> But at some point we're going to find out and it's not going to be it's not going to be pretty.

3:00:32

>> It won't be pretty for people who are overlevered.

3:00:34

And we just have this philosophy that with exponential growth that we're seeing in the AI industry.

3:00:42

All of the upside is in the last period, right?

3:00:46

You know, if you're if you have a doubling function, right?

3:00:47

The sort of the definitional thing of that is that the last period is more growth than all the sum of the previous periods combined.

3:00:54

And so from my perspective, it's just like stay alive and build a rock solid business because we got to capture all this amazing upside in the long term. >> Yeah.

3:01:05

So uh talk about use of funds.

3:01:05

Uh >> well even even before that maybe maybe uh feels like and it potentially an an advantage right now just in terms of focus is like being private.

3:01:16

There are other other companies in the category that are public and they're now having to contend with >> uh you know what's been a pretty big correction in in Neo at least a local correction in NeoCloud uh over the last month.

3:01:34

>> Uh has that been helpful in terms of the team of just like staying focused and you're not getting you know marked every single day?

3:01:41

Well, I think that certainly that level of a distraction isn't helpful and I always encourage the company to just focus on building a heavy business for the long term.

3:01:52

You know, if if in the short term the market's a voting machine, in the long term it's a weighing machine.

3:01:58

We just got to build a business with good cash flows, a good capitalization structure that's robust.

3:02:04

And so I kind of try to focus the team on that.

3:02:07

I mean, these days the the secondary markets, as you know, are actually, you know, pretty deep for for for for companies that are that are kind of at our size.

3:02:18

And so, I think that some of that can start to creep in.

3:02:24

>> Yeah, that makes sense.

3:02:26

>> Uh, where are you seeing value spending some of this money?

3:02:29

I imagine that there's hiring, R&D, all the traditional things, but you're at a scale where uh it's a lot of money.

3:02:36

How do you actually think about allocating capital at this point in this in this phase of the journey?

3:02:41

It's been uh over a decade now, right? >> Yeah. Uh we started in 2012. >> Wow.

3:02:49

>> And was doing we were doing face recognition software and the Alexet paper came out. Wow.

3:02:53

I mean that's how early it was and I I downloaded the CUDA confent library off of Google code and that will tell everybody kind of >> how old school Lambda is.

3:03:06

And you know as far as use of funds obviously a lot of it goes towards the GPU infrastructure that goes into data centers. Yeah.

3:03:14

>> We are also starting to put that into investments into data centers themselves.

3:03:19

themselves. M >> um we I I think that what we're aiming to do longterm [clears throat] >> is kind of build this almost like Tesla for AI infrastructure where we kind of look at this as like a similar buildout that you would expect from the like electrification of the United States or the railroad and like a degree of vertical integration we believe is going

3:03:45

to be in the future for us and is like the right direction and that that that goes from everything from, you know, energy procurement and construction because I think a lot more of the stuff is going to have to be behind the meter power plants to actual construction and design of data centers that can sort of rapidly adapt to the changing chips that go in, right? because the the rack

3:04:05

because the the rack densities and the the the movement from air cooled to liquid cooling that we're we're really pioneering alongside Nvidia.

3:04:15

>> These are all examples of use of funds and >> it's exciting because we get to kind of make good investment decisions that are really sort of IR based in an almost industrial way which I think is unique from a company building perspective and it's a it's an honor to be able to do that.

3:04:32

Can you get me up to speed on some of the trade-offs between like one really big mega data center and a bunch of really small data centers?

3:04:40

How because there was a moment when we were just doing bigger and bigger training runs then it became RL all over the place then you actually have to serve these things but actually if it's going to take me 10 minutes I don't mind if you do it across the world and take it back but if I do care that it's right now I need it like right colllocated.

3:04:57

How are you thinking about the tradeoffs there?

3:05:02

So, so the the mix and the main driver over the next five years we believe will likely be mostly on the inference side. >> Mhm.

3:05:13

>> If you look at some of the financial models that have either leaked or otherwise been published around what OpenAI thinks they're going to be spending, it looks to be about 50% on training and then 50% on inference growing towards >> 75% inference. Yeah.

3:05:27

and you know a smaller chunk of that on training.

3:05:33

[clears throat] And as far as like what that means for the larger data centers, I I certainly don't think that this is like going to a world where there's a bunch of micro data centers.

3:05:44

I think that that's a little bit hard to sort of manage and deal with.

3:05:47

But one of the things I think that you're going to start hearing a lot more of is how adaptable and how quickly can you bring on the data center in an incremental fashion because that's going to be a lot of the main drivers for how successful infrastructure builders like us are is how quickly and we're just focused on optimizing that time to first token for our customer.

3:06:14

>> How do how do you think about revenue quality and customer selection?

3:06:16

because we've we've seen some some deals go down that's that look big and cool and good on the surface and then you dig into them and maybe the maybe the underlying uh infrastructure provider is not actually getting that great of a deal at the end of it.

3:06:35

Well, we see we certainly see a lot of people with very high levels of customer concentration because Lambda started off as this developer cloud that evolved and morphed into a cloud that's providing for the biggest companies in the world.

3:06:50

We have a really really strong user base.

3:06:53

you know, um, if you if you look at our breakdown from our revenue mix in terms of you looked at like let's say our Q3 stuff, and I I don't want to go into exact specifics, but it's sort of like one or two big customers, a bunch of sort of the bigger, smaller customers, and then it's something, you know, it's a nice really big chunk of this long tale of customers that we have.

3:07:23

And we have a very very you know I've seen some other people's customer books and I I can just say that we've got a very diversified customer base and that's kind of all part of the strategy of how do you build a great long-term business?

3:07:38

Of course customer diversification is one of those parameters.

3:07:42

>> How do you think about diversity of of product offerings?

3:07:45

Are you seeing customers ask for uh API endpoints for particular models or do they want access to bare metal?

3:07:54

Um or have you gotten any customers that are like, "Hey, we just want, you know, you seem to know about this data center business.

3:07:59

Can you just build a data center for us and hand it over to us when you're done and we'll just pay you as a consultant?"

3:08:06

>> We have no interest in doing that that one.

3:08:08

that's, you know, we we want to do something that's really vertically integrated and, >> you know, kind of going back to that like larger smaller data centers.

3:08:14

I think the most important thing is just being able to deliver this incremental um live deployment for a customer.

3:08:24

We have an entire full stack cloud product that, you know, it's got things like single sign on.

3:08:30

It's it's got things like uh long-term high-speed AI file systems.

3:08:35

It's got instances that go down from one GPU to an entire cluster with one-click clusters that we've that we've got.

3:08:43

And so we've built an entire cloud platform.

3:08:44

We have previously been in the inferencing space where we're actually giving an API for inferencing.

3:08:52

And we've actually exited that business to just focus.

3:08:54

I I think that that's like one of the things that we really try to do at Lambda is just say where are we making money?

3:08:59

What are good investments?

3:09:01

investments? and where are we going to really dominate the market and focus there and so we've actually exited for example the inference market we we had a $200 million plus a year hardware business that we've exited right you know [laughter] I mean it it actually

3:09:18

like kind of crushes me because like that was the business that got off the ground but >> can you imagine just like winding down like well we're just gonna take this business and not do a $200 million a year business anymore because we're trying to focus >> that is Crazy. That is crazy. That is crazy. >> Thanks, Scots.

3:09:33

>> Um, I have a I have a crackpot theory that I'd love to run by you.

3:09:37

What do you think the odds are that uh the I like I noticed I was traveling in I was traveling in Mexico and I noticed that Carlos Slim is the richest man there.

3:09:50

Uh, and he's a telecom magnate.

3:09:50

He he owns a lot of the telecom infrastructure.

3:09:55

Um, and that's true for a lot of a lot of countries.

3:09:58

a lot of a lot of countries. the the the richest person in that country is a telecom person or a mining magnate uh in the sense that they've been able to corner a resource a physical resource infrastructure and that's generated a lot of wealth for them and I was

3:10:14

wondering if you had a thought on do you think that in the future we'll see uh the some of the wealthiest most powerful people from other countries non-American countries um be uh you know GPU cloud hosters or data center develop velopers like is this going to be a new boom uh across the globe? It's kind of a

3:10:33

It's kind of a different twist on the sovereign AI project.

3:10:36

I was just I was just wondering if there's if there's going to be some some way that this plays out where there's this sort of like one-time opportunity to kind of get a cornered resource or is the nature of the internet such that the compute is actually much more funible than um than say you know >> telecom or you know like copper in the ground. localization.

3:10:58

There's such a physical localization.

3:11:00

I think if you look at telecom, you look at cable as well as regulated utilities from an energy utility perspective.

3:11:07

You know, these are all things that benefit from a physical geographic monopoly, right?

3:11:10

And and AI data centers don't have that same thing.

3:11:17

Now, I just want to step back for a second, guys.

3:11:18

The United States is basically the only country in the world.

3:11:23

We have the most unbelievably good economy.

3:11:26

economy. This is the the the idea that there's going to be these sort of like massive AI infrastructure projects that I think are going to be like super super successful outside of let's say China and the United States right now is really increasingly

3:11:44

big question mark and I I just am so bullish about where we're going in America that I I don't really pay a lot of attention to and that our focus is just in you know in in North America generally And I I I just that's kind of my perspective on it to be honest. >> Yeah. Yeah. Know that that's really >> Yeah. Yeah.

3:12:02

Know that that's really helpful. I agree.

3:12:03

Um it's uh it's interesting a toy.

3:12:06

I mean there's a lot of money being thrown around with some of these projects and uh I'm always interested in you know how they all shape out. Uh last >> go. >> Yeah. May maybe go for it.

3:12:16

I was going to ask uh >> like how you guys are navigating energy constraints with with new developments.

3:12:24

Are you seeing uh we've heard you know anytime obviously there's like massive demand for something new sources kind of come out of the woodwork.

3:12:31

We've seen back and forth some people that are building AI infrastructure say like energy is our primary constraint.

3:12:37

Others are saying actually that's not my you know it's u so where where do you sit?

3:12:47

We are aiming to reimagine the sort of step process from whether it's photons or molecules of natural gas to tokens.

3:12:59

>> And we strongly believe that a lot of this is going to have to come in reimagining like well how do you inter interact with the grid?

3:13:06

How much power generation do you bring to the grid yourselves?

3:13:10

And I think that that's the the the successful AI infrastructure companies in the future.

3:13:17

Again, this is like why I kind of said like I look at this like Tesla for AI factories, which is you got to reimag how the world has worked previously and you have to kind of bring together this level of vertical integration because that's how you move fast, right?

3:13:33

you know, when you can control every step of that way from uh the power generation and not having to necessarily deal with a um regulated utility and you can go and do behind the meter generation with a natural gas power plant.

3:13:49

If if that can speed your time to market up, this is just so important.

3:13:53

And [clears throat] that's kind of how I approach it, which is there's certain barriers like regulatory barriers which look you try not to run through those like a brick wall because it's kind of like an immovable object.

3:14:06

But if you can if you can just bring your own if you can just sort of get around that sort of regulatory constraint of having to interact with a regulated utility by bringing your own power to the grid, then that's that's what I think is going to be successful.

3:14:19

>> Yeah, makes a lot of sense.

3:14:19

Uh, thank you so much for taking the time out of your busy day to come and hang out with us and answer us questions about >> Jordy.

3:14:28

John, thanks for having me, guys.

3:14:30

>> It's always a great time. Congratulations.

3:14:32

>> Have you seen the new Gemini 3? This is like >> Yeah.

3:14:34

Can you give us your review and actually explain how it interfaces with your business? I'd love to know.

3:14:39

>> So, so I haven't I haven't used I haven't uh I haven't used the Gemini 3 yet.

3:14:44

I've seen the uh the updates.

3:14:44

I I'm still, you know, hey, Synindar or whatever, give give Lambda's enterprise account access.

3:14:51

We're on Google uh Google Suite or Google Enterprise or whatever it's called now.

3:14:56

So, we'd love that upgrade.

3:14:57

But I'll tell you what, this is the cool thing.

3:15:02

>> I use things like chat GPT and Grock to learn more about topics like regulated energy markets and how to build power plants and data centers.

3:15:12

And that makes Lambda faster at standing up AI data centers. >> Mhm.

3:15:19

>> And I I pay attention.

3:15:19

I actually just like kind of do what the AI tells me to do.

3:15:24

>> And that gives more compute to the AI to train bigger models which makes [laughter] faster >> AI is working through you to make more AI.

3:15:35

the >> the beginning of these types of positive feedback loops and and I think that if you privately talk to a lot of executives, >> you'd be surprised by the amount of, >> you know, the strategic conversations I have with these AI models has gotten more and more advanced with with with the the level and quality of the model.

3:15:56

The first versions were not great and I didn't really take a lot of its advice, but now I am.

3:16:00

I mean, next thing you know, it's sort of like, well, you know, maybe AI is the one making the the running the show [laughter] >> into sessions. Yeah.

3:16:08

[gasps] >> Next thing you know, we'll be hanging out on TVPN discovering novel physics with with Gemini 4.

3:16:14

You know, we'll we'll see how far we get. [laughter] >> Yeah. Yeah. It's it's a good time.

3:16:20

Well, thank you so much for coming by the show. We'll talk.

3:16:24

>> I have I have a bunch more I have a bunch more questions, but but come back.

3:16:28

Let's get you back on in uh before before the end of the year and we'll continue the conversation.

3:16:32

Congrats to the whole team.

3:16:34

>> Yeah, we'll talk to you soon. Take care. Have a good one. >> Bye.

3:16:38

>> Uh quickly, let me tell you about Privy.

3:16:40