0:35
[Music] Hey, hey, hey. [Music] Heat. Heat.
Search this video
1:02
[Music] [Music] Hey, hey, hey.
1:32
[Music] [Music] [Music] Team Deathmatch.
1:58
[Music] [Music] [Music] [Music] [Music] Heat. Heat. N.
2:56
[Music] [Applause] We came to this world. Reach the stars.
3:22
We came to this world to see a future.
3:33
We came to this world [Music] to reach the stars to shape our future.
3:47
[Music] We came to feel the new [Music] [Music] I see multiple journalists on the horizon. Stand by.
4:35
[Music] [Music] You're watching >> TVPN.
4:59
>> Today is Thursday, September 18th, 2025.
5:02
We're back in the TBPN Ultradome.
5:02
the temple of technology, the fortress of finance, the capital of capital.
5:08
>> Hello everyone, Bobby C.
5:10
>> Long since we have podcasted. >> It has been.
5:12
We had a bit of a dry spell. We fixed it. We're back.
5:14
Uh thank you to everyone in the chat who's committed to never missing a stream.
5:21
>> All of you, Andrew Miller, Bobby Cosmic Parker, Burger, >> good to see you guys.
5:24
of >> we uh yesterday we were at Meta Connect 2025 and they launched uh the Meta Ray-B band displays.
5:33
12 years since the launch of Google Glass. Google Glass was 2013.
5:39
>> Did you ever get a pair of >> I never owned them.
5:40
They were $1,500 at the time. I was dead broke. Could not afford them.
5:44
Uh probably would have bought them if I had the money, but >> I would have cut your runway by like 15%. >> For sure. For sure.
5:50
It would have been rough.
5:52
Um, but I did get to try them in San Francisco at the time and it was it was cool.
5:56
Uh, very small screen, very limited interaction.
6:00
Um, some of the same demos honestly, messaging, maps, music, the usual, but uh, way less stylish. Way way less stylish.
6:09
And and very quickly uh, the world was not ready for the glass hole, the the Google glassware.
6:18
Are you familiar with that term?
6:20
It was uh one of my friends actually I think he got I think he got like assaulted or something or like >> for wearing them.
6:27
>> I don't know if he actually got like punched but he definitely got accosted at a bar in San Francisco for wearing them.
6:33
He was an associate at Andre and Horowits at the time I believe.
6:36
>> Would they be perpetually recording? >> No.
6:38
>> No. So it was the same thing where you would turn them on and record but they had this vibe to them in the like in the world where it was like oh this is like the surveillance state like every there's going to be a camera everywhere and this is before I mean 13 years ago
6:54
people or 12 years ago people were not have their phones out taking photos constantly and so all of a sudden this idea that like you could be in a bar and someone would be taking a video or photo of you like automatically was pretty rattling to people and so it was it was like pretty severe like rejection. >> It's a terrible feeling to be filmed by
7:10
>> It's a terrible feeling to be filmed by a stranger just in general.
7:12
I was at a restaurant a week or so ago >> on a date with my lovely wife and this person was >> basically doc documenting like the entire outside of the restaurant.
7:25
We were sitting at an outdoor table and like repeatedly like panning over us. >> Interesting.
7:32
>> While we're just sitting there.
7:33
>> Yeah, it's suspicious.
7:34
>> It Yeah, it's just it it is a violation.
7:37
Yeah, >> it I I'm all good if I'm in the back of of the picture of the video that you're taking, but if you're like panning over and I become the subject of the video, >> we haven't and we don't >> Yeah, because it could be that you're the rest of the video is just window dressing for an excuse to actually be filming you for some reason and you're like, why why do you want to pick? >> Yeah. What's going on here? >> Just ask. Um, yeah.
7:57
At the same time, I've been on the other side, like we were at uh the airport and I saw this guy absolutely crushing sales calls with a full-on headset on instead of just Bluetooth.
8:08
You know, people wear the AirPods or the wired headphones.
8:09
This guy had a full headset with the microphone down in front of his mouth, just barking orders, locked in.
8:14
And he was a LinkedIn general. >> This is a champion. I love this guy. I'm so into this dude.
8:20
I should have just asked and gone up and said like, "Can I do a professional photo shoot with you?"
8:25
Like I I love I respect your culture. Check out your rig. >> Can I Yeah.
8:28
Can I can I can I check out your rig and and enjoy that.
8:31
Um but I I restrain myself and I didn't take a photo.
8:35
I did I did go up to you and say, "Hey, you got to check out this guy. This killer.
8:38
This we maybe we got to recruit this guy. It's pretty sick." >> I know.
8:41
We we we did take a lap and >> Yeah.
8:44
>> Uh and and kind of investigate the >> Yeah, he had uh he had he had his fingers in a lot of pies.
8:48
Much like Rune, we printed a post today. We got Rune.
8:52
We printed a post of our own quote. >> Yes.
8:55
Run Run says uh Run has got his fingers in a lot of pies from the God whispers of TVPN by John Jordy and John.
9:06
And I I remember saying this, but I don't remember what pies I was referring to. I he has a job.
9:12
>> We don't even know what what it means.
9:14
But it is uh it is certainly provocative.
9:17
>> But I'm glad he enjoyed it and and quoted on the timeline.
9:19
And you know, now it's printed forever. Um, anyway, uh, ramp. com. Save time and money. Time is money. Save both.
9:28
Used to use corporate cards, bill pay. >> It was a whole lot.
9:32
>> Absolutely fantastic watching you at your absolute best doing a ramp ad, right? Zuck was walking up.
9:37
I was watching Zuck walk up the stairs.
9:40
He's feet away from us and you're just ripping. >> Barking ramp. >> Just bark and ramp. >> I love it. Like a carnival barker.
9:46
Of course, yesterday uh the stream was only made possible by Reream one live stream 30 plus destinations multiream and reach your audience wherever they are. Thank you.
9:56
>> We were actually everywhere.
9:56
We were over on an Instagram.
9:59
>> Yeah, we did our first Instagram live stream.
10:01
We're going to do a lot more there.
10:02
Build out a full vertical layout.
10:04
I I think it's going to make a lot of uh distribution a lot easier.
10:06
Get on reals, get on TikTok, all these different things.
10:11
>> Uh glad that everyone showed up for return to form TBPN classic.
10:13
Just uh just a couple good boys hanging out.
10:19
>> TBPN at its best right here in the Ultra. >> Technology Brothers.
10:21
Well, let's go through reactions to Meta Connect 2025. Uh TJ Parker had a post.
10:26
Meta Execs are 20 years younger than Apple's and still willing to do live demos.
10:31
Unfortunately, very bearish for Apple. It's a good take.
10:35
There were a few >> Yeah.
10:37
I mean the the the wild thing uh from from the interviews obviously you could find this out uh from just doing a little bit of research but if you were just watching and realizing how young the Meta executive team is >> and how long they've been working together. >> Yeah.
10:53
Cuz they all were like they were like 22 and they started the company and they joined like most of the people the average tenure of person we interviewed was probably 15 years and they were all like in their 40s at most.
11:03
Um, but the the drama and the timeline was around the failed demo.
11:09
I don't know if we want to play the actual video.
11:11
We were we were reacting to the keynote live, but we were talking and we kind of picked up on one of the demo fails.
11:18
I didn't realize the second demo fail happened.
11:22
Uh, very bold, but >> yeah.
11:23
By the way, we're going to figure out how to do better keynote sort of like reactions. Yeah. Yeah. It's kind of crazy.
11:30
>> We were struggling with the audio.
11:31
>> Like we want to give we want to give you the facts.
11:33
We want to let you hear what's happening in the keynote because that's when the news actually breaks.
11:39
Like they like they don't want anyone talking about the fact that the new red meta ray band display is $7. 99.
11:46
>> We kept almost >> until Mark says that on stage so you got to wait until he says it.
11:50
So you might as well just watch along.
11:51
So anyway, uh bit of a bit of something to practice for future.
11:57
>> It was it was obvious.
11:57
I mean, we were surprised, especially re-watching the keynote afterward, seeing the moments where where the demos failed.
12:05
We were surprised because we got those demos. >> Yes. >> Multiple times. >> Yes. >> And they worked. Yep.
12:11
>> And so we were kind of trying to um uh we were trying to figure out why that might have been.
12:16
I think our our theory for why Zuck's demo wasn't working on stage is that the glasses were simultaneously live streaming. >> Yeah.
12:26
and trying to carry out the regular functionality >> with video call.
12:30
>> So we were we had done the video calling functionality with WhatsApp. >> Yep. >> Multiple times.
12:36
>> Didn't have a problem at all.
12:36
>> Had no problems at all.
12:38
>> Um and so and and the first demo that we did was months ago. Yeah. Right.
12:41
So >> surprised that that uh that happened.
12:44
Uh but also just the like real time live streaming from first person is like very cool functionality. >> Yeah.
12:52
Although that's not something that they're launching. Yeah.
12:54
So, it's very odd that they >> they successfully live demoed a product that doesn't exist or a feature that's not live yet, which is live streaming from the glasses >> and you could tell the OS was like thought that the video was already pulled up. >> Yeah. Yeah.
13:12
>> Because it was streaming or something.
13:14
>> I I think that one of the stories is like they clearly went super hard on weight reduction. the the display glasses.
13:22
Uh the the Orion demo that launched last year, which actually also had a failed a botch demo as well.
13:27
Um >> it was um it was it was chunky but still very thin and light and when you try it on it feels great.
13:36
Um but this is like remarkably lighter like 69 grams.
13:42
Carmarmac I thought Carmarmac's quote was 100 grams for a headset and $100.
13:47
Turns out he was saying 250 grams and $250 which is I mean 69 grams is way less than 250.
13:54
So uh you you really can wear these all day without any serious strain.
13:58
They're like just slightly chunkier than just normal Ray-B bands.
14:02
Um and so let yeah let's pull up the actual de uh demo and uh and and see exactly what happened here. >> WhatsApp video call. >> There we go.
14:10
>> So he's trying to kick off a WhatsApp video call is to boss. >> Yeah.
14:14
>> Well I Well that's what happened.
14:14
This is so bold and you can't see it because we have our there.
14:21
>> I don't know what happened.
14:21
Um, >> at this point he's done this so many times.
14:26
>> Maybe Bos can try calling me again. >> Whatever. They're having fun. >> All right.
14:32
Well, I got a missed video call.
14:34
Okay, there's the the actual video.
14:36
>> And in the top corner when he does the live demos, they put this like massive bubble that says like this is a live demo. This is live. We're not faking this. >> Get out. It happens.
14:46
>> Well, at least, you know, you didn't think it.
14:47
So, anyway, >> it's 4D chess. The next demo. >> Yeah. What do you think? >> Will be pre-recorded. They want to train.
14:54
They want to No, I'm kidding. >> No, no.
14:56
The real 4D chess is that this sets expectations low.
14:58
It makes you go into it thinking, "I'm getting a devkit-like experience and and then they can overd deliver."
15:05
So, if you put on a if you put on a pair of these, you're like, "Okay, yeah, like uh I'll probably be able to listen to some music, but the WhatsApp video call functionality is probably not going to work."
15:15
You try it, it works, and you're delighted.
15:18
>> So, so it's a hard product to like demonstrate, right?
15:21
It really is something >> that you have to experience >> because like even when we were talking to Bos, Bos had the display up the whole time.
15:30
He got a message from Zuck like right as he was walking up. really can't see.
15:35
>> You cannot see the display at all from the outside. >> Yeah.
15:37
Um well, there were some uh there were some good reactions in the timeline to uh the the failed demos.
15:41
But first, let me let us tell you about our newest sponsors, Privy Wallet Infrastructure for every BR bank.
15:49
Privy makes it easy to build on crypto rail, securely spin up white label wallets, sign transactions, and integrate onchain infrastructure all through one simple API.
15:59
Of course, we had the Privy team on the day they uh announced their acquisition with Stripe, >> part of Stripe now.
16:06
>> And uh we are happy that Privy is the official wallet infrastructure provider of TBPN of the Ultradome. >> Yep.
16:15
>> And uh >> and we also have a new sponsor that the sharpeyed viewer might have noticed.
16:21
They jumped in the ticker during the stream yesterday.
16:24
Uh, Cognition, the makers of Devon, the AI software engineer, Crusher Backlog with personal AI engineering team with your personal AI engineering team.
16:34
Uh, we're huge fans of Scott Woo and the team over at Cognition.
16:38
We've had him on the show a bunch.
16:39
>> Organic super intelligence.
16:39
Scott Woo productized it. >> Yes. Yeah.
16:43
He's had some fantastic calls about uh AI achieving an IMO gold medal.
16:49
He's been ahead of the curve on tons of this stuff.
16:50
Obviously went viral with the initial Devon launch.
16:52
Uh, and he's been He saved his partner >> with the when Sururf acquisition.
16:58
>> Oh yeah, that one's legendary.
16:59
>> He saved the social contract. >> Yeah, really.
17:01
He he saved Silicon Valley.
17:03
Uh and so we're we're we're proud to support uh Cognition and Devon.
17:07
So go head over and sign up. They're in GA.
17:10
Uh so let's go to Mark German.
17:10
Uh remember Face ID failed demo, but essentially flawless for users.
17:14
I wish Apple would go back to live, but I don't think it'll ever happen.
17:18
much more reach now and they and they definitely see taking risks for this type of thing as pointless.
17:24
I don't think they're wrong.
17:27
Um but on the flip side Elon Cybertruck demo.
17:32
>> Oh yeah, the Elon Cybertruck demo. That was a crazy one.
17:33
And you haven't heard about I mean there's been a lot of you know varying responses to the Cybertruck but no one's complaining about their winds their their uh their windshields breaking or their >> fortunately you don't have to use you don't have to test your windshields as much as you have to test if your device can handle video calling.
17:51
>> Yeah >> but >> that's true.
17:53
>> Uh but yeah certainly >> so yeah on the on the flip side Shiruya says uh it takes a lot of courage for someone in Zuck's position to go live and make a fool out of himself.
18:03
Everybody gangster in their comfortable controlled environments. It's true.
18:08
And and you can see here the red pill for live demo in red there letting you know this is not faked.
18:13
And there's been a lot of demos that have been sort of faked.
18:16
I mean we we're coming off the the Apple Intelligence where there were a lot of demos that that shipped on varying timelines.
18:22
There were ads that went out and features were promised and Mark German wrote that big article there's something rotten.
18:30
Certino really put his whole reputation on the line to to kind of >> every Apple every Apple focused journalist basically put their >> Yeah. to say anything. It was rough. It it was bold.
18:43
and uh and and then Google had something similar where they demoed an AI uh a real-time video model that would interpret what you were seeing and then um run through Gemini, but they but they sped up the the wait time for the LLM, which some people were like, "Oh, that's not fair."
19:02
Although, you know, with that it's like they're going to speed it up over time.
19:05
Um, but there's always been uh there's always been like, you know, criticism of the various tradeoffs of keynotes.
19:12
And I feel like I mean, this is not as bad as it could have been. It's pretty good.
19:18
And overall, it just felt like being there on the day.
19:21
It felt like we weren't watching a movie or like an ad or a video.
19:25
video. we were like part of a play because like you Mark Mark Zuckerberg and Diplo go running by you and then there's like the skateboarders and there's like this this like demo with the crowd happened in one place and they move they're moving around and like while we were on stage we could see Mark
19:42
going from one thing to the next like actually >> very much participating in like the keynote or the you know the event but in a very like moving around physical >> yeah it's tough it's it's failing failing on stage is tough because no matter matter what we say, right? When I
19:55
When I was using uh when I was demoing uh the display glasses, >> I could say, "Hey, Meta, find me restaurants.
20:05
>> What What are some restaurants that are nearby?"
20:07
And on the heads up display, it would pop up some options.
20:10
I'd be able to select one with the neural band like super super easily. >> Yep.
20:16
>> And and it would just pull up a map and then it would give me perfect navigation. I could just walk.
20:19
It would guide me to it, right?
20:21
And so having this experience like what what they're selling is >> there's a lot of moments that you need to pull out your phone and what if you didn't what if you didn't have to sort of like take yourself out of the moment >> um and uh you know go go like this.
20:33
Uh I think it's very telling that the Verge wrote this morning.
20:39
Uh they said I regret to inform you Meta's new smart glasses are the best I've ever tried.
20:45
Uh the new >> really know your audience.
20:49
>> really know your audience. total audience capture just like some >> I'm sorry guys so tech optimist and such a technative blog and Neilia Patel I believe is like still running it I think he started it and so like I I don't think of the verge as like having this adversarial relationship with tech and yet like they
21:07
had to frame it this way it was very funny uh Dylan Field founder of Figma had a more positive uh take he said congrats to meta to the meta team on today's launches I saw some of the tech while it was still in devel it's still in development the glasses AR Our interface, neural band, and overall capabilities are extraordinary. And yes,
21:21
And yes, live demos can fail. We've all been there.
21:25
The tech is still awesome. Onwards.
21:26
Dylan is, of course, the CEO of Figma.
21:29
We're partnered with Figma.
21:29
Think bigger, build faster.
21:31
Figma helps design and development teams build great products together.
21:34
You can get started for free.
21:35
Um, uh, Ben Thompson also had a positive take.
21:40
He says, "Live demo fails are good tech karma >> because if you get rewarded, >> you get rewarded."
21:46
I mean, for setting for setting the bar low.
21:48
I there is a there is a world it's it's >> you also have to understand like how confident you have to be in a product to go do a live demo for like Zach Zach didn't go out there being like oh there's this is a coin flip like 50% chance >> it's shipping in two weeks it's it's crazy >> yeah and it's just I I can't I can't say it enough.
22:09
It's so painful that like we used this functionality months ago at this point and >> so it's the way it goes.
22:16
Uh well um I I was trying to break down like we we we talked to a lot of the meta team.
22:23
>> Wait, did you did you see the screenshot too of of inside the uh they're doing the Gaussian splatting? >> Yeah. Yeah.
22:30
>> Uh with uh >> Oh yeah.
22:33
>> And you see the Lucy logo. >> This was super cool.
22:35
So 3D gausian splatting is cool.
22:38
And now you can capture real world real world spaces just by walking around in your Metaquest headset.
22:44
We did this demo as well.
22:44
You walk around in the headset, it scans everything.
22:48
They actually make it into this like kind of fun game.
22:49
It takes about six minutes, so you can recreate your room, recreate anything.
22:51
Tyler made a version of this uh just using his phone camera. It looked really good.
22:55
It actually runs in a browser.
22:56
Uh and uh but you can imagine this being really fun.
23:01
But this was such a funny Easter egg that they put um my nicotine pouch brand, nicotine gum brand in the octagon.
23:07
Uh and I don't think this was intentional.
23:09
I think this is literally just they went and scanned one of the octagons and like our logos there, which is cool.
23:14
Um but but fun to see and it made me do a double take in the moment.
23:18
Um the real-time feedback ensures you don't miss a spot.
23:20
Apple needs to get on this ASAP. I agree.
23:21
I at the same time like I was standing in the uh in the octagon and it was like it was cool to feel like okay if I was here how high would it be?
23:31
It gave me a sense of scale but there there wasn't really like a game mechanic to it or anything. >> Yeah.
23:36
To me, to me, this what I brought up the I I just care about what what kind of why would I why would I want to do this other than pure novelty?
23:45
The best example I could think of is is a real estate agent.
23:49
They go they scan a property. >> Sure.
23:52
>> That they're that they're marketing for for lease or sale. Yep.
23:56
>> And then people can just drop in and experience the home. >> Yeah.
23:59
>> Yeah. like actually from >> I see it I see it probably more as like a pipeline to traditional 3D games and like traditional 3D geometry uh years ago decades ago now there was this game called true crime the streets of LA I think it was called and it was like
24:17
Grand Theft Auto nowhere near the level of game mechanics nowhere near as like fun of a game but they did a full scan of Los Angeles and so you could go and find like the TBPN Ultra Dome because it was a street by street rep like replica of of LA. Yeah. And I imagine that that Yeah.
24:32
And I imagine that that if you're able to do gaj and splatting, you could bring that through to a generative 3D geometry pipeline and then create a game out of it.
24:43
Uh seeing how these two technologies kind of merge, I think will be will be pretty important.
24:50
Um, I I had a lot of fun with it and and it was Yeah, it was it was it was cool, but yeah, it's definitely in like the novelty demo.
24:57
I think overall I tried to synthesize like the takes that I was pulling out of the meta team and I had three uh one is just that wearables are underhyped because of how much focus there is on AI.
25:09
And so they I feel like they're all the entire meta team is like very happy to swim in that lane where they're like doing something cool that's not like oh it's got to be fast takeoff got to be god in a box like the expectations are so high agi.
25:25
It's like it's like do they look nice? Are they affordable?
25:30
Do they do the the cameras work?
25:32
Like it's pretty it's pretty low.
25:35
>> Yesterday the the display heads up display is getting more hyped than live AI. Totally.
25:40
>> And and even though every Gen AI product today is not >> 100% reliable and and can have issues, >> the fact that you're just going to be able to walk around with this perpetual intelligence humming along getting to experience everything that you're experiencing, building that context in your life >> is underrated.
25:59
And it says a lot that even in the current state they can run that you get like an hour of battery life for like the truly live AI where it's just processing everything in real time doing the translation functionality.
26:12
>> I didn't realize there was a full hour >> but um the other the other takeaway we were digging into personal super intelligence trying to understand what that means.
26:19
This wasn't really a full AI event.
26:21
This isn't llama con but uh it feels like personal super intelligence will definitely be a personal shopper.
26:28
uh you'll be able to take a picture of something, say, "Order that, it'll go straight to the brand, maybe a Shopify integration, something like that, check out for you, ship it to you."
26:36
Um or if you just see something on Instagram, you're going to be able to fire that off.
26:41
Like we've seen the agentic commerce like outlined in semi analysis.
26:46
We've also seen uh just yesterday or the day before Sam Alman sort shared a screenshot where in the dashboard for your chat GPT app uh one of the tabs was orders and so they're clearly thinking about helping you order things.
27:01
Yeah, it's a huge way to take a cut of commerce on the internet.
27:05
Extremely valuable, extremely monetizable.
27:07
valuable, extremely monetizable. though uh it would be truly shocking if uh Meta was like oh yeah all of our customers every brand that advertises yeah we don't want to continue the relationship in the AI age like obviously they're going to do that so they didn't share
27:22
too much there but it's but uh nothing nothing you know uh nothing was uh shared that felt like that was not on the immediate roadmap it felt like it was and then uh the last the last thing that that I was taking away was that uh basically like lindy pieces of cinema will be the first real killer app for VR. I think your prediction a year ago
27:41
I think your prediction a year ago uh will wind up being correct. Multiple glasses.
27:46
You will have something for action sports that plays in the in the GoPro realm like the Vanguard, the surfing, the the the the uh ski goggles, that type of stuff will be action sports.
27:59
Then there'll be the heads up display that you're maybe using when you're, you know, walking around, driving, doing whatever, answering messages, listening to music.
28:07
Um, and then if it's time to sit down, watch a movie, there will be an alternative to the to the TV on the wall at a similar price point.
28:18
Right now, you can get 65 inch TV probably for like 500 bucks.
28:20
Um, I think you'll be in that same territory pretty soon.
28:25
We talked to uh James Cameron about this.
28:27
Um the like the level of fidelity watching his films Avatar in 3D is just better in VR.
28:35
He was talking about the Quest 3.
28:39
I think he's seen a demo of the Quest 4.
28:41
I think the Quest 4 is going to have a screen that is at the level or beyond the level of the Apple Vision Pro, which was phenomenal.
28:52
But they're going to bring all of their engineering prowess to they're going to take the battery out.
28:58
They're going to make it plastic.
28:58
It's not going to have a screen on the outside.
29:00
It's not going to have all this extra stuff.
29:02
Maybe they'll do a neural band.
29:03
So, because if they do the neural band, the big thing with the neural band is that it takes it's distributing the compute across your body.
29:09
You don't want all the compute right on your bridge of your nose. That is terrible. That's painful.
29:15
So, take the battery, put it in the back.
29:17
You know those are they called croakies? The things like this? Is that right? >> Yeah. >> Yeah.
29:21
So, like I could imagine croakies going over your metaray bands or your or your Oculus where the battery is actually kind of hanging dangling from the back of your head and you're getting extra battery life out of that and and maybe that runs down down your back.
29:34
You throw it in a pocket like what you do with the Apple Vision Pro.
29:38
I think that was a good paradigm.
29:39
I think the Apple Vision Pro it uses hand tracking as input, but that means it has to have cameras.
29:46
I thought it was smart that the charging case for the display like folds up flat so it can act as a as a regular uh case if you have it at the right angle like those kind of iPad >> iPad screen protectors. >> Yeah. >> Yeah.
30:01
Neural band is also uh under hype from yesterday.
30:03
Everybody wants to focus again on like >> the actual glasses and the display.
30:06
But yeah, >> uh I said it before and uh I hope people >> um you know, you go to a a retail store and and demo these, try them out, but like trying the neural band, realizing how quickly you adapt to this interface, it's like this pretty much the same motions you're used to on an iPhone.
30:24
You just don't have a phone >> and uh takes a little bit of getting used to.
30:28
Their pitch for it was was, >> "Well, because it's not using a camera, you can you can use it out of view of what the camera would be if it was on your face.
30:38
You can use it behind your back."
30:39
I don't actually care about that.
30:41
I think it's fine to put my hand in front when I'm adjusting the volume.
30:43
I don't really see a value of putting it behind my back.
30:47
I think the value is that all that like you need you need some input sensor to capture what's going on with your hand.
30:54
And it makes sense to just put that right next to the hand.
30:57
And I think that if even if it's just a couple grams, taking that off the face and putting it on the wrist is way better, way more natural.
31:03
I do think it was interesting that Bos was saying that long term the neural band could be uh an input device for other applications and eventually they could open that up to developers.
31:13
So you could have an app that use the neural band as an input or it could be a little bit more platform agnostic at some point.
31:20
Very unclear where that actually goes.
31:22
We were debating a little bit.
31:23
We didn't get our firm uh percentages down, but uh between voice input, just speech to text versus handwriting, I think most people will be doing speech to text.
31:35
But I don't know if that's just because I've been so I dictate like a ton these days.
31:41
I open up the chat app, I click on the audio mode, and I just talk and talk and talk.
31:47
And then even if I'm not, a lot of times my prompt is just like clean this up and make sure it's grammatically correct. Like don't change it. Don't rewrite it. Just reproduce it.
31:56
And then I can just copy that into wherever I need to go if it's a longer thing.
31:59
I think that long-term people will just be comfortable talking and and we you're able to talk in a very low whisper and it picks it up just fine. >> Yeah.
32:09
>> The the the secret handwriting, I don't know how I don't know how popular that that's going to be.
32:14
Maybe like 20% of total input.
32:16
I think it's going to be popular for short responses where you just need to be like >> like, "Hey, do you want me to grab you a sandwich?" >> Yeah.
32:26
Well, there's also just >> you want me to grab you a coffee. Yes. >> Yeah.
32:29
Well, there's also like the thumbs up.
32:30
It it dynamically selects emojis there.
32:33
So, you can just swipe and be like heart or thumbs up or thumbs down.
32:37
Uh, so I it is like in between just like send an emoji response, dictate something longer, handwriting for something in the middle.
32:44
I don't know how popular it's going to be.
32:46
It's definitely like a lift.
32:48
People are gonna have to learn.
32:49
It's like an entirely new input medium.
32:51
But then again, speech to text is is somewhat of a new input medium and people have figured that out.
32:55
So, uh maybe maybe over the few years people bring that back, but I I'm not sure.
33:00
I'm not sure if I would be leaning on that all the time.
33:02
Maybe I'm just a terrible uh terrible handwriting. So, I don't know.
33:06
Um anyway, it was a fun uh fun event and thank you all for watching.
33:10
Let me tell you about Vanta.
33:12
Automate compliance, manage risk, prove trust continuously.
33:13
Advances trust management platform takes the manual work out of your security and compliance process and replaces it with continuous automation.
33:19
Whether you're pursuing your first framework or managing a complex program, Tower of Babel.
33:24
>> Uh yeah, let's start there.
33:26
>> So we were talking to Chris Cox, the chief product officer at Meta, and he said that AI translation is tower of babel level.
33:33
Uh and it was a very funny, he used a very funny word, a judgment.
33:40
He was like with AI you can take a judgment and and scale that was in a different context. >> It was in this clip. It was in this clip.
33:45
AI can be used to scale a judgment.
33:48
And I and I forgot what word he used and I asked him and he didn't remember and we kind of just moved on in the conversation.
33:53
But um I was uh but Skooks is saying me when I don't know what the story of but Tower of Babel was about.
34:00
Um the Tower of Babel story of course is is humanity in in the book of Genesis used to all speak one language and then God uh fragmented the languages.
34:10
Um but uh >> and so I think he was he was generally alluding to going back to >> Yeah.
34:17
like like is it hubris for humanity to try and speak all the same language that that there's always these uh Meta loves these nicknames for projects that have >> or or go either way a a a massive project that collapses under its own weight.
34:33
The other way to interpret that is that okay >> the glass you know every code word is a double-edged sword is a double-edged sword.
34:43
I mean, every every uh, you know, Greek myth is a double-edged sword.
34:47
Like, just the Orion headset, their full augmented reality headset.
34:51
Orion is the story of the hunter who has uh gets so cocky, so much hubris, he decides he can kill all the animals and he's struck down by God and and then sent to uh sent to the uh the the heavens to exist as a constellation.
35:07
uh the the uh the the the the new data centers called Pr Prometheus, which of course is uh stealing fire from the gods.
35:14
And and there's another Hyperion.
35:17
Uh in in one of those, the the the perpetrator of the Hubris is sentenced to uh have his kidney liver eaten by an eagle every day forever.
35:27
There there's a lot of like bad endings in these stories, but they all also are like very powerful and cool names.
35:35
Tyler, what's your favorite Metacode word?
35:39
>> There there's a lot, but I think there's there's some Straussian reading where maybe Zuck is actually a doomer. >> Oh, yeah. >> Secretly.
35:46
>> He thinks Yeah, he thinks that uh that Prometheus, he will steal fire from them gods and be smited >> perhaps.
35:52
I mean, the Behemoth thing was crazy because it's it's like behemoth llama and do you remember the AI generated image that they used to promote it?
35:59
It was like this very like uh demonic like like beefy bulky llama. It was crazy >> guys.
36:06
How did this one get out?
36:07
>> And then and then it didn't get out and then it was actually like a it was it was too big to tame and like that is the story of the big >> but simply the image.
36:14
>> Yeah, the image was crazy too. use this.
36:16
>> And so, uh, yeah, maybe just stick with like the version four, version five. I don't know.
36:23
But then we get into numerology. Who knows?
36:23
Um, >> I got to try to find this.
36:25
I I have this I'm I'm putting this uh image in the chat, guys.
36:31
>> Uh, Tyler, do you think that they're subsidizing the cost of this thing?
36:33
So, Joanna Stern said, uh, wow, $7.
36:35
99 for Meta's new screen equipped Ray-B bands and the neural wristband.
36:40
I thought that was shocking when they told us on Friday.
36:43
Uh Ben Thompson was shocked by this as well.
36:45
Everyone's been reacting to just how cheap that is.
36:47
The Orion headset when they demoed it last year, the rumor that was going around was that it costs $10,000 to produce those.
36:53
And so people were like, "Okay, yeah, so they'll get they'll get twice as uh twice as cheap every year for five years or something and then we'll be in the 750 territory."
37:03
But this has a lot of the same tech and it's $799.
37:06
It's half the price of the of the original Google Glass, which is crazy.
37:10
Uh do you think they're subsidizing it?
37:12
Uh, it definitely seems like really cheap.
37:14
I was surprised that it was under $1,000.
37:17
I don't know enough about the actual technology to tell like what the price would be, but it does seem like really cheap.
37:21
>> There's so many reasons to make it as cheap as possible, right?
37:25
Just like one, user feedback.
37:27
Two, it can uh help them improve.
37:31
Um, >> oh yeah, there's the llama for behemoth preview.
37:35
>> Yeah, this is a picture. Hey, big llama.
37:37
I'm glad they >> It's kind of fun.
37:38
I I think it's kind of funny, but it is a fun. >> It's kind of funny. A jack llama.
37:42
That's something I would do.
37:45
>> You think somebody's sleep paralysis monster? >> It was kind of crazy. Uh it looks cool.
37:51
It looks It definitely looks And it had great vibes when you chatted with it.
37:53
It was goofy and laughing about the fact that it was >> extremely metal.
37:58
>> Uh but yeah, it is very ext. It's hardcore.
38:00
Uh anyway, much like graphite, graphite.
38:03
dev, dev code review for the age of the AI.
38:05
Graphite helps teams on GitHub ship higher quality software faster and get started for free.
38:08
So on the subsidization thing, I mean they're like they're they're saying like the best selling product, but they're at the $300 mark, $400 mark.
38:17
And I don't think they're selling tens of millions of those.
38:22
I think they're selling like millions.
38:23
So if you sell a million and you're subsidizing by 200 bucks a pop, it's like, okay, that's 200 million bucks. Like that's fine.
38:29
Like Meta has no problem with that.
38:31
In fact, they can two AI researchers. >> Exactly. Exactly.
38:35
So, what they can do is they can say even if they're even if they're like, "Let's subsize this by $1,000 and let's sell a million pairs."
38:42
It's like, "Okay, they're going to lose a billion."
38:43
They probably can't sell 10 million or or 50 million pairs.
38:45
But they could just do a run of a million pairs at a huge loss and then just cap it and be like, "Yeah, we're out of stock. V2 is coming. V2 comes."
38:54
And then they raise the price and they say, "It's even better and even lighter and even thinner."
38:58
And then they slowly creep together.
39:00
They never need to make money on hardware. >> No, I don't think so.
39:04
>> Like they never need to make money here.
39:06
I'm sure they will if they keep executing like this, but ultimately it's just about it's about owning owning the next platform.
39:11
Uh Signal uh screenshotted uh the stream and said Meta is so rich they have a VP for just fashion partnerships.
39:20
Um and Michael Mirllor, I don't know if he's in the chat right now.
39:26
Ava Chen has been at Meta since 2015 and is the primary reason why Instagram is Instagram.
39:30
This should be common knowledge.
39:32
You do not know ball if you do not know this.
39:37
Especially if you've ever tweeted about tech and taste.
39:40
>> Uh I thought I mean I thought I thought the the post was funny.
39:42
Uh Signals post was obviously is a friend of the show, but I thought it was funny purely because >> Instagram is the most influential platform in the world for fashion.
39:52
They generate billions of dollars a year from fashion brands advertising on the platform.
40:01
>> Like of course they should have somebody to do like >> and you don't have to be a VP.
40:04
We have a VP of logistics, a VP of production, a V we have a chief intern officer CIO and so you know you could titles are free.
40:14
>> I don't know if you need to worry about that.
40:16
>> Yeah, I would be I would be bearish if they didn't have somebody that was just focused on building because it's it's on both sides, too.
40:22
It's not just companies, it's it's the creators, right?
40:26
And think about the most the biggest creators in the world.
40:30
>> Many of them are are advertising.
40:33
>> Also, I mean, think about the fashion partnership they've done.
40:35
It's like a multi-billion dollar deal with Lexodica.
40:38
They've taken a position, a major equity position in the it's we can make we're going to make face computers, >> but they have to look good otherwise people won't wear them.
40:48
Should we have somebody that understands this world? Sure. >> Yeah.
40:51
Al, also also just in terms of like the VP title, like you you want like a fashion partner analyst showing up to uh do a multi-billion dollar deal with like the number one sunglasses brand in the world.
41:03
Like, no, you want to send a VP.
41:05
You want to send someone with a real title. Uh anyway, very silly. Uh but but fun.
41:08
Everyone's having fun on the timeline. We'd love to see it.
41:12
Um >> we're promoting Michael Mirlo to vice president of taste.
41:17
>> Yes, vice president of taste.
41:17
Um Mark German gave his take.
41:20
He says he has a strong feeling these will be popular.
41:21
I wonder if this will cause Apple to speed up its timeline.
41:25
Right now, I'm not anticipating Apple glasses with displays for a few more years.
41:29
Their first non-dis model is likely being announced late 26, early 27. That is slow.
41:34
That's a lot of time for Meta to to iterate on this and actually get through it.
41:40
It It's going to be a big It's going to be a big fight.
41:43
Uh Google's talking about doing display glasses.
41:45
Wait, so Gurin is saying they'll do a pair of glasses that just have a camera in them. >> Yes. Yes.
41:51
A competitor to Meta Ray-B bands. Yeah, I know. What do you get? Not that much.
42:00
You do get slightly tighter integration with Bluetooth.
42:02
So, you will uh you'll probably be able to pair them and unpair them a little bit more easily because AirPods are a little bit easier to pair reliably than the Meta Ray-B bands.
42:13
But in but as a competitor meta ray bands it's camera headphones access to AI. Okay.
42:19
So you get the Apple uh Ray-B bands which won't be Ray-B bands.
42:25
They will look like Apple products and they will have maybe a better camera because Apple's fantastic at cameras.
42:31
Maybe better Bluetooth connectivity and reliability there but you don't get any other features.
42:36
And then instead of talking to Meta AI, which is probably going to ship something really sick out of MSL soon, like you're dealing with Siri and Apple intelligence, which we don't know the timeline for the V2 of that.
42:46
They're kind of moving slow on anthropic partnership or something like that.
42:49
But um even though they have the chatp integration, like I have not become comfortable using the Siri button to reliably query chatpt, I still open my phone, open that app every time. >> Yeah.
43:03
Um, and so there was also a rumor that they might partner with Gemini, which actually would be great because that model is fantastic.
43:08
And so, um, it'll be interesting to see where they where they pencil out on that, but um, uh, Meta seems to I I would be surprised if if Apple can just come from behind and dominate. >> Yeah.
43:22
You have to look at Meta's advantages, which is like we own the platforms that you share content on, too, right? >> Yeah. Um, well, Julius. ai AI.
43:30
What analysis do you want to run?
43:34
Chat with your data and get expert level insights.
43:40
>> Julius, we love Julius.
43:40
The AI data analyst that works for you.
43:42
You can join millions who use Julius to connect their data. Ask questions.
43:47
>> You should have had Raul show up at Metacon and then just like hop on the stream.
43:52
Yeah, that' be great in the background. >> Surprised. Uh, what else?
43:57
>> Loavv was saying, uh, start collecting training data right now.
44:00
put these on every manufacturing worker in America. Good.
44:05
Uh >> yeah, >> probably probably smart. May as well. You could just do that.
44:09
You don't need the display either.
44:12
>> The >> But again, Amazon is already saying we're putting the they're developing their own pair of glasses to put on their >> workforce.
44:18
Hopefully, it's not a death nail.
44:20
When when Google Glass pivoted to B2B, it was sort of like going out to pasture because they they were doing Google Glass for consumer.
44:28
They thought it was going to be this massive like consumer adoption moment. There was a ton of hype. Didn't go anywhere.
44:32
And then pretty soon it was like Google Glass for enterprise.
44:36
We'll use it in manufacturing context.
44:37
That's not what Aeron's saying.
44:39
Um but but I would be personally worried if if uh Meta started talking about, oh yeah, these are going to be really great enterprise use cases.
44:50
They're not ready for consumers.
44:50
Like no, they need to be ready for the the the Instagram crowd.
44:54
like it has to it has to integrate with meta platforms in order for it to be uh successful in my opinion. Yeah.
45:00
Uh let me tell you about fall the generative media platform for developers.
45:05
The world's best generative image, video and audio models all in one place.
45:09
Develop and fine-tune models with serverless GPUs and on demand clusters.
45:12
>> We should do a soundboard partnership with Fall. >> Oh yeah.
45:16
>> Anyone can uh >> Great.
45:18
>> Can generate TVPN like sound effects.
45:21
>> Well, we talked about Scott Woo earlier.
45:23
He's in the timeline showing some love to Mark Chen over at OpenAI.
45:25
Mark, he said, "So insane.
45:28
You have you guys have no idea how hard this is."
45:30
Mark Chen said, "We wrapped up this year's competition circuit with a full score on ICPC after achieving sixth in the IOI, a gold medal at the IMO and second at the ATCoder heristic context."
45:43
>> I never I never see you at the heristic contest. John >> contest.
45:48
>> I was at the heristic contest and they didn't know your name. >> No. Yeah.
45:51
We Yeah, we got to send Tyler to to be our on the ground correspondent at the IMO and the IO IOI next year. I think that'd be great.
45:59
Do some uh postgame interviews for the for the contestants and then uh get them directly routed into the tier one VC firms.
46:07
Uh they're already getting calls from from from investors.
46:11
>> This was probably the biggest news of yesterday, but it went under the radar.
46:15
Uh there is a Rolex uh Oyster that uh that's Domino's Pizza branded.
46:23
And uh >> Andrew Reed says he's not a watch guy, but he might get this one.
46:26
Uh it's a 1970 uh vintage Rolex Oyster Precision Classic men's Domino's Pizza watch.
46:33
And we have the story behind it, the Domino's Rolex Air King in 1997.
46:37
They've done a apparently.
46:41
So, Domino's Pizza founder Tom Monigan uh began incentivizing franchises with Rolex watches when a high earning franchise owner earned the watch off his wrist by hitting a $20,000 sales week. Let's go.
46:59
Monigan wrote in his 1986 biography, Pizza Tiger.
47:02
What a great name for a biography.
47:04
Uh send it straight over to David Senra.
47:07
I wore a Bova with our Domino Lo Domino's logo on its face.
47:10
A franchisee asked what he had to do to get that watch from me.
47:15
I told him, "Turn in a $20,000 sales week." He did it.
47:21
And so he won the watch off of his wrist.
47:24
After that, >> I wonder what the what do you think the market value of that watch was at the time?
47:29
Because he might have might have been an arbitrage just realizing, hey, >> well, we're about to find out.
47:33
After that, Monigan started rewarding top performers with Seikko watches and later upped the stakes by giving away, okay, hundreds of $800 Rolexes.
47:40
So, in the 80s, we're talking a couple hundred bucks for these, but that's still a lot of money back then.
47:47
Uh, initially turning in a $20,000 sales in a week at Domino's would earn a Rolex.
47:53
However, as Rolex prices increased, so did the challenge. Domino's c Domino's.
48:00
>> Did you get it every time?
48:00
So theoretically you could get onelexes a year. >> I don't know.
48:05
Yeah, I imagine if you just like pick the right location, set the right prices, have good traffic, like you should be able to put up, you know, fantastic numbers. Very motivating.
48:16
The Domino's in like Time Square has got to do tons and tons of revenue, right?
48:19
Uh Domino's continued to give out branded Rolexes, but a franchisee eventually needed to achieve $25,000 in sales per week for four consecutive weeks to win the watch.
48:29
But the air king and uh these apparently there is one available on bezel. Go to getbzel. com.
48:36
Your bezel concierge is available now to source you any watch on the planet.
48:40
You can ask them for a Domino's pizza.
48:43
>> Meta Meta Exact team had some tasteful absolute hitters.
48:47
>> Yesterday I clocked >> Bos had a great watch on. >> Adami had a Nautilus.
48:52
>> Vintage uh vintage aqua >> Aquinaut.
48:53
The Aquinaut on a black strap. >> It was great.
48:57
What do I have to do to win a a TBPN air king? >> I don't know.
49:02
We'll have to figure that out.
49:04
Chat, let us know if you have any ideas for for Tyler.
49:07
>> We were we were actually debating whether or not we should do Q4 bonuses in watches.
49:12
And John's just for the record, John's point of view is that you guys would probably just want cash, but >> it might be delivered in watches before. >> We'll see. We'll see.
49:22
Um uh in in other >> guys the guys are clapping. See I >> Rolexes.
49:28
Okay, maybe Jord's on to something.
49:30
>> I think I might be on to something.
49:32
>> Well, uh we have our next guest joining in just a few minutes.
49:34
Uh but first like there there's big uh breaking news yesterday.
49:37
Disney's ABC is pulling Jimmy Kimmel indefinitely after late night uh after the late night hosts recent remarks about Charlie Kirk.
49:43
The move comes as ABC affiliate groups told network uh they would be dropping the host.
49:49
So um this didn't come from top down from Disney.
49:51
came from the uh from some of the affiliates.
49:54
Next media is one of the largest n the nation's largest TV owners said it would >> open AI having a complex corporate structure but media on media company Sinclair is also involved.
50:06
So Sinclair put out a statement uh Mike Salana says so this didn't have anything to do with the FCC.
50:12
There was a lot of debate over what the FCC's role was in this who was really uh putting pressure on there.
50:19
There was a viral video, a clip of uh of uh of Kimmel's monologue that went viral on Tuesday after he delivered it Monday night.
50:30
There was a lot of backlash to that and then this came up.
50:32
Uh Jason Calacanis, the host of the All-In podcast, said executives at ABC, like those at CBS, wanted to fire these money- losing late night franchises for years.
50:42
Trump has even given them cover.
50:45
Trump has given them the cover they needed when it was wildly profitable to back these same comedians a decade ago.
50:51
The networks had no problem letting them wa mock our wonderful, amazing, tremendous, beautiful, brilliant, and astute President Trump. Wink. Uh, every single night.
51:00
And so, uh, I we we got a little bit of detail on Co Bear's financial situation.
51:06
I have no idea what the financial situation was like at Kimmel.
51:10
>> Viewership ratings were dropping though. >> We did hear that.
51:12
I think it probably tracked the um basically the the the >> I mean the the audience on on these networks is like retirement age, right?
51:24
And so these audience >> 67 was the median age at CNN and that was like the the lowest or something like that.
51:30
Uh it does feel like it gets harder to harder harder and harder to monetize an older audience um just because they're not as they're not as proflegate with their money.
51:38
They're not just going and spending money on all sorts of things.
51:42
They're not building AI companies.
51:44
They're not using turbopuffer.
51:45
They're not even in the They're not even in the ICP for turbuffer.
51:48
They're not using serverless vector and full text search built from first principles and object storage.
51:54
Fast 10x cheaper scalable linear >> and our friends over at Rewise too.
51:59
Uh some very important breaking news.
52:02
Uh Chinese Joe Weisenthal has hit the timeline.
52:07
Chinese Joe says >> familiar with American Joe When we first started raising awareness about Chinese Joe Weisenthal just a few months ago, the legacy media and the corporate establishment laughed in our faces.
52:19
Something tells me they aren't laughing any longer. Chinese Joe Weisenthal.
52:21
Uh so I >> and Geiger Capitals is very bullish China. >> It is. It is.
52:29
>> They're catching up to the US much faster than anyone anticipated.
52:34
>> Uh every country should have a Joe Weisenthal.
52:36
They would be lucky to have one. >> Yes.
52:38
uh and every every country needs a profound get your brand mentioned in Chachi PT reach millions of consumers who are using AI to discover new products and brands and we have our first guest of the show from Palunteer coming into the studio. Welcome to the stream.
52:55
>> How are you doing Lewis? Good to see you.
52:59
>> Very good to see you guys.
52:59
Thank you for having me on. >> Thanks so much.
53:01
Uh would you mind kicking us off with an introduction on yourself, a little bit of your backstory, history of Palunteer, and we'll go into the news. >> Yeah, sure. So, I'm Louis.
53:10
I run Palunteer out here in the UK and Europe.
53:15
Uh you can tell probably from my accent that I'm British.
53:19
>> And I've been at Palanteer like almost a decade.
53:21
Um so seen a seen a seen a lot of change.
53:24
Um a lot of growth out here in the UK and Europe.
53:27
um that journey from a business that was very focused on defense and intel with a tiny bit of corporate to now a business that's serving every bit of the public sector and every corporate sector you can think of.
53:44
>> And what's the news today or yesterday?
53:47
>> Well, the big news is well technically no technically it was uh early UK time this morning so you're not out of date.
53:55
Uh uh we announced a big uh big deal with the UK Ministry of Defense. U billion dollars. >> A billion.
54:03
>> First billion dollar deal. >> A billion US dollars. >> Congratulations. >> Thank you. Thank you guys.
54:09
Uh well, it's the first billion dollar deal that Palanteer has done outside the US.
54:13
So it's a big significant milestone.
54:15
Uh and alongside that deal, we also announced uh a big investment into the UK, $2 billion over the next five years.
54:23
Uh the creation of 350 new jobs. >> Mhm.
54:29
>> And uh our European HQ for defense will be in London.
54:33
>> What is the $2 billion uh investment look like?
54:36
Is that uh just the investment in the team opex capex?
54:39
Like how are you thinking about that? >> It's all of those.
54:44
And uh London is already um little known fact it's already home to about 20% of Palanteer's headcount. >> Oh wow.
54:52
>> It's uh so it's actually our second largest office globally.
54:54
So we do a lot more than just support you you know British customers from this office.
54:58
We do a lot of product development.
55:00
Uh and it serves as a as Yeah.
55:02
as like the European and and broader Emir headquarters. >> Yeah.
55:07
When did you realize that billiondoll deals were possible?
55:12
Was it 10 years ago when you started? Was it 5 years ago? Was it more recently? It's a big number. >> Yeah.
55:21
I I think I always believed um and and I I think we're only just getting started.
55:26
You know, this is uh you know, we'll look back in 5 years time, I think, and and we'll think, yeah, those were small deals.
55:32
Um you know, the power of the software is such and it's meeting its moment.
55:37
uh you know the significance of this deal is is it's you know we are the operating system for the modern battlefield >> and we've seen the US make that move and the significance of the news today is is the US's closest ally the UK >> making the same move. >> Yeah.
55:56
What what's the mood like in the UK relative to the rest of Europe? We talked to Dr.
56:01
Karp uh was that just last week or was it the week before?
56:06
uh about uh the reception and what's going on in Germany.
56:09
He of course studied in Germany and he was uh kind of joking about the lack of uh entrepreneurial talent and adoption and uh kind of getting with the program in Germany, but it seems like the UK might be a little bit more forward thinking.
56:25
Walk me through sort of the view in Europe technology right now.
56:30
>> I think I think that's spot on.
56:30
I think the UK is an outlier.
56:32
the UK is an outlier. Um we've got uh especially here in London an incredible talent pool >> uh especially in the computer science software engineering domains you know that's why Palenteer has so many people here it's why we do product development
56:48
here it's it's really access to the talent >> and you've got deep mind you've got you know you've got a key key parts of the of the broader AI supply chain ecosystem are are here and I think it means UK is is like the only country in the west
57:04
broadly defined outside the US that does have that kind of talent pool uh and obviously the language helps speaking English the connection to the US and uh you know we weren't alone right today Palanteer is not the only company to
57:18
have announced a big investment in the UK alongside um President Trump's visit we saw Microsoft we saw Nvidia we saw open AI we saw Google a whole raft of big tech make big investments >> uh how are you thinking about the work that you'll actually do. How how How how concrete is it?
57:35
How much can you share about uh what's actually in scope for this contract?
57:42
>> I I can share bits of it.
57:42
Obviously, a lot of it is sensitive operational details, but um a key part of it will be what the UK is calling the digital targeting web.
57:54
>> Uh we we Palunteer will be a component of that.
57:56
It'll involve many many other companies and players.
57:58
And you could think of that a bit like what the Maven smart system does for the US.
58:03
So it's really your targeting infrastructure, how you connect up all your sensors, your satellites, your drones, all of your various data feeds with your aectors, you know, the stuff that you're going to use to shoot airplanes, tanks, missiles.
58:16
And uh it's that data infrastructure that sits in between that.
58:20
It's it's the harness in which you then run all of the sophisticated AI and computer vision algorithms and and so forth.
58:27
forth. And uh a lot of it is inspired and frankly lessons learned from the war in Ukraine >> where this the same technology our our platforms have been as you'll know deployed by the Ukrainians day in day out now for nearly 3 years >> and those lessons are being learned and
58:46
you know we're seeing the future of warfare play out in real time and uh you know the significance of this is is is the UK government making a multi multi-million pound investment in in that >> what was the uh what was the precursor technology to Maven. I remember when I
59:01
I remember when I uh when I dug into it, I mean it was like controversial program here in the United States.
59:08
And um and when I looked into it, it seemed like the precursor to using image recognition and computer vision to identify objects and images was uh thousands of Air Force airmen tagging and basically doing something that you would expect like a data labeling firm to do.
59:30
And it's not like the government wasn't trying to identify images, objects, and images.
59:35
Uh they just weren't using technology.
59:37
Is it the same thing over in the UK?
59:39
Like what what's the lineage of this? >> Yeah, exactly right.
59:42
And and if anything, the the the problem is more severe because the UK is smaller and has fewer resources than the US.
59:49
>> So, uh you know, you're never going to have enough eyeballs to watch all of those feeds.
59:53
M >> so you need to find some way of of automating that and surfacing something of interest to the human when when that occurs when that matters that's how you're going to scale up.
1:00:04
Does this type of uh deal make uh the next five deals with other western allies easier? Right.
1:00:13
I'm assuming that uh it would be you know given it just becomes much more difficult to coordinate if if allies are operating on different systems but I don't know if I have the wrong framework for that.
1:00:24
I think no I think that's that's a that's a I mean obviously I'd hope that's a likely scenario but it's you know it's critical and again this is a lesson from Ukraine right the interoperability is everything >> the ability to t pass targets seamlessly between uh between units between allies that is that is the way we're going to uh confront and deter the adversaries that we now have in the west >> when we talked to Dr.
1:00:51
>> when we talked to Dr. Karp he mentioned that um because of the current structure of Palunteer because of the the where we are in the technological adoption curve artificial intelligence uh he did not expect headcount to grow significantly
1:01:08
does this deal change anything for Palunteer's UK office I imagine it's like a whole lot more business it might justify some hiring are you hiring uh how do you think about actually supporting 350 new hires but that didn't feel like a huge number in the I mean it
1:01:26
just you know the efficiency of of the platform >> it will it will mean we hire uh we will be creating jobs but >> the output per head >> y >> is going to grow even more significantly you know I think uh you you know Ted Mabry my colleague who runs uh runs the
1:01:44
US or the global commercial business was just tweeting today about AI FDEES So the the forward deployed engineers we had that have historically been human beings, we're starting to to to replace some of the work they would have done manually in the past with AI. And you
1:02:00
And you can just see a path now where like 90% of what used to be the day-to-day job can be done by AI.
1:02:09
So then suddenly you've got 10 FTEEs for every one you used to have.
1:02:14
I mean the the the exponential here is is crazy. >> It's amazing. Uh well congratulations.
1:02:20
Thanks so much for coming on and staying up late to join the stream. Not a Yeah. billion dollar deal.
1:02:24
We have to hit this button for you. >> Overnight success. >> There we go.
1:02:31
>> Thanks so much for coming on the show. We'll talk soon. Congrats. >> Thank you. Thank you. >> Linear.
1:02:38
Linear is a purpose-built tool for planning and building products.
1:02:40
Meet the system for modern software development.
1:02:42
Streamline issues, projects, and product road maps. Head over to linear.
1:02:46
Uh the bond market has completely rejected the Fed's rate cut.
1:02:50
We talked about this a little bit yesterday.
1:02:51
Yields have ripped higher following an initial drop on the rate cut announcement.
1:02:55
The rate cuts had opposite >> up prayer circle at this point.
1:03:00
>> Yes, mortgage rates will now spike.
1:03:03
Mortgage rates were creeping down.
1:03:03
Uh this is not very good news.
1:03:05
We're still talking about.
1:03:07
1% of a move over the last day, but not in the right direction.
1:03:12
So >> yeah, >> not fantastic.
1:03:14
But of course that that account is from QE infiniti uh which is quantitative easing of course. >> No no bias there.
1:03:21
Um uh our post earlier >> for uh on August 10th uh Leupold uh >> age like a fine wine. >> Yep.
1:03:32
>> We went we we were extremely bullish on Leupold and situational awareness.
1:03:34
uh back in August 10th, his fund topped $1.
1:03:39
5 billion and posted 49% gains just for the first half of 2025.
1:03:45
This was from uh filings.
1:03:48
Whenever you're running a fund, you need a >> remember at that time there were some people saying like >> long Intel, what could >> his fund already blown up? >> Yes. Yes. Yes.
1:03:57
And so, uh Swix shares a screenshot of what Intel has been doing over the past year.
1:04:03
They're up 28% and Swix says, "Oh my god, he's going to destroy H2 20252." Uh, >> good.
1:04:13
>> And now, >> good year to have some situational awareness.
1:04:16
>> Mise says everything reminds me of him and it's all green lines up and to the right with a variety of stocks up 50% 100%.
1:04:24
He has fully nailed the AI trade.
1:04:29
Um, Will Brown says you could buy so much diet, Dr. Pepper with $2 billion.
1:04:31
I don't know if that's like Leopold's favorite drink or something.
1:04:34
I I don't get the diet diet type reference, but >> this stat was insane and a little disheartening.
1:04:41
Yes, >> the top 10% of US consumers now account for half of all consumer spending, which is a record high, >> up from about a third in the early 1990s.
1:04:53
Okay, so >> creeping up different way to understand um >> wealth inequality.
1:05:00
And it just goes to show if you're if you're building consumer products, you probably want to be building for that that top 10% that are spending half of all dollars >> for sure.
1:05:12
And you got to pay your sales tax on numeral numeral HQ. Sales tax on autopilot.
1:05:16
Spend less than 5 minutes per month on sales tax and >> numeral just announced a new round today. Congratulations.
1:05:20
$350 million up from their series B uh earlier this year or series A earlier this year.
1:05:30
Um and uh we're going to have Sam on tomorrow live from the studio for his uh first uh first ever guest appearance and it'll be in pumped for that. >> Uh still on Intel. Funny journey.
1:05:43
If you bought $10,000 worth of Intel 25 years ago, it would be worth $10,000 today.
1:05:50
Uh absolutely zero movement over the past 25 years.
1:05:57
Basically, it's a store of value.
1:05:59
I guess uh it there has been movement.
1:06:01
Of course, you got to factor in inflation.
1:06:04
>> And then it actually sort of ripped from 2010 uh up to 2021, but then it fell and went back up, but now it's climbing back up.
1:06:14
And uh the the the the big news on Intel World is that Nvidia invested $5 billion.
1:06:19
Um >> they marked Trump up. >> They marked up Trump.
1:06:22
They marked up the US taxpayer.
1:06:24
Uh since now we all own Intel shares.
1:06:27
Uh but this could be the start of something cool.
1:06:30
Um this was what uh uh who was it?
1:06:32
John at at uh at Asianometry was proposing the idea that uh Nvidia would second source the CPUs from Intel because uh >> Nvidia is up 22% today.
1:06:49
>> Today what is there China news? What? >> Sorry, sorry. Not Nvidia. Intel. >> Intel's up 22%.
1:06:58
>> I'm so used to saying Nvidia.
1:06:58
No, Nvidia can't be up 22% a day. It's so big.
1:07:01
It went down 3% yesterday on the China news. >> Back up 3%. >> It's back up 3%. >> 3. 7%.
1:07:08
So back to where they started.
1:07:11
>> Market thinks they're going back into China, baby.
1:07:12
That Blackwells are going to be shipping the B30s. They're It's happening.
1:07:16
Um Bobby Cosmic is rooting for Intel and says he loves.
1:07:21
>> Of course we're rooting for Intel.
1:07:21
We're all rooting for Intel. >> We're having Pat.
1:07:26
>> This post uh is hilarious. Every AI app today.
1:07:29
And >> it's a box truck with a sprinter van inside.
1:07:33
And then inside the sprinter van is a a subcompact car.
1:07:39
>> And this is just the nature of uh recursively calling various API abstraction layers probably um the agent that calls the underlying LLM.
1:07:49
So, >> laughing at >> South Park uh delays new episode hours ahead of airtime because creators didn't get it done in time.
1:07:59
They say apparently when you do everything at the last minute, sometimes you don't get it done.
1:08:04
Do you think this is because they were about to do something that that was going to get them taken off the airwaves? >> Maybe.
1:08:09
I think >> I think a lot of people that um are used to joking around are >> understanding that uh at least in this period of time >> jokes can have uh consequences. Blowback. >> Yep.
1:08:28
>> Cancel culture is back.
1:08:29
>> Cancel culture is back.
1:08:29
It is remarkable that they uh that they uh ship an entirely new episode that they that it's a it's a produced animated show, but it's created around the news cycle and so >> real time.
1:08:44
I believe when the when when the Obama election happened, they had I think they prepped two different episodes and then they were able to air the one that w that corresponded with the the correct winner or something like that because they had that episode up like the same day that Obama won, which was crazy.
1:09:01
>> Um >> but uh hats off to them considering the their track record. It's impressive.
1:09:07
This doesn't happen more often.
1:09:09
>> That's what I'm saying.
1:09:11
You really you got a >> tinfoil hat moment. tinfoil hat moment.
1:09:15
>> Yeah, they Yeah, I don't know that they've ever missed uh Have they ever missed an episode?
1:09:18
They must have at certain point.
1:09:19
And they do and they do go off air for a little bit and I'm sure they have uh stuff in the backlog.
1:09:22
But it's crazy how long they've been at it.
1:09:27
This has been what 20 30 years on the air.
1:09:28
Um it's it's one of those shows that I keep forgetting about and then somebody shares like, "Oh, they did one about AI.
1:09:33
You got to you got to watch this.
1:09:35
They did this about the thing that you really focus on. So you got to do it." >> Fin.
1:09:39
AI, AI, the number one AI agents for customer service, number one in performance benchmarks, number one in competitive bakeoffs, number one ranking on G2.
1:09:47
>> Founder, >> I'm only able to find one other instance where they had a official production delay for a new episode. Really was in 2013.
1:10:00
>> Well, um, Bryce Robert says, "Don't wait for more resources to have kids."
1:10:03
John Woo, uh, broke it down. got 9,000 likes almost.
1:10:09
>> Yeah, this post went super viral.
1:10:10
>> It said, uh, "My only regret in life is not having kids earlier.
1:10:13
When I was 27, I was going out four nights a week, working 12-hour days, six days a week, waking up at 5 a. m. daily for CrossFit.
1:10:22
K would have been a breeze.
1:10:22
Don't wait for more resources to have kids.
1:10:24
Energy is scarcer than money." It's a good take.
1:10:29
Thanks for blowing this up, guys. Follow me. Yeah.
1:10:31
The other the other thing here is if you have kids younger, it just means you're younger when your kids are fully functional and independent, right?
1:10:40
>> Also, I mean, there's just this like idea that you like, oh, you need to like like you you don't you actually don't get to stop working when you have kids.
1:10:48
Like you actually do go to work and and make more money.
1:10:51
And so it's it's you know it's not just like oh you need to be retirement age or have like retirement money before you >> I mean I think it's entirely a fixation of people have had in their mind forever this idea of like you buy your first you get married you buy your home and then you have kids >> and that's so ingrained in the culture that people just don't even consider having kids. >> Yeah.
1:11:14
Because of the house >> because house prices are so high. >> Yeah.
1:11:16
They're not in a position to buy a house.
1:11:18
But >> I think uh yeah >> well >> kids are nothing more nothing to uh motivate the grind set like uh >> indeed >> couple of pups >> deeply motivating >> we got John soundboard finally he's uh he's trigger happy >> uh Amit who he uh met at Palunteer dev day said uh >> number one Palunteer retail soldier >> yes uh he said imagine general told Jensen that If he didn't pump his Intel bags, then he wouldn't be able to sell to China.
1:11:50
Then Jensen saw the 60 billion they had in cash and was like, "Screw it.
1:11:54
Put 5 billion into Intel so we can ship to China."
1:11:57
>> Do you think this solves China problem? >> I don't know. >> I don't know.
1:12:00
I don't >> because it's not a Trump issue anymore. It's a Beijing issue.
1:12:04
>> It's a Beijing problem.
1:12:04
So >> So, um, yeah, I don't know what I don't know what Trump has left.
1:12:11
>> I mean, Trump up 36% on his first real trade size. >> Pretty good.
1:12:16
makes you want to he makes him want to get him set up.
1:12:20
>> But there's a lot of uh there are a lot of other like chips on the table that Trump can pull.
1:12:26
There's rare earth, there's Tik Tok, there's all sorts of what >> what did you say? >> Chips on the table. >> Oh yeah, yeah, yeah. Of course.
1:12:32
Uh it was an accidental pun.
1:12:33
Um but uh there are other places where I if Nvidia rises to this to this top of the stack and becomes one of the most important American geopolitical issues and the mood around exporting chips and chip bans just kind of completely cools and China has an ask well then Trump can art of the deal and say hey uh you got to let in you got to unban Nvidia locally.
1:13:00
It is funny that we flip from from America banning Nvidia selling to China to China banning Nvidia selling to China and it's like they both at one point were banning but at different time time horizons and we're kind of going back and forth.
1:13:12
But anyway, uh I think Jensen will will figure something out.
1:13:16
Uh I this is not the end of that story for sure.
1:13:20
There's going to be there's going to be chips flowing flowing flowing flowing.
1:13:25
Um, uh, Near shares that you can just copy trade Leopold's fund and triple your money in a day. The forms are public. It's crazy.
1:13:34
The January 16th, 2026, $24 call that Leopold put on is up 200%. So, triple. >> Is that good? >> It's fantastic.
1:13:45
Says absolute kings I kneel.
1:13:48
Congrats on the extra billion.
1:13:52
Anyway, >> the uh Coachella lineup dropped two months early and over 60% of attendees were on payment plans last year.
1:14:01
>> What does that mean that it dropped early? >> I don't understand.
1:14:05
>> Yeah, just I think I think they maybe anticipate it's going to take longer to >> to sell all the tickets.
1:14:11
>> It used to be back in my day in the heyday >> it was uh Weekend One was always fully sold out.
1:14:17
sold out. they'd be the the tickets would be selling at a pretty significant premium >> premium >> uh to retail pricing >> and so uh and I think last year if I remember correctly you could just basically buy you could basically buy weekend one ticket at below retail
1:14:35
>> um >> I wonder if that's a I wonder if it's like a dynamic around like who is actually playing Coachella who's hot like are there are there is >> they drop the lineup you have >> is the lineup crazy >> Sabrina Carpenter and is headlining day one Justin Fever Day Two. >> Okay. But it's not Taylor Swift and the >> Okay.
1:14:49
But it's not Taylor Swift and the Backstreet Boys.
1:14:50
They're over at the Sphere.
1:14:52
So maybe the Spheres sucking the energy out cuz the Sphere tickets are selling great and they're they're super expensive.
1:14:57
So maybe people want want something new. Maybe Coachella's wash.
1:15:02
>> Could we do a live show from the Empire Polo Club this year?
1:15:08
>> I don't even know what that is.
1:15:09
>> That's where Coachella takes place.
1:15:09
It's on the polo field, >> but not during Coachella.
1:15:13
Just like at a different time or during Coachella.
1:15:15
You want to be on stage next to You want to be going up against >> I just want to be >> podcasting next to >> Yeah.
1:15:21
next to >> um interesting.
1:15:25
Uh Daario went out on stage and said he does not think we should be selling AI chips to China.
1:15:34
Quote, I think it's completely nuts.
1:15:35
It's very important that we defeat China in this technology.
1:15:38
Well, >> you got what you wanted because Beijing decided to make that decision themselves. >> Yeah. Dario.
1:15:47
Anyway, Adio customer relationship magic.
1:15:50
Adio is the AI native CRM that builds, scales, and grows your company to the next level. For free.
1:15:54
Also, >> team just went absolutely wild in here.
1:16:00
>> Uh, uh, Max Seagull over at Privy, our latest sponsor, uh, says, "Wow, OG Privy customer won an Emmy."
1:16:07
Uh, shout out Dylan Abberriscato and People Pleaser doing some serious damage getting onchain projects represented at the pinnacle of media accomplishment.
1:16:17
Uh, White Rabbit just became the first crypto project to win an Emmy.
1:16:21
Imagine telling >> outstanding innovation and emerging emerging media programming.
1:16:28
>> It's cool that they have a new form like a new award that has to be a new award, right?
1:16:31
That wasn't that award can't have been around for that long.
1:16:34
Um, but uh, People Pleaser says, "Imagine telling 2021 crypto Twitter that in a few years our producers and ETH would get a shout out on stage at the Emmys." Uh, pretty remarkable.
1:16:47
>> Let's pull up this video. >> Yeah, let's play.
1:16:50
>> Let's check in with Europe.
1:16:51
>> Let's Yeah, let's play the Europe. Wait, why Europe?
1:16:54
>> Oh, I I I was I was on to the next post. >> Oh, okay.
1:16:56
I wanted to watch the Emmys video if we have that one.
1:16:58
I want to see I want to see exactly how they frame this.
1:17:01
But we can watch the this >> I don't think that's the right video.
1:17:03
Yeah, I think that's just a different humanoid robot. >> Got the wrong guy.
1:17:07
>> Let's pull up the peopleleaser Max Seagull post and I'll tell you about eightleep. com. Get a pod five.
1:17:13
5-year warranty, 30 night risk-f free trial, free returns, free shipping.
1:17:16
It's good because you're home and you get to sleep in your eight tonight. I know. >> Finally.
1:17:22
>> It's always hard being away.
1:17:24
>> Uh anyway, can we pull up the Emmys video from the time?
1:17:30
>> I think that might have just been a screenshot. I don't people pleaser.
1:17:34
Okay, >> we'll circle back.
1:17:34
Let's pull up this video.
1:17:36
I put it in the chat of this uh humanoid.
1:17:39
>> This is Calvin 40, the humanoid developed by the French company Wandercraft.
1:17:43
And Chris says, "Europe is cooked." LOL.
1:17:47
I don't know why they're cook.
1:17:47
They're cooking if they're making robots. >> Let's see. >> Is it bad?
1:17:52
I didn't actually see the whole video. I hope it's not bad.
1:17:54
It doesn't Okay, it's picking up a tire. This seems fine. This is good.
1:18:01
This is good performance. What's not to like? Picks up a tire. This seems useful. >> Headless.
1:18:10
>> It is weird that it doesn't have a head.
1:18:11
>> I mean, >> it's not exactly Pixar level cute.
1:18:18
>> This is a literal clanker. It's clanking around.
1:18:21
>> One of the comments is, um, was this trained on folks over 40?
1:18:25
>> Yeah, it does seem very slow.
1:18:25
Uh, >> I mean I I swear almost every humanoid company to date has just been training on like Biden.
1:18:33
>> They haven't been training.
1:18:36
>> They're trying to train on Usain Bolts, but you know, it takes time to integrate all that.
1:18:42
>> Maybe Bolt is holding out.
1:18:42
He's saying you can't train. You can't train.
1:18:44
He's He's >> I'm I'm locking up my training data.
1:18:48
>> He's like, you got to pay up.
1:18:49
>> What do you think, Tyler?
1:18:49
I mean the difference between this and the the unitry demo or the demo on the unitry I don't think it was unitry that did that where it was fighting and then it fell down and immediately got back crazy the difference. >> Yeah.
1:19:00
And someone was saying that that robot was uh like not was being teleoperated.
1:19:04
Like there's there's someone in the background with like a controller controlling it.
1:19:08
And and Run had a good point that was like well like telling it to move forward is like the easiest part.
1:19:14
Like >> the hard part is like actually having a button to press to like pop up. Yeah. Exactly.
1:19:21
>> You're not mainly controlling that motion of >> No. No.
1:19:23
It clearly has some incredible incredible ability to like respond and understand its position and hop back up. Uh, it's pretty good. Well, I don't know.
1:19:31
We'll have to check in with Wandercraft, see how they do.
1:19:32
I I you know, this is better than nothing. I don't know.
1:19:37
I think they're >> It's a good start.
1:19:39
>> They could keep going.
1:19:39
Um, we have our next guest, Brendan from Meror coming in the studio.
1:19:48
>> Brendan, welcome to the stream. How you doing? >> I'm doing great. How are you? >> I'm great.
1:19:52
Sorry we couldn't uh link up on Tuesday.
1:19:55
You have some massive news.
1:19:57
Uh uh, introduce yourself. break it down for us.
1:19:59
What's the latest and greatest? >> Yeah.
1:20:01
I'm glad you finally, you know, it's good you get to the $500 million revenue mark and you decide, all right, I'll do some podcast. >> Yes, >> exactly. Yeah.
1:20:09
Um, so I'm Brendan, the CEO and co-founder of Meror.
1:20:13
Started the company with my best friends from high school when we were 19.
1:20:17
Um, scaled up a little bit to a million dollar revenue run rate. Dropped out of college.
1:20:23
uh started working with all of the AI labs and scaled from 1 to 500 million in revenue run rate in 17 months. >> Congratulations.
1:20:34
>> Which has been pretty wild. Uh yeah. No.
1:20:36
And super super exciting and surreal as you can imagine.
1:20:38
Um so uh very excited about all of the um progress with the business and the best is yet to come.
1:20:47
>> What was the first task you did?
1:20:47
Like what was the first data you labeled?
1:20:52
What was the first project to get from go to how do you go from zero to one basically?
1:20:56
>> Yeah, name every piece of data.
1:20:59
>> Well, so I remember the very first oh let's see there there were a few but the first meeting that really jumped out uh was when we were hiring Olympiad medalists uh because everyone was interested in how the models could become superhuman at Olympiad math.
1:21:14
And so we turned around 25 Olympiad medalists in 24 hours.
1:21:20
A and I I bring up this example cuz I wasn't completing the task successfully.
1:21:25
I was uh obviously haven't don't have an Olympiad gold medal in math.
1:21:29
But it was uh it was incredible to just see like how capable the models are.
1:21:34
uh a and this indication of the huge trend underway in the market away from low and medium skilled talent towards this super high-skilled sourcing and vetting paradigm ranging from Olympiad math all the way to the fang software engineers and top investment bankers and consultants that helped to push the frontier of models.
1:21:57
>> What was the what was the onboarding process for those 25 medalists?
1:21:59
Was this just like cold outreach or something like how did you actually meet these folks?
1:22:04
largely through referrals because we have a big pool of people that we've already hired on the platform.
1:22:08
And so, uh, the largest sourcing channel by far is that people we've previously worked with, we'll send the link to their friends and we'll pay them 250 bucks as for a successful referral to help grow the talent network.
1:22:22
>> It feels like most of the labs are I mean, we saw this with uh, OpenAI.
1:22:25
Mark Chen was just posting that they basically dominated at every hard programming and math competition this year.
1:22:33
Um what are the labs interested in next?
1:22:39
Yeah, I I think the largest transition is and we'll share more about this in one of our product releases soon, but it's away from academic evals like Olympiad math or GBQA for PhD level reasoning uh and moving towards all of these professional domains of how do we measure what it means to be a great software engineer to build products?
1:23:01
How do we measure what it means to be an investment banker that can do thoughtful financial analysis or a consultant that can help to uh you know segment a market and these endto-end evals over all of the professional capabilities I think will be one of the largest most exciting trends in the market over the next year or two.
1:23:23
>> What's the shape of that task then? I mean it's so vague.
1:23:26
It's it's so much like less quantifiable than what we what your score was on the IMO, although that's incredibly impressive if someone can do that level of math.
1:23:35
Um, go build good software feels really unverifiable, feels really broad. >> Exactly.
1:23:42
And so that's why you need humans to define the stasis points.
1:23:44
It's so much more difficult to measure.
1:23:47
And so one way of doing it could be to build a rubric where the model can use the criteria to score the deliverable that's being produced.
1:23:56
Like imagine you want the model to be really good at building a web app that looks beautiful.
1:24:00
You could have rubric criteria for you know all of the different elements of of said web app or or whatever um you're ultimately building.
1:24:10
And so having humans define the success criteria and using that those verifiers as part of in our environment to train models iteratively to learn how to optimize for those criteria is one of the enormous trends that we're seeing across all of the frontier labs.
1:24:29
How do you how do you sort of plan with the team when you have overwhelming demand for a current product yet simultaneously need to try to predict future demand in a way that I think normally when companies are doing traditional demand planning it's like
1:24:48
very clear of like just how many customers can we reach with our current set of products and future products but in your business it's not always entirely obvious what the needs are going to look like you know even a couple years out. >> Totally. I think that the most important >> Totally.
1:25:01
I think that the most important thing is always working on the frontier and understanding what are the leading indicators of what the entire economy is going to be doing soon and adopting soon.
1:25:13
And emphasizing that frontier in all of our product development and all of our investments is one of the most important decisions that we've made historically as sort of a a framework for resource allocation. That makes sense.
1:25:29
>> Uh we were just watching a video of a of a French company, Wandercraft, uh making a humanoid robot.
1:25:34
People were kind of joking about it.
1:25:36
I thought it was pretty impressive.
1:25:37
I haven't seen any humanoid robots out of uh out of France.
1:25:39
Um but uh can you talk about like what the data collection process for humanoid robots looks like now?
1:25:48
Yeah, specifically a lot of, you know, we were at Meta Connect yesterday talking with Zach and Bos and the team and a lot of people saw the announcement and they said, "Okay, a lot more people are going to have cameras on their faces soon." Yeah. How valuable is this?
1:26:01
Should should frontline factory workers be, you know, >> uh be collecting data today or or is that not necessary?
1:26:10
>> Yeah, it's interesting.
1:26:10
The key thing for the models to learn is having a clearly defined reward.
1:26:14
And so I'll give like a couple of examples of that and sort of the role that humans can play.
1:26:20
The first one actually without humans is that uh I was at the um this like robotics office where they had robots that were folding laundry and then they would have a vision model look at the laundry to see if it was folded properly as the reward.
1:26:34
So right having that stasis point where you can have a 100 different model trajectories see which five trajectories are right and then reward those trajectories so that the model increases its probability of doing that correctly in the future is very powerful.
1:26:47
All the way to another example where models are proposing scientific experiments and then we need humans that we hire as contractors to run those experiments in the physical world report on the results and say how they did.
1:27:00
And so I very much believe that models will learn from their experience and interacting with the real economy.
1:27:07
But so much of that experience similar to the way that you or I learn is curated by humans in the way that you know we help models with running the experiment in the physical world and giving feedback etc.
1:27:21
>> You tweeted the letters IPO a while back.
1:27:23
What did you mean by that?
1:27:26
Well, I did I did comment in parenthesis below that kidding. Uh >> oh, okay. I missed that part. I missed that part.
1:27:33
Okay, that >> it's all good.
1:27:35
Well, people people took it seriously because I remember that >> I'm sure you got a lot of like frantic calls from bankers being like, Brandon, you told me you'd tell me when you're ready. >> Yeah.
1:27:46
Well, it's funny because last year when we were a seedstage company, I tweeted IPO by end of year and everyone thought I was, you know, a little bit crazy because we'd raised our our $40 million seed round.
1:27:56
Uh or sorry, I tweeted Yeah, Unicorn by end of year.
1:27:59
And uh we, you know, ended up making it happen.
1:28:03
And so people thought maybe this time was was real, but I was I was just kidding about the IPO.
1:28:08
>> Got to keep >> What's the uh what what's the shape of the business now?
1:28:11
Obviously there's a huge amount of focus on on these expert networks and and these really highskilled specialized talent uh getting data from that and working through different problems and all the examples that you gave.
1:28:24
Uh is there still a need uh from big labs for just the more uh traditional RHF? Is this good? Is this bad?
1:28:34
thumbs up, thumbs down or has that been completely absorbed by uh just >> users of the of the >> the users or also just the models themselves like is is are GPT5 or or you know any of the other models the frontier models are they able to deliver the the the thumbs up thumbs down if you need to do some sort of fine-tuning on a specific uh problem?
1:28:55
>> Yeah, it depends on the lab.
1:28:55
There's definitely still large investments that are happening in RHF.
1:28:59
However, it seems like it's more efficient to collect those via data flywheels in the real world.
1:29:06
Uh, and where you really need expert human involvement that's incredibly valuable is someone that will think about a problem for 5 hours and come up with this like very structured framework for how to evaluate model success in a way that it's difficult to expect uh users uh of products to do in a reliable way.
1:29:26
a reliable way. Is that the is that like a reasonable way to think about a task a single task like five hours of >> that might be one way but one thing I'm very excited about is that the time horizons of agentic trajectories will go up dramatically right and so I think a lot of people initially think about AI
1:29:46
in the context of what can they see on their screen on chat GBT at any given time but over time it's and maybe using one tool with like online research uh but over time we're going to have the models working on problems that would take a human 30 days to do or 90 days to do that are using 10 different tools are interacting with various employees in
1:30:06
the companies and we need environments for all of that right we need ways to eval to measure success to define the rewards and that's going to be a very exciting problem space to continue pushing the frontier of >> uh how does how do you or just humans generally fit into solving the problem of like booking a flight or uh or you know ordering Door Dash we've heard
1:30:30
about these uh simulated environments RL environments verifiable rewards like uh is it is it mostly designing the environment with the reward uh or is there actually a process for um someone who's just a fantastic travel agent to uh you know create a a rubric or create a uh or just actually do a ton of tasks to generate data. Yeah, it could be. So, Yeah, it could be.
1:30:53
So, one way you would do it for those kinds of browser use workflows is that you could have a simulated application and then a unit test that measures if the model effectively, you know, completed the task.
1:31:11
It changed the state and like booking the applica the flight or whatever the action is.
1:31:15
And ultimately, you do need a human expert to help write what is that unit test.
1:31:21
Um, but my guess is that computer use will be solved relatively quickly in the next like or at least in the next like two or three years and then there's going to be this much longer tale of sort of the broad space of knowledge work and everything that we want to do of how do we get the model to build a startup or uh help help uh prep for a podcast episode or or whatever the workflow is. >> Yeah.
1:31:49
How do you think about uh solving problems that take like decades?
1:31:53
Like I I always go back to like you know health there's certain things where you know the the FDA does a lot of work to try and understand uh you know if you're consuming this particular ingredient for 50 years how does that affect you?
1:32:07
It feels like until we can simulate the entire human body and and run it at a faster clock rate, like that seems like something that you just can't really short circuit.
1:32:18
We're already getting to like longer and longer rollouts, and that feels like that might be some sort of like damping function on how quickly we can compound.
1:32:26
But are there any promising uh strategy that you've heard for dealing with uh problems that just take a long time to actually understand the reward or understand the did the past fail?
1:32:40
>> Yeah, the concern I would have with that scenario is it's so difficult to perfectly simulate like the human body and how that would play forward.
1:32:47
And so my guess is that for a very very long time models will more so look at empirical analysis.
1:32:53
However, models might survey people that took certain uh you know, vitamins or or whatever drugs uh when they were a certain age and see the impact that that's had in their analysis, but it'll it'll be difficult to simulate in that case.
1:33:11
There are other cases where it's like a well scoped physics simulation of how well uh does you know this like ball roll down the plane or or whatever uh we're modeling out.
1:33:24
that'll be easier to simulate uh and therefore for models to have an accurate understanding of how things will play out in the real world.
1:33:34
>> What's uh are you guys uh naturally 996?
1:33:39
Do you do you I always get this question but but I but I feel like you guys are probably more on like a >> like 997 like six is almost for the week.
1:33:48
you guys are 20 21 like what what else do you have to do besides uh besides this?
1:33:55
>> Well, it's funny because we have actually never really mandated ours the company.
1:34:00
It was just >> that's I figured but there's like if you're ramping revenue from one to 500 million and >> it's a lot of work. Yeah.
1:34:06
And a little bit of time. >> Exactly.
1:34:09
So I think it the way it started was our initial core team was working like seven days a week.
1:34:14
um and everyone was in the office like super late, all this stuff.
1:34:18
And so we initially gave that rough guidance because we wanted people to have a little bit more balance and going home earlier, etc.
1:34:25
Uh but obviously as the companies developed, I think it's important to be able to hire people that have families and even though they'll work hard on the weekends, they might still be at home.
1:34:37
And so there's um there's some of that as well.
1:34:40
Still emphasizing a lot of intensity and you know moving mountains for customers but at the same time not necessarily being as input oriented and and much more focusing on outputs.
1:34:53
>> How big is the team now?
1:34:53
Like the full-time core not the network >> relatively large. We're 250 people now. >> Wow. Congratulations. >> Fast. Yeah exactly.
1:35:01
Um across the US and India.
1:35:05
um and then a little bit across Latin America and the UK as well.
1:35:08
But um yeah, it's it's been exciting.
1:35:11
Certainly a crazy feeling to start having all these people in our our new SF office and and new faces that I have to meet.
1:35:21
So, um lots of >> What kind of predictions are you going to make?
1:35:24
Uh uh any numbers you're throwing out?
1:35:27
You you said unicorn by end of year last year, but what about what about going forward?
1:35:31
anything you're willing to I'm willing to to take >> we could say duckorn by end of year this year so >> feels like you might uh might have already had it in the bag but >> yeah we'll have to call R for rock to get this congratulations we'll talk to you soon on Bye >> next we have Darren Mauy from Google Cloud coming in big Google announcement today big event uh over at Google in the uh global startups at Google Cloud World. Let's bring in Darren. How you doing, Darren? Good to see you. >> Good to see you guys. Thanks for having me.
1:36:14
As you can tell, I'm in a pretty cool spot here in Mountain View, so it's great to be able to spend a few minutes with you.
1:36:19
>> Thanks so much for hopping on the show.
1:36:20
>> Looking sharp as well.
1:36:20
Thank you for wearing wearing a suit. >> Thank you.
1:36:23
I had to brush off the suit.
1:36:25
I had to kind of dust off show somewhat presentable today, guys. >> Yeah, it looks great.
1:36:28
uh take us through uh what what uh the event, what's going on today, what's been announced, who's there, everything. >> Yeah, that's great.
1:36:35
So, we're actually in a in a couple hours going to be kicking off our first global AI builders summit.
1:36:41
And so, we're going to have a couple hundred founders and builders here in this room, which I can tell you about if you're interested.
1:36:46
A little bit of cool Google history here.
1:36:47
And then we're going to have um a few uh actually tens of thousands of startups and builders around the world joining us digitally as well.
1:36:54
So today we're having customers like Lovable, Replet, Fireworks, and others on stage talking about, you know, the problems they're trying to solve, but how AI is actually helping them complete, you know, completely revamp the industries that they're in.
1:37:08
We also have some Google Cloud and some DeepMind folks joining us as well to talk about kind of how quickly we've seen the evolution, where we are now, and we'll look around the few corners into what's coming.
1:37:17
So some really uh great sessions today over the next few hours.
1:37:21
So we're we're excited about that. >> That's great.
1:37:23
Tell me about the history of the building.
1:37:24
You said that there was something special about it. >> Yeah.
1:37:27
So, we're actually in this place called Charlie's Cafe, believe it or not.
1:37:30
And you guys probably know Google has been known for having good food, good cafeterias for a long time.
1:37:33
This was our first cafeteria named after our 57th employee, Charlie.
1:37:38
And so, in all seriousness, although it is a little bit of an urban legend, we actually do deeply believe in creating great spaces for people to come together and challenge each other, have good conversation, and eat a little bit of good food, too.
1:37:51
I'll definitely admit that.
1:37:53
But we're here in this space.
1:37:53
We thought it's a perfect spot to kind of bring everybody together, talk about building and innovation.
1:37:58
So, uh, it'll definitely be a great few hours.
1:38:03
>> What, uh, give give us a high level on what the last year has looked like in your role specifically around, you know, supporting startups that just have an insat insatiable demand for uh, inference and uh, you know, all the other infrastructure needed to scale these types of applications.
1:38:22
No, it's a really good question.
1:38:22
You know, you guys um should know I've been in cloud computing for a while.
1:38:26
I was at another hyperscaler for a long time at the early days of cloud and I thought we were moving fast then until we've entered this AI revolution, right?
1:38:34
Which I think the compression and the speed is like nothing I've ever seen before.
1:38:38
Um over the last 18 months, we've definitely felt what I think we would really call like a seismic shift, frankly.
1:38:45
Um the old school way of thinking about cloud through the lens of infrastructure as a service.
1:38:49
All of a sudden, people talking about all layers of AI from chips and infrastructure, right?
1:38:55
Are we going to use Nvidia chips?
1:38:57
Hey Google, what's up with this TPU concept that Anthropic is relying on?
1:39:01
Kind of what does that mean?
1:39:01
Kind of what does that mean? When you get to the model level, you know, the fact that Gemini and what we're releasing with VO, these are first class citizen models as I think you guys can see in terms of performance, cost efficiency, but the fact that we're
1:39:13
building and going to continue to innovate on our models, but also partner with, you know, folks like Anthropic, folks like Meta to make sure Llama and Sonnet and others are also first class products that startups that are building on Google are able to use in a super integrated fashion. And then this
1:39:27
integrated fashion. And then this concept more recently right which is this agentic this agent you know agent AI this is not a theoretical pursuit I think it's interesting that when we talk to founders and they give us some feedback the feedback they're telling us is the agent capability from Google cloud is a real capability it's not a
1:39:45
planned release of products hopefully in the future right it's a fully integrated stack where startups have an SDK and an ADK they can build these agents as I said using firstparty and thirdparty models they can publish them distribute ute them and we can even help them go to market right so to this to your point
1:40:03
the last 18 months break neck speed I think we're all learning quite a bit as we go but I would say the feedback that we're getting from founders and builders especially are a telling us we're on the right track doing the right things and frankly they're saying go even faster right help us do even more even more quickly so uh with nano banana you guys
1:40:22
may have seen that was released recently these are things that are coming out of our deep mind team uh and there's going to be further announcements again around agents and new models fairly shortly that are going to keep us very top of mind and very much a part of what startups are building every day. >> Yeah, it's fantastic. Um I Google's >> Yeah, it's fantastic.
1:40:36
Um I Google's always been very uh is a great partner to the startup ecosystem giving out credits.
1:40:44
Uh do you feel like startups are burning through credits faster now in the AI age?
1:40:48
I feel like uh I remember going through Y Combinator and getting some huge amount of credits and not really knowing what to do with them. Exactly. Yeah. Good.
1:40:57
Well, they usually expire after a year.
1:40:59
Um, but I feel like with with AI, like you can you can actually accelerate much faster.
1:41:04
What's the mood been like from startups that you work with?
1:41:09
>> No, it's a really good point.
1:41:09
When you think of our Google Cloud for startups program, which does have a credit component, an engineering component, a support component, training, etc.
1:41:16
We definitely have had a dramatic increase in the amount of startups coming to the platform.
1:41:22
So, that's first and foremost a really good signal.
1:41:24
To your point though, which I think definitely shows you understand this space is getting into these programs is one thing.
1:41:29
Even being approved for credits is another thing, but actively consuming the credits and building value.
1:41:34
That's what the startups care about and that's what we care about. Right?
1:41:38
So what I've seen now over my almost five years at Google Cloud is in the last 18 months, not only do we have more startups than ever in the program, they're consuming the credits extremely quickly.
1:41:49
credits extremely quickly. And more importantly for us at least and I think for these startups they're staying with us right this concept of I'm going to jump from one platform to another to use credits worked in a commoditized cloud world where you could go from a virtual
1:42:03
machine to a virtual machine to a virtual machine now that I'm able to come to these startups and do what we talked about a moment ago of GPU TPU Gemini claude sonnet llama agents wrapped in Google cloud we're finding these startups are using the credits but then they're like I'm not going anywhere Right. And so I think that is the true
1:42:19
And so I think that is the true business case and value proposition behind these credit programs. >> That's great. Uh anything else, Jordy?
1:42:28
>> No, thank you for joining and uh have fun out there.
1:42:30
Wish wish we were uh going to be able to catch some of the talks ourselves, but uh we'll have to be there next time.
1:42:36
>> Yeah, we'll catch up with you soon.
1:42:37
Thanks for taking the time.
1:42:38
>> Yeah, that's all right. Great seeing you guys. Have a good day.
1:42:40
>> We'll talk to you soon. Cheers, Darren. Have a good one.
1:42:42
>> You up next, we have Kavon from Macroscope coming in the studio.
1:42:44
Also the founder of Periscope.
1:42:47
Um, we will bring him in in just a minute.
1:42:50
In the meantime, let me tell you about public.
1:42:54
com investing for those who take it seriously.
1:42:57
Multi-asset investing industry leading yields are trusted by millions. I like noise. Thank you, Jordy.
1:43:04
Uh, what were you about to say?
1:43:05
>> Uh, Periscope was acquired by Twitter. >> Yes.
1:43:10
>> And did that become >> live streaming? >> Live streaming.
1:43:13
>> We get it from Kon directly.
1:43:13
Let's bring him in to the TBP Ultradome from the Reream waiting room. Kavon, how you doing? >> Welcome to the show.
1:43:21
>> Hey guys, great to meet you.
1:43:24
>> Clarify Jord's question.
1:43:24
Did tell us the story of Periscope.
1:43:27
We'll work through to Macroscope, but I'm super interested in in some of the Silicon Valley lore here. Yeah, just kick us off.
1:43:34
>> By the way, huge opportunity for somebody to buy up all the scopes with scope at the end. Just every English word.
1:43:40
So your next company you're going to be like h I got to I got to keep double down.
1:43:44
>> I might have I might have already done it. You never know.
1:43:47
>> I got to I got to tell you actually funny funny thing.
1:43:49
9 years ago almost to the day 9 years ago we launched a feature called Periscope Producer >> which is the same infrastructure that is powering this very broadcast right now.
1:44:00
And our our dream when we built Periscope Producer was literally for a show like TVPN to exist.
1:44:04
It's sort of like that perfect blend of eye production live streaming content blended with the conversation of what's happening on Twitter.
1:44:14
>> And you know, it took a while for that to to come to fruition, but it just like makes me so happy and proud in a very emotional way to see everything you guys have done and to see it happening on on Twitterx. Um, >> that's awesome. >> That's awesome.
1:44:26
Congrats on everything you guys.
1:44:27
Thank you for for building the bedrock.
1:44:30
>> Yeah, I mean I I I want to get to macroscope, but tell me more of the lore.
1:44:33
Uh what was the what were the early days of Periscope like?
1:44:35
Uh what was the first like go to market motion?
1:44:40
There's this there's all these famous uh back then there was the era of like go to South by Southwest.
1:44:44
This is the story of uh of a bunch of uh social apps including Twitter uh where you get the early you kind of create a ground swell of tech early adopters.
1:44:53
What what was the first product build?
1:44:55
What was the story back then?
1:44:57
Uh, how'd you launch the product?
1:45:01
>> Well, I mean, if you want to go way back, the very first version of Periscope, A, it wasn't called Periscope. It was called Bounty.
1:45:05
And B, it wasn't it wasn't actually live streaming.
1:45:09
Our first prototype was essentially um I sort of think of it as like a reverse marketplace for for Google Maps.
1:45:14
Like you would you would drop a pin somewhere in the world and someone would take a photo.
1:45:18
someone would take a photo. you would you would you would have some prompt like you would put a bounty on you know what's happening at the Tokyo fish market right now and someone would respond with a photo um and that to us was like a really cool way of trying to attempt to build like a teleportation device um we didn't know how to we didn't know how to actually build
1:45:35
teleportation so many fascinating like social app social mobile local apps at that time it was a big it was a big trend and there were a whole bunch of different ideas that were experimented with it was such a fascinating time
1:45:46
>> well and it's yeah interestingly enough like Instagram has like delivered effectively that functionality now or Snapchat where you can teleport and see how what's the vibe at this restaurant right now. >> And if you click on the location that
1:45:56
>> And if you click on the location that you tagged, I can see who else put pictures there if they're public.
1:45:59
So yeah, those features you were clearly very early.
1:46:04
>> Yeah, I think Snap Map actually was like probably one of the best manifestations of that early on.
1:46:07
The our stories feature really brought that use case to life for Periscope's journey.
1:46:11
You know, we we when we built that prototype, we realized it a just wasn't really interesting and b you have this like liquidity problem if you're just dropping pins randomly in the world and c it didn't really feel like teleportation because of static photos by definition old by the time you see it.
1:46:26
>> So that's that's when we were like let's flip this and make it press a button go live and and rather than using static photos, let's try making it live video.
1:46:34
And so that was one key thing that we did.
1:46:37
And then the second key thing that we did that really made Periscope click was the the floating hearts.
1:46:40
I don't know if you guys remember, but like it was the first social network we had seen where you could have an infinite form of expression.
1:46:48
It wasn't just like pressing a button to like it.
1:46:50
It was sort of an infinite infinite amount of love. >> Yeah.
1:46:53
>> Um and you know, we had this beta of 20 users.
1:46:57
It just like our friends basically and some you know, family and investors.
1:47:00
And when we when we shipped that version of the of the build that had live video with the floating hearts, it just it just was so clear to us that there was something here.
1:47:09
Um, and one of the early beta users happened to work at Corp Dev at Twitter.
1:47:14
She invited Jack and Dick who was the CEO at the time and that sort of like put us on the radar uh with Twitter and so we actually ended up getting acquired before we launched the product.
1:47:24
like there was no go to market motion that got us and it was just like a beta of 30 people and Twitter was like the 31st user.
1:47:32
>> Um >> that team was crazy about the early acquisitions.
1:47:35
I mean Vine had launched but they acquired Vine as well.
1:47:37
They were like very aggressive about picking stuff up early.
1:47:41
>> Top tier picking companies, not top tier at landing the plane on the integration. That's my TLDDR.
1:47:48
>> Someone called it a clown car, but we're not going to go into that.
1:47:50
Someone you met with yesterday called it a clown car that fell into a gold mine.
1:47:55
>> Gold mine, which is a great a great quote.
1:47:57
Anyway, um uh I want to know like uh what do you think about live stream monetization?
1:48:04
Like we've we we have we have ads running on a ticker.
1:48:06
We've we've we've uh we've brought through a bunch of like the TV era aesthetics, but then we also do just host red ads.
1:48:13
Um we don't really that I'm aware of get a big share of like programmatic ads.
1:48:19
We don't do I know Twitch you can do like I'm going to an ad break and it will play programmatic ads. We haven't done that.
1:48:26
Did uh we've heard a ton of stories about Tik Tok shop and the live streaming sales stuff.
1:48:31
Uh we've joked that we want to be like that for enterprise >> live commerce for macroscope. I got one license. I got one license. >> Got one license here. Pick it up.
1:48:41
>> I got five seats left. Buy now.
1:48:41
I got 25 credits over here. I got an SDR.
1:48:49
I I think you guys are an interesting place where you can you can sort of benefit from all of these models, right?
1:48:54
You can do the pre-roll, you can do the midroll, you've got obviously incredible brand placements from some of the best tech companies in the world.
1:48:59
tech companies in the world. Um and then I think also you have a unique vibe going where you can benefit from a lot of the monetization techniques that like have been become popularized on you know uh Twitch and Tik Tok and you know even Periscope early on we had this thing
1:49:14
called super hearts which was like people could pay for inapp purchases to you know fly Ferraris off the on the screen or whatever like I think you have enough super fans that watch the show that there's like an enduser monetization component on you know in addition to what you're able to do with big brands. I don't think there's many
1:49:29
I don't think there's many types of content that can benefit from all of those forms of monetization.
1:49:33
all of those forms of monetization. Did you uh did you feel like you were at at at what point like now it feels like Periscope was like extremely early even though even though like like I think people anticipated live streaming would be big but I feel like only in the
1:49:51
last few years people at least the tech world has like woken up to how big live streaming is even outside more more so outside of tech right >> and on mobile too I mean that that that was one of the unique insights like there there there were there was a there was a host of companies that were really
1:50:06
focused on solving like I mean Flickr existed and then Instagram was huge and then uh there were a whole bunch of video apps Vine one of them like YouTube existed but no one had cracked it on mobile and it required actual deep insight into the user experience and also the engineering to understand how
1:50:22
to get it to work on a phone which wasn't as powerful as a laptop back then but yeah >> yeah I mean I think there's a lot of tech there were a lot of technological problems to making uh mobile based live streaming work well at the time a lot of those problems are just solved now and and somewhat commoditized. I mean,
1:50:36
I mean, there's just like SDKs that let you do this really easily.
1:50:39
I think my my big takeaway, and you know, call me somewhat jaded on this.
1:50:44
Um, but I I I think what we learned the hard way is that a live focused social network on mobile that's like short form live video um isn't tenable, right?
1:50:55
And that's what Periscope was.
1:50:56
It was live only as distinct from like Instagram or Facebook live at the time.
1:51:00
that was live was a feature amongst the social network that let you communicate and keep in touch with people asynchronously.
1:51:06
And I think it took us longer to build async forms of connection than it did take Instagram and Facebook building all of our features into their existing platforms.
1:51:17
And so that was our sort of like hard lesson learned.
1:51:19
Um because you know we had a parallel track where we were trying to make Periscope integrate into Twitter.
1:51:24
That was the whole thesis of the integration and it just took us way too long for for reasons that we can get into if you're interested.
1:51:29
um to to make that integration come to life.
1:51:31
And so as much as we you know we grew from zero to 100 million users in like a year and a half. It was insane.
1:51:36
But just everyone else built all the feature >> Yeah.
1:51:41
>> Everyone else built all the features um you know uh quickly. >> Yeah.
1:51:46
What what do you think about the lack of uh Sorry, last question on Periscope.
1:51:50
Just >> talk about macroscopes. I know.
1:51:51
But uh uh lack of screen sharing API on mobile.
1:51:57
I felt like uh during the Clubhouse era that was something that was sort of missing was uh like you go to Twitch, yes, you're watching someone live stream, but a lot of the work is done by the video game that they're playing or the video that they're reacting to.
1:52:11
And being able to put something else on the screen so that someone doesn't need to just stand there and do an eight hour standup routine with no support for eight hours straight, that helps.
1:52:21
And I felt like Apple kind of nerfed that or never really and maybe it was just a hardware thing, but uh what was your take on like like how important that was?
1:52:31
Am I misunderstanding that?
1:52:31
And and like how how how how would it have played out if it was easy to screen share?
1:52:38
>> Um I don't know the state of the current APIs, but I know at the time and this is probably like 2018 I want to say we did a lot.
1:52:44
We we actually built a bunch of integrations that let you share your screen including from mobile.
1:52:50
>> I think just the reality is it's such an edge case.
1:52:52
relative to what like the the vast majority of the use case for our product was people just was people talking to people, right?
1:52:58
It was like 98% was that type of broadcasting and 2% was what you guys are doing which is like professional broadcast whether it's from the NFL or TVPN or anything in between.
1:53:09
>> Um and so I I just I don't think that would have had a material impact for us as a use case.
1:53:14
>> Um but I don't know if the if if Apple if Apple did indeed nerf those APIs.
1:53:19
That's that's news to me. >> Yeah.
1:53:20
I talked to a YC company once that was trying to do uh mobile Twitch.
1:53:22
So you would screen share from the phone.
1:53:27
Now people do that with like you take the video feed out of USBC, you route it through a PC.
1:53:31
There are huge mobile gaming Twitch streamers, but they basically like are screen recording with a third party device. It's very complicated.
1:53:40
It's not something they can do on the go. Um anyway, sorry. I want to move on. Jordy, what do you got?
1:53:44
>> I I I I want to continue that conversation. Limited time.
1:53:46
Uh let let's yeah let's switch gears to macroscope.
1:53:51
What uh give give us the I don't know give us a hybrid investor slashc customer pitch.
1:53:56
I want kind of a bit of both kind of long-term vision as well as like why somebody should sign up today. >> Yeah, totally.
1:54:03
So I'll start with the like the sort of customer focused angle because it's it's I think what resonates the most with me.
1:54:09
You know, we think of Macroscope as um X-ray vision for your company.
1:54:14
You know, we help you understand what's happening.
1:54:15
um how's the product changing?
1:54:17
How's the codebase evolving?
1:54:19
What's everyone working on?
1:54:21
But just answered automatically um and answered via the source of truth, which is the codebase.
1:54:26
If for any company that builds software, the source of truth is the codebase.
1:54:29
If it's not in the codebase, it hasn't happened yet.
1:54:30
And if it isn't in the codebase, we you know, AI and state-ofthe-art LM can do a really good job of articulating how things work, who did it, when it happened.
1:54:38
happened. Um and sort of our observation having worked at many companies both small startups that we've started and you know very large companies like Twitter is it's actually extremely hard to answer these basic questions like the classic what did you get done this week which is ironically very relevant to Twitter's history is something that every leader thinks about constant like
1:54:58
so much of my job as a head of product at Twitter was literally just understanding what the people were working on >> um and usually it's like the state-of-the-art solution to this problem is meetings issue management systems, spreadsheet trackers, just bugging engineers and asking them and sort of multiply that out by an organization that's hundreds if not thousands of engineers. There's a lot of
1:55:16
There's a lot of human capital waste that goes into this problem.
1:55:19
Um, and so our thesis is that this is silly like in a world of LM um you know there's a lot of amazing AI tools tools built for engineers not a lot of great AI tools built for leaders and so that's what microscope is trying to do.
1:55:32
trying to being be an an understanding engine for your company that simultaneously gives leaders clarity while saving time for engineers, right?
1:55:41
It's sort of this interesting hybrid where we're solving problems that are paper cuts that engineers, you know, feel 50 times a day, whether it's automating their PR descriptions, doing AI code review, avoiding them having to go to status updates or write status updates, which like there's nothing an engineer hates more than, you know, getting distracted from building something and instead reporting status through some game of telephone.
1:56:00
Um, and so we are simultaneously helping the leadership team get automated visibility while saving engineers a bunch of time.
1:56:09
And we think that like I mean we're obviously biased but like there's just no way every company in in 5 years like every company is going to have a tool like this whether it's macroscope or some other tool.
1:56:18
It is just complete insanity to imagine that we are doing this the oldfashioned way.
1:56:24
>> So >> so when when did you actually start the company because you announced around yesterday with light speeded our friend Michael but uh I imagine and you've been at it for a while.
1:56:34
>> Yeah we started the company in July of 2023.
1:56:36
raise a seed round from um uh our mutual friends at Thrive Capital and Adverb and GV and some amazing angels.
1:56:45
Um nice and then and then yeah >> but to go back at that point in time at that point the the the the AGI pill folks were saying AGI by 2025 SAS is not going to matter fast take off.
1:57:01
So did you always did you never lose faith in enterprise SAS?
1:57:06
It feels feels like you had you had conviction that this type of thing was going to be important for a long time.
1:57:14
>> I I think there's a lot of dramatization that goes into like the shifting of the you know landscape and they make for great headlines and all that.
1:57:21
I actually think that as every engineer gets turbocharged by AI and as like coding agents completely revolutionized how software gets written.
1:57:30
software gets written. I think this problem the problem that microscope is solving only becomes more important right like if if we if companies are writing 10x more code and humans are writing less of it and humans are reviewing less of it then it becomes even more challenging to understand what's happening and ultimately like humans are still accountable for the outputs of what a company is shipping
1:57:51
and building and so I think having this AI air traffic control system and understanding engine for what's happening with your company becomes even more imperative um so I the company's built like assuming that agents will get better and better and better and that humans will just stay at this like more and more at this global level of just kind of witnessing okay what are what are all my play what what's my whole
1:58:13
team doing what what what's this player doing it doesn't really matter if they're a person or or an agent >> yeah today it's like 95% of the use cases are what are my humans building assisted with AI and in tomorrow five years from now it might be you know 90% of it might what are what have all my agents produced and maybe 10% of that is like what have humans produced but I think the the problem that's being
1:58:37
solved is still fundamentally the same which is what changed what impact did it have um and um you know and where do we go from here like that's that's a neverending thing that is the highest leverage thing a leader whether you're an engineering leader or a product leader or a CEO like those questions are always on your mind >> how do you think about the level of integration into the systems that you want to pull data for. Like I could
1:59:00
Like I could imagine like a Slackbot that talks to every it effectively acts as a middle manager literally asking people what did you do this week and then they write their little status update and that gets rolled up and then that can be queried.
1:59:13
Uh, I could also imagine something like, you know, a screen recorder super integrated into everything the employee is doing and you can query, hey, did my employee do send this email that you have full transparency and then a wide swath of tradeoffs in between for the level of abstraction integration you want into the systems. >> Yeah.
1:59:33
Well, like the first thing I'll say is like we're not we're not big fans of the there's a surveillance state. >> Yeah.
1:59:41
We're not fans of that angle and like the last thing we want to build is a is a spy tool.
1:59:44
a is a spy tool. So we don't we don't imagine you know doing screen reporting or anything like that but I do think having sort of extensive integrations with the the stack that a company uses to build and manage their product is really important like today we started with a few systems we integrate with
1:59:59
GitHub we integrate with your issue management system so whether you use Jur linear we integrate with Slack >> um and we think those are the critical starting points but it's just the beginning because you know the codebase can tell you what you did and how it works like how does our billing system work the code can answer that question. It can't tell you why you did something
2:00:16
It can't tell you why you did something like what customer problem we were resolving with this feature is not in the codebase, but it's probably in a Google doc or a notion.
2:00:22
Um it might be in in a linear ticket.
2:00:24
Likewise, like >> who can see this feature is not necessarily in the codebase, right?
2:00:29
You might have a launch darkly flag or a static flag that tells you, oh, this is available to 1% of users in Japan, but like the the codebase can be the glue that then stitches into all these other systems.
2:00:40
And we're building macroscope in a way that allows it to be a conversational interface to all those questions and answers.
2:00:47
Like today we have a Slackbot that lets you ask a question like the one I asked.
2:00:49
Like have we launched this feature? If so, who can see it?
2:00:52
And we imagine over time building all these other integrations that let you essentially get more insight into what's happening and how things work.
2:01:00
>> How do you how are you kind of uh setting goals with the team and forecasting with Periscope?
2:01:05
you built like a viral consumer app.
2:01:07
And so uh and and today even in uh developer tools like there's this intense like you know a lot of companies are growing ridiculously fast pretty unprecedented for for B2B products.
2:01:22
Uh, and so there's this intense pressure to like, you know, show massive traction and adoption quickly, but Macroscope feels like a pretty complicated product that you're still going to need to be like iterating around and and figuring out where different types of companies are getting value.
2:01:39
So, how do how do you how do you kind of like set goals with the team and and what is success look like over the next 12 months?
2:01:48
>> Yeah, I mean, it's a good question and it's obviously early days for us.
2:01:50
If you really sort of simplify our product down into two components right now um there's really two pieces.
2:01:57
really two pieces. We have a code review feature um which is um relatively speaking it's a it's a more mature space right like we are not the first code review tool um but code review is an enormously important problem for for companies to solve and I think that the sort of like heruristics around whether
2:02:14
our product is working well for them or are are much easier to quantify right like we we released a benchmark um as part of our launch yesterday which sort of is one indicator of what we use to evaluate whether our code review tool is working like how what percentage of bugs can detect um in a customer's pull request. And so like we think about
2:02:29
And so like we think about goals very differently for a product area like that where we can measure ourselves very in a very quantifiable way relative to competitors um than we do the other part of our product which we sort of refer to as status.
2:02:41
Like we help you understand the status of anything happening in your product development process.
2:02:45
That's way more green field, right?
2:02:47
Like it's we're not aware of really any other products like technological solutions to that problem.
2:02:53
And so both our road map and how we think about goals for for that part of the business is a little bit more greenfield.
2:02:59
We're just sort of excited to push the envelope on where this goes and how we can solve bigger and bigger problems for customers.
2:03:06
>> Talk to me about where the budget is coming to buy macroscope.
2:03:08
It feels like with even zooming out broader to just the AI enabled SAS market, there's there's a lot of products that don't neatly fit in with okay, I'm going to rip out this and replace.
2:03:23
It's a lot of adding something on top.
2:03:26
We're seeing a lot of like the SAS apocalypse, the seatbased models going away, but it feels like it would be hard to quantify this with like value based pricing.
2:03:34
How are you thinking about justifying a budget internally if you're dealing with a larger customer who's trying to kind of underwrite the value that Macroscope brings relative to the cost? >> Yeah.
2:03:48
Well, so our you know our buyer is I would say half the time it's the engineering leader.
2:03:52
So this would be like a CTO or head of engineering and the other half of the time it's the CEO.
2:03:57
Y >> and you know I think I think um from a from a value standpoint like if you're an engineering leader you want your team spending as little time reviewing code um and and as little time dealing with production issues as a result of shipping bugs into production as possible.
2:04:13
And so like the the value of even catching one production incident from an AI code review tool um I think is intuitively very easy to understand.
2:04:23
like we don't we don't see customers asking the question of is a code review tool valuable what they want to know is like why is this tool better than all the competition um so I think you know and that's just going to become that's going to be more and more true over time just given what's happening in the AI coding landscape I think for the other
2:04:38
aspect of our product which is the sort of understanding layer it is you know relatively speaking it's it's it's green field right there is no product solution they're ripping out with ours and so what we've seen resonate with with our customers is you can you have an intuitive feel for how much time your team is spending in meetings and dealing with all the work around the work. Um, and so I think the the the
2:04:57
work. Um, and so I think the the the value proposition really is um do I do I buy that this tool is going to help my team focus more on building and less on doing that work around the work um and is that is that worth the the you know the the price of entry which from our standpoint like again the questions that our customers are asking is not like is this valuable the question is does it
2:05:19
work the way you say it does um and but obviously that's where we have to do our job well um but I think anyone who's worked at a big company and has you know invested in solving this problem the oldfashioned way knows that it's like the worst part part of working at big companies and they would gladly pay any amount of money to solve the problem if it actually works. >> Yeah. How do you think about uh >> Yeah.
2:05:37
How do you think about uh generative UI or actually like I could imagine some people want text result of they want to chat and ask what you know how are things going.
2:05:47
Someone in the chat said, uh, this is more of a progress bar.
2:05:51
And I could imagine someone wants to see a dashboard, a progress bar, stats.
2:05:55
Like, do you think in the future you'll be able to like instantiate exactly what the particular manager wants to see kind of like a a dashboard of of what's going on in the organization?
2:06:06
>> I think there's there's lots of interesting vectors here.
2:06:07
One is just often times we're describing in in words how a product is changing.
2:06:14
And there's nothing more powerful than just showing how the product is changing, right?
2:06:17
So whether that's like integrating with Figma and showing you the intended mockup that just got shipped or whether it's actually running the code.
2:06:22
Like a lot of time, like think about what an engineer does when they ship a feature.
2:06:27
They ship the feature, they go into Slack and they literally record their local branch and like a mockup of the product and they post it in Slack and say, "Hey, I just shipped this thing. It merged into staging."
2:06:34
We should just automate that, right?
2:06:37
we should literally automatically run the product and show you the thing that just got shipped and save the engineer the time from having to do all that.
2:06:44
So that's like one angle that this can take and the other is sort of what you what you were saying um which is kind of like in the appropriate time generating dashboards or some other visual manifestation of some status update.
2:06:56
I think all of these things are are possible.
2:06:59
We have to sort of pick our punches in terms of where we start.
2:07:01
Um but um but yeah, humble beginnings.
2:07:06
Well, thanks so much for coming on the show.
2:07:07
It was great catching up with you.
2:07:09
>> Come back on again anytime.
2:07:10
>> Yeah, I I love I love Silicon Valley lore and I I'm very excited about what you're building. Congratulations.
2:07:16
>> Thanks for having me, guys.
2:07:17
>> We'll talk to you soon. See you. >> public.
2:07:19
com investing for those that take it seriously.
2:07:21
I already told you about public.
2:07:23
So, I'm going to tell you about. com.
2:07:24
Out of home advertising made easy and measurable.
2:07:27
>> I time the double call button perfectly.
2:07:29
>> You did say goodbye to the headaches of out of home advertising.
2:07:30
Only adqu combines technology out of home expertise and data to enable efficient seamless ad buying across the globe.
2:07:35
We have our next guest in the reream waiting room makund from emergent AI. How you doing? >> Hey, I'm doing good. How are you guys? >> Welcome to the show. How you doing? >> What's happening?
2:07:48
>> Super excited to be here.
2:07:50
>> What's going on today?
2:07:52
>> We actually launched a product three months back and one of our goals was to be on TVPN. Super. Yes. I love the show. >> Fantastic. Thanks so much. uh break it down for us. What are you building?
2:08:03
>> Yeah, so I'm mukun at emergent.
2:08:03
We are building world's most advanced wipe coding platform uh for consumers uh for nontechnical users.
2:08:09
They can come in, prompt an idea and get a fully built live app uh that they can take to production.
2:08:15
Um and we launched a few months back.
2:08:17
Over a million users have tried the platform.
2:08:19
More than a million and a half apps have been built so far.
2:08:22
We're growing pretty crazily right now.
2:08:23
>> How do you think about sub like like beach head markets and fra uh like like submarkets?
2:08:28
Uh there are platforms that focus on get an iOS app in test flight.
2:08:35
Uh build a game, vibe code, a a website like are are the people that are signing up saying I'm going to start a business.
2:08:42
I'm going to build an app that will be the foundation of like you know some monetary thing or is it you know are we in the era of like vibe coding apps as memes somewhere in between? Are you still exploring? What do you think? Yeah.
2:08:53
So I mean we we are the only platform that supports uh web app, mobile app and back end all integrated in one and we have consumers coming from all all parts of life.
2:09:01
We have business owners trying to digitize their business.
2:09:04
We have entrepreneurs building their startup on on emergent right now.
2:09:09
Um a lot of people are building apps that are moneti monetizable right now and they are shipping them.
2:09:12
Uh and so so it's a it's a crazy spectrum of ideas on the platform right now. >> It's amazing.
2:09:18
A million users is a lot.
2:09:20
like what's the what's the acquisition funnel?
2:09:23
Like how do you actually get people?
2:09:24
It's it's it's a it's a it's a buzzy space.
2:09:26
It's a cool tech, but I imagine it's hard to get people to go to your specific website. Like what's working? >> Yeah.
2:09:32
So, I think people really really love our product.
2:09:34
I think that's what's sort of working really well for us.
2:09:36
A lot of the acquisition is word of mark today for us.
2:09:38
Uh that people are refering each other on the platform.
2:09:40
As they get successful app, uh they refer each other.
2:09:44
We also do bunch of influencer marketing.
2:09:46
So we use Tik Tok, Instagram, X uh to promote our our brand and we partner with influencers and they're able to sort of create stories which resonates with their audience.
2:09:54
Uh that's something that is working really well for us as well.
2:09:59
>> That seems to be working really well these days.
2:10:00
I feel like there's this interesting flywheel where maybe for the first time we're seeing influencers drive drive software adoption, software downloads.
2:10:08
I mean we've all seen like VPN ads, but uh we're in a new age where people are are pushing like 5.
2:10:13
people are are pushing like 5. is what how what what's good retention in your view because obviously a bunch of people are going to come in create stuff and then uh but uh you know 12 months from now how many of these people what what is success look like in terms of routine
2:10:28
so now I'm a real business you can have a relative I think I think >> the beauty of these things is somebody can come in in 20 30 minutes an hour they can create something very cool >> you don't need to retain all of them you don't even need to retain the majority of them but how how are you thinking and kind of planning planning around that. >> Yeah. So we we have uh good retention on >> Yeah.
2:10:45
So we we have uh good retention on the platform with month three like roughly 40 50% of the users are still on the platform.
2:10:51
We have a very strong power user behavior where top 10% of the users are spending a lot more money on the platform who are building serious apps.
2:10:58
They're taking them to production and there our retention is north of like 80% right now.
2:11:01
So uh so we think it's still early days but I think what's going to happen is like a lot of people are going to come on the platform lot of because we also take care of deployment.
2:11:09
We host these apps on our platform.
2:11:10
they're going to stay with us on on the platform and uh what's also happening is as they build more they they want to add more feature as they get more feedback from their users they come back to build more on the platform.
2:11:19
Uh still sort of early sort of days but but we think that you know like uh you can build a a pretty large business around these power users uh who are trying to launch their app launch a new business uh or a new idea uh to the world.
2:11:32
>> What do you think for business model uh consumptionbased?
2:11:33
Uh, do you want to get people to spin up an app, subscribe, and then you're acting as their hosting provider long term?
2:11:39
Um, a lot of people build websites, and then they're happy with them and they don't actually have a lot of feature requests.
2:11:44
Uh, we've seen Squarespace and a bunch of other companies, you know, they're they're not they're templated based, but they they wind up being great businesses because people are happy to continue managing on WordPress or Squarespace.
2:11:56
Uh, what what what business model makes sense in the in the vibe coding AI era? >> Yeah.
2:12:03
So for us like almost uh most of our apps are actually uh like full stack web apps which has backend databases on so mobile apps.
2:12:11
Uh so what we are seeing uh you know is is a lot of users are coming and adding adding features to them and building them.
2:12:16
Uh I think in a steady state like a lot of these users you you'll have a power law where like you know top top x% of users are going to deploy big apps have breakout success on the platform and those are going to be continuing to be on the platform and and the business is going to get built around those people.
2:12:32
Do you expect enterprise vibe coding or consumer proumer vibe coding to be more competitive in the long run and do you think that companies should try to do both or pick a lane?
2:12:41
both or pick a lane? I mean we are very focused on consumer uh we think these are two separate markets uh require separate separate sort of uh personal for example like a lot of our users they call GitHub Jitub right so we have to like really you know like think think through like who we are catering to we have a lot of education baked into the
2:12:59
product which tells them what is API like you know how to sort of you know look through errors and things like that um so I think these are two different sort of swim lanes and of course there's an intersection in the middle where like small teams use us uh as well but but largely like like we are very focused on consumer market right now. >> Very cool. >> Very cool.
2:13:15
>> That's refreshing to hear because we've had a number of of other platforms on the on the platform and they always say, "Oh yeah, we're we're doing consumer but we're really like, you know, a lot of, you know, basically not not really admitting that they're two separate markets in my field." >> Yeah.
2:13:31
There's a lot of growth masking like a lack of actually understanding what the product will be when it's mature. >> Yeah.
2:13:37
What the long-term market looks like. >> Exactly. >> Yeah.
2:13:40
Yeah, I mean when we started we we we thought like I mean realization that we had that like a billion people have ideas in their head like and how do we sort of bring that out?
2:13:46
How do we sort of help them launch >> and we think it's going to sort a new uh sort of economy and we just want to you know be part of that. >> Well congratulations.
2:13:54
Thanks for coming on the show. We'll talk to you soon.
2:13:57
>> Keep us posted on your progress.
2:13:59
>> Have a good rest of your day. >> Cheers.
2:14:01
>> Victoria has a post here. Uh it's happening.
2:14:04
For the first time in history, AI designed a complete genome and it has been proven functional in the lab.
2:14:12
>> Many of the most from Samuel King.
2:14:12
Yes, this is going to get a lot of people going.
2:14:16
Many of the most complex and useful functions in biology emerge at the scale of whole genomes.
2:14:20
Today, Samuel King is presenting the preprint generative design of novel bacteria phages with genome language models where they valid validate the first functional AI generated genome. Uh, fascinating.
2:14:33
Other news, uh Dylan Field shared a leaderboard >> of uh ranking uh a bunch of different vibe coding tools.
2:14:42
Figma make came in second place, number one in our hearts. >> Yes.
2:14:48
>> Uh but also featuring uh Lovable at three, Bolt at five, Bass 44 at 8, and V0, and uh Replet of course.
2:14:55
So >> this is so interesting.
2:14:59
I didn't know that you could that you could do uh like rankings here, but I guess they they did uh blinded pairwise matches.
2:15:05
So I guess they're human judged as an as an eval.
2:15:11
They they sent them out prompt or something like that. >> Yeah.
2:15:14
And it's like how much do you uh Yeah.
2:15:15
So so uh various prompts and then and then show two two results to >> this is just like how LM Marina works.
2:15:24
>> It's LM Marina for for vibecoded websites. >> Yeah. >> Very cool.
2:15:28
Well, congrats to Dylan Field over at Figma.
2:15:31
Um, Run has a great post here, uh, talking about the the anthropic mayulpa.
2:15:36
Schultto posted, "We're sorry and we'll do better.
2:15:39
We're working hard on making sure we never miss these kinds of regressions and building trust with you."
2:15:44
And Run says, "Shto has a Japanese sense of honor to his customers." I love it. >> Love it.
2:15:53
>> Such a >> Yeah, the comment the comments.
2:15:54
Uh >> yeah, you were saying that people were really really upset about anthropic and it's interesting because >> Yeah, it was hard to sus out.
2:16:00
Was it like extreme power users or were average user actually mad?
2:16:05
>> What was your What was your take, Tyler?
2:16:07
>> Uh I think it was definitely like power users um like API um >> so if you like kind of built a business on top of anthropic and then there's regression down >> unreliable. Got it.
2:16:18
>> Sometimes it's like really slow.
2:16:19
Sometimes responses are just like >> Yeah.
2:16:21
And they and and and they they did not say that this was like they ran out of GPU capacity, something like that, right?
2:16:29
>> Um >> this is this was more in like they they shipped a new uh a new version and it resulted in downtime, right?
2:16:36
So it was it was very much like not on the not on the capex side of the business.
2:16:41
Uh it was more on the actual like development that they did. >> Yeah, I think so.
2:16:46
Well, I mean they made a they made it very clear like we do not we will not degrade performance of the model uh just to serve more people.
2:16:52
So I think there might have been some aspect of there being like not like just not enough GPUs.
2:16:59
>> Yeah, I mean it's certainly been like demand's been skyrocketing.
2:17:00
Uh brown paper bag in the chat says clawed code subs are mad. >> Big mad. >> Uh big big mad.
2:17:06
Uh well, you know what's not down? Wander. Find your happy place.
2:17:13
Book a wander with inspiring views, hotel, great amenities, dream beds, top tier cleaning, and 24/7 concept service.
2:17:17
It's a vacation home, but better.
2:17:19
>> Um, uh, also in other news, Limewire has acquired the Fire Festival brand for $245,000, and liquidity said, "My phone got a virus from reading this tweet."
2:17:30
That is I I have no idea what's going on here.
2:17:36
We need to do a deep dive on Lime Wire.
2:17:38
We need to get the red string out and figure out if we're uh like what is Limewire still doing? What is their business?
2:17:44
Why are they acquiring the Fire Festival brand?
2:17:46
What are they going to do with that?
2:17:48
I feel like there's something to be done with the Fire Festival brand.
2:17:51
Like it Fire Festival is a big enough brand, there's been a documentary at this point, people know it, that like if you were able to buy the brand for 245K, you could potentially do a marketing stunt with it that would generate half a million dollars in value, maybe.
2:18:04
But it does seem like it's got to be really really inspired because >> Fire Festival, the tech conference, >> the brand is pretty bad.
2:18:12
Like the brand is associated with >> Yeah.
2:18:15
It's one of those things like if you're going to put all the effort into getting to creating a a realworld event that's not embarrassing. Yeah.
2:18:22
Why not just kind of create a new brand that >> totally >> that's not uh >> I don't know maybe toxic.
2:18:30
>> I mean you could >> it's attention getting. >> Yeah.
2:18:32
Um, but yeah, the I mean this had this had been on the market for a while.
2:18:37
It was like >> Wait, the brand >> Fire Festival.
2:18:39
>> Oh, I had no idea it was on the It was on the market. >> Billy McFarland.
2:18:41
>> Did you Did you go to Fire Festival, Tyler? >> I did not.
2:18:43
But I I think the main like whenever I see stuff like this, >> he was too poor and then he was too rich.
2:18:51
>> Um, but I I think uh you could see some kind of like Enron thing where it's like, yeah, it's a marketing thing and then it's crypto. >> Yep. Yeah.
2:18:58
>> Oh, be careful out there.
2:18:58
or watch out for fire festival coin. Uh, how is Enron doing? >> The memecoin. >> Yeah, the meme. It was memecoin.
2:19:05
It was a publicity stunt.
2:19:07
Uh, Taylor Loren did some crossover video production with them.
2:19:11
Um, the team behind it seemed like genuinely great uh like comedian videographers, but then it it it took a really odd turn um as as things always do when they turn into into meme coins.
2:19:26
And uh it it was a funny bit for a little bit.
2:19:28
It was it was a funny idea, you know, uh, leaning into to Enron.
2:19:33
Certainly fertile ground for jokes.
2:19:35
>> Everyone's familiar with Enron.
2:19:35
I saw people having like buying old Enron merch like years before this like brand revitalization.
2:19:41
You could see them doing something funny with Fire Festival.
2:19:43
Or you or you could honestly see the value of that brand just being like producing more content about the failure of Fire Festival, like instantiating that.
2:19:50
like there's been a documentary, but maybe you buy the brand, you get the rights to do more with it, and so you can you can monetize that way and not and and actually lean into the fact that it's it has a it has a lot of baggage.
2:20:02
It's like a negative it's a negative association. >> Yeah.
2:20:06
You you would hope they could have done something more with this this iconic brand. >> Yeah.
2:20:12
Did you see this comparison of the thickness of recent iPhone cameras next to each other?
2:20:16
We were really in the bulky iPhone era.
2:20:18
The iPhone 5 completely flat. No bump.
2:20:21
The iPhone 6 gets >> bulking season.
2:20:25
>> The iPhone 6 they called they they they said it was a camera bump or something and we didn't know what we were in store for.
2:20:32
This is a Well, now it's called a plateau.
2:20:34
That's the official term. >> The plateau.
2:20:38
>> I don't think Applecoms wants you to call it a bump.
2:20:40
>> I just ordered I just You ordered it? >> I just ordered Pro. I I just got the silver. >> You got the silver?
2:20:46
>> I I think I think uh that the orange >> if it was in ramp yellow, it would have been done deal. Orange is a lot.
2:20:53
>> Yeah, I kind of liked the orange less and less as time went on. Yeah.
2:20:57
>> Uh unfortunately, I ordered it late.
2:20:57
Now I have to wait till like mid October.
2:21:01
>> Oh, they're they're selling out.
2:21:01
That that's bullish for Apple. >> I think so.
2:21:06
>> Like we like it for the stock. >> Yeah. I mean that seems good. What about you, Tyler?
2:21:09
You itching for one or are you satisfied?
2:21:11
>> He just got a new phone.
2:21:11
Yeah, I just got a new phone. >> So, you're happy? >> Yeah.
2:21:14
I mean, I was at I was I had an 11 before, so this is amazing. >> Oh, yeah.
2:21:17
You jumped forward to the future like seven years. Yeah, >> that was great.
2:21:20
Uh 17 Pro looks pretty good.
2:21:23
Uh did you see the the latest on the OpenAI brand exploration? >> The latest.
2:21:27
This is from February of 2023.
2:21:30
>> That's a decade ago in AI terms area.
2:21:33
I've never heard of this firm before, but really really cool work.
2:21:37
>> John Palmer's >> Yeah. >> uh firm. >> Amazing.
2:21:39
He had a has a crypto company. >> Okay.
2:21:43
>> Backed by uh Aras says brand exploration for OpenAI with Sam Alman February 2023.
2:21:50
two logo concepts, a circle and a monogram alongside broader exploration for Chachi PT across brand and product.
2:21:57
And this just like threw me back to the 90s.
2:22:00
Like it's just remarkable.
2:22:00
It feels like old Microsoft, old Apple, old Mac.
2:22:06
Um I kind of get why they didn't go this direction, but um like sort of beautifully nostalgic.
2:22:10
And uh Sam Alman commented on it.
2:22:13
He said, "Fun to look back at this exploration with area on the OpenAI brand.
2:22:16
This work partially inspired the circle that we use and love in our products.
2:22:21
Um the the SIF open AI font just feels like a boxed piece of software.
2:22:29
This feels like throw this logo on it, sell it at Best Buy. One chat GPT, please.
2:22:34
I'll take version five and I'm good for the next two years.
2:22:39
They got a CD key in there. Lock it in.
2:22:43
But the the O AI logo on the on the right there in the top right is is wild.
2:22:49
This is like it's a very loud design.
2:22:52
>> I think they should re-release these as as like uh merch for the >> It' be great. Yeah. Yeah.
2:22:55
It's a it's a little like of of like Yeah.
2:22:58
It's just so retro in my opinion.
2:23:00
Uh >> you see this post from Cass over at Open Doors and he's this morning we filed an 8K to say that my and Open Doors X accounts would be used to talk with our investors.
2:23:11
We've also parted ways with our former external PR agencies.
2:23:14
When we want when we want to talk, you will hear from us, not paid professionals who don't share our mission.
2:23:21
>> I didn't realize that you had to actually disclose that in the AK that you're planning to post.
2:23:26
>> Should be God-given, right?
2:23:26
Um but yeah, Cass is on a tear.
2:23:30
I'm excited to hear more from him and his plan.
2:23:33
>> The top comment, you're a god. He says, I'm not. There's only one God.
2:23:36
I'm just praying very hard that he gives me the power and the perseverance so I can fulfill his will. >> Let's go, Cass. Good job. >> Rallying retail.
2:23:44
Uh, did you have an did you did you see this? Um, >> what else?
2:23:49
>> This this movie coming down the line.
2:23:51
>> Oh, I I I haven't seen the movie. I saw it.
2:23:55
>> Jackson Doll said, "Ladies and gentlemen, we see we have what seems to be an absolute banger on our hands.
2:23:59
The movie is called One Battle After Another.
2:24:03
It stars Leonardo DiCaprio.
2:24:03
Oh, it's in theater team 26.
2:24:06
That's what I was thinking. It's in It's in IMAX.
2:24:09
Uh, so get the tuxes ready, boys.
2:24:13
>> Task for the team, guys.
2:24:13
Figure out when we can all go see this movie together in IMAX.
2:24:18
Get the tickets today and >> get us without further ado.
2:24:21
We have our next >> We have our next guest in the reream waiting.
2:24:25
>> Got carried away to the >> DVPN.
2:24:29
Sorry to keep you waiting, Noah.
2:24:31
>> Welcome to the stream, Noah. How you doing? >> Lost.
2:24:34
>> We got lost in the timeline. I'm good.
2:24:36
>> Uh, are you going to the Google event today? Break it down for us. >> Oh, not right now.
2:24:40
I got busy with work, but Okay.
2:24:42
I'll be at the dinner tonight. >> Oh, fantastic. Fantastic. >> Too locked in. >> Locked in.
2:24:47
Well, what what are you locked in on? What are you building?
2:24:51
>> Um, so soft and we're building computer use agents. Okay.
2:24:53
>> Uh, >> so we're training foundation models on how to use a computer like a Cuban. >> Yeah.
2:24:59
>> To really automate any type of work.
2:25:03
What's uh what's the secret sauce?
2:25:03
How do you differentiate?
2:25:05
It's a it's a it's a complicated industry.
2:25:08
>> You train the model and then you say, "Hit the keys in the right order to make me $und00 million. Don't make mistakes." >> Don't make mistakes. >> Yeah. Exactly.
2:25:15
We don't allow it to make mistakes. >> That's the secret.
2:25:19
>> There's there's big consequences.
2:25:21
>> Uh but I mean right now is is the is the barrier to computer use just scale? Is it algorithms? Is it data?
2:25:26
Like what are the key inputs to getting good results?
2:25:32
Um, it's really a mix from what we've seen a lot of these things kind of combined, but a big part of it is actually the data.
2:25:37
Um, there's not a lot of like really good data to use.
2:25:44
>> Where do you get more data?
2:25:44
Who are you going to call?
2:25:47
>> Well, I think a lot of it can be synthetically generated.
2:25:49
Like we we worked on these pipelines back in April for our first model release where uh like pretty much our entire data set was just synthetically generated. >> Yeah.
2:25:59
And then I think like obviously you have things like you can label your own data.
2:26:03
Um you can collect it or you can get it from products.
2:26:06
So you need to use a combination of everything.
2:26:09
>> Computer use is so broad.
2:26:09
I mean it can mean everything from like you know ordering Door Dash to playing a video game.
2:26:15
Uh what's the lowhanging fruit?
2:26:18
Like what what is the next step that you know feels solvable in like the 12 to 18month time horizon?
2:26:26
Yeah, I think the first few tasks that you really need to solve are these monotonous type of tasks which are like pretty easy for a human to do like cleaning up your email inbox to like scheduling your calendar. >> Sure.
2:26:39
>> Um to finding all of your receipts in your email like things that are very low level which like cognitively it's not like super advanced.
2:26:47
>> Why do you need a why do you need a computer use agent for that?
2:26:48
I feel like uh email calendar these products have existed for decades.
2:26:53
They have APIs like they can be interacted with via JSON.
2:26:59
>> Yeah, there's actually a lot of like software out there which like either has really bad APIs or things or uh different types of products where you need to use computer use to really interact with it efficiently.
2:27:09
>> Uh usually because they might just have broken APIs straight up.
2:27:11
So do you expect in the future people will be like uh oathing into their email on a virtual machine and then letting you hang on to the the the cookie or the login in in some VM for like a long period of time because it feels like uh Apple's not going to let an AI agent run on my app and that's how I use email mostly.
2:27:37
You know, I think like um I think what we're going to see is like uh we're going to see a mix between like the current LLM paradigm where it's like you use things like structured outputs, you use APIs together with computer use.
2:27:48
Um and like in the future I think like you're going to like models are going to jump between both uh and they're going to do so like really efficiently.
2:27:58
So it will just kind of depend on what the input is.
2:28:02
How do you think about uh your place in the stack?
2:28:05
Do you want to partner with Google?
2:28:07
I know you're you're talking to them.
2:28:09
Is is uh is the goal to build applications on top or or stay as this like point solution work with foundation model companies?
2:28:18
Like how do you see your customers developing over time?
2:28:24
>> Um the way I view it is that like computer use is still very new. >> Yeah.
2:28:28
>> Yeah. So um there hasn't been like any product that has really taken off in the space like there was SHAT GPT operator there was cloud computer use but like none of these have really taken off uh like immensely like just normal SHA GPT and I think like initially it's going to just be that you provide an API to a
2:28:46
very good computer use model which you sell to enterprises you sell to startups that are working on in different industries really uh across with different customers and different vendors um and then eventually Like what we're looking at is like if this can somehow transition into a more consumerf facing product as well. >> Well, it sounds like the dinner tonight
2:29:04
>> Well, it sounds like the dinner tonight will be a steak dinner, a sales dinner.
2:29:08
Hopefully you can get some customers.
2:29:10
Thank you for stopping by. >> Yeah, thanks.
2:29:12
>> We'll let you get back to your busy day. We'll talk to you soon. >> Cheers. >> Have a good one.
2:29:17
>> Did you see the Arizona iced tea account posted?
2:29:19
Legendary link up Arizona and the Costco CEO and then it just inflation. What?
2:29:27
Because Arizona's always 99 cents and the Costco hot dog's always 99 cents. Look at these two.
2:29:32
Look at the founder of Arizona is still cooking too and the Costco CEO. They're just boys.
2:29:39
>> Inflation literally doesn't exist. >> It doesn't. It's fake. >> It's made up. >> It's in the computer.
2:29:45
>> You can just not print a sign that says the hot dog costs more >> and it will stay the same the same cost.
2:29:51
>> Just hardcode the pricing into the into the register. >> Yeah. Look at this. Costco Don. Don and who? Who? And and Rob. Don and Rob. >> Dream.
2:30:03
If you know them, introduce us.
2:30:06
We want to have them on the show.
2:30:07
>> Had you ever seen this picture before? Tracy Aloway. Uh, >> this is iconic.
2:30:10
I had no idea this existed.
2:30:12
>> Financial crisis crisis most iconic picture.
2:30:16
>> This is a fantastically iconic picture.
2:30:18
>> Is she like like uh like >> hamming it up a little bit? I think a little bit.
2:30:23
I mean, it's not like she lost, >> but the chart everything's looking red.
2:30:27
>> If anything, you know, crises are are are pretty good when you're in the news business, but we love to see uh Tracy in an iconic photo.
2:30:35
What a fantastic what a fantastic insight with the Dell computer rocking the Dell stapler, got the papers, and it's unclear what she's looking at.
2:30:45
Some sort of Bloomberg terminal, but it's just all red. But I have no idea.
2:30:49
It doesn't look like a stock chart that's red.
2:30:50
just looks like a lot of red cells in a spreadsheet or something. I have no idea.
2:30:55
We'll have to ask her what she was actually looking at and and how real this was.
2:30:58
I mean, it was a stressful time for everyone, even if you're in even if you're in media and you have more to report on as the financial crisis is unfolding.
2:31:05
Like, I remember reading the Wall Street Journal every day during the global financial crisis, trying to understand what was going on.
2:31:11
And like the the there was no there was no there was no lack of news.
2:31:13
There was tons of interesting things happening as people understood what was going on at Bayer Sterns, what was going on at Lehman Brothers.
2:31:20
Like these are storied institutions that are collapsing.
2:31:22
You want to hear about what was what were the decisions made by the board members, who was who was involved, like what was the plan for the bailout, what's the government's response.
2:31:29
Um, but it's still stressful no matter what job you're in.
2:31:33
>> But what an iconic picture. >> What do we have next?
2:31:36
Ryan Peterson is in >> Arena Mag.
2:31:39
You got to go subscribe to Arena Mag. Ryan, >> I saw this.
2:31:43
I thought I was looking at the cover of GQ. Fantastic. Looking at the cover.
2:31:48
>> Max Meyer stepped in to write Ryan's Worlds.
2:31:50
Flexport CEO Ryan Peterson on the craziest year for global trade.
2:31:53
Uh with trade chaos and the rush for companies to implement AI.
2:31:58
There are few executives in higher demand than Ryan Peterson.
2:32:01
Flexport builds technology for logistics. Freight forward.
2:32:03
He's been on the show a bunch.
2:32:05
The Flexport office is full of curios from the worlds of from the world of logistics. I love a curio.
2:32:12
We have a lot of curios in our museum of business. >> We do.
2:32:16
>> Uh the hallways are decorated with posters that Ryan Peterson himself generated by Chad GPT.
2:32:21
>> Uh on the engineering floor, cardboard model of the Evergiven that that you remember the Evergiven story that >> back back to your point on uh on on uh relics.
2:32:30
Uh uh Dylan was saying that the a friend of his the only Zuck signature signed object that he's aware of that traded hands traded at at 100K. >> Uh so Oh yeah.
2:32:45
>> So so some there was a Zuck signature that hands >> went for about 100K.
2:32:48
Uh which would put the the game hit Gong probably in the >> eight figures. Eight figures.
2:32:53
I'm thinking for sure >> and never sell it.
2:32:56
It's cuz it's going it's going in the room.
2:32:58
>> We did have a plan uh to uh to engrave on the gong a contract and not show him that.
2:33:06
So when he was signing the gongle he would accidentally be signing a contract to onboard ramp.
2:33:11
That would be the real goal.
2:33:14
I don't think that would go over well.
2:33:16
But >> uh Nathan keep it keep it in the comedy sketches.
2:33:20
I suppose >> we'll have have much problem there. >> Yeah.
2:33:23
Um, Built Rewards built uh their own TV show, short form series called Roomies.
2:33:29
They release on Tik Tok and Instagram.
2:33:31
I'm gonna guess that uh Adam Faze did this. Oh, yeah.
2:33:34
Because he does he does these types of things for different companies.
2:33:40
>> I'm going to message him.
2:33:41
>> I got Yeah, I got to meet him.
2:33:41
I His name keeps coming up and and I and I want to know more.
2:33:45
>> Yeah, we should have we should actually >> Yeah, I'd love to have him on the show and meet him.
2:33:47
Uh but yeah, built Built Rewards, founded by Anker Jane.
2:33:49
um cash back or uh what uh credit card points on paying your rent is the pitch.
2:33:58
>> Um but uh pretty pretty remarkable to actually get uh get a uh like a scripted series actually running.
2:34:04
Uh it's pretty hard to do.
2:34:06
Uh most people that can that can do this level of we'll see we'll see how quickly he gets back.
2:34:12
>> Get him on the show right now. We'll see.
2:34:14
>> Uh Derek Thompson says AI is overrated.
2:34:19
Uh he says, "My baseline case is that the AI being built right now is overrated.
2:34:22
Soon it will be a disappointment.
2:34:24
Then it will be a bubble and by the 2030s it will be world changing.
2:34:28
Self-driving cars are a model for this.
2:34:30
In 2015, I heard autonomy was 5 years away from taking over the roads.
2:34:34
In 2020, they were nowhere.
2:34:34
Even in 2022, you could say they were a huge disappointment.
2:34:38
Now they're quietly a revolution.
2:34:40
Driverless taxi usage in California grew 8x in one year and Whimo is expanding to other cities." >> Truth Zone.
2:34:46
Tyler Cosgr, what do you think?
2:34:49
He's not just wrong, he's unpatriotic. >> Whoa. >> Oh, shots fired. >> Fired.
2:34:55
Where's Where was shots fired?
2:34:55
Do >> we know how that >> amateur over here? >> First time. First time. >> First time.
2:35:07
>> First time on the decks. Yeah.
2:35:07
So, you're now you're just hitting random stuff.
2:35:09
That's That's uh that's that's offensive. >> Hey. Hey. Easy. Don't go.
2:35:17
Uh well, I mean it's not that crazy.
2:35:17
It does feel like we're in some sort of like the last the last 1% takes 99% of the time.
2:35:26
You know, we got 80% of the value and but that's not enough.
2:35:28
And so the the last 20% going can take 10 times as long.
2:35:34
>> I don't I mean we we don't have a Dyson sphere.
2:35:37
How are we getting 80% of the value?
2:35:39
We we're only getting like a 10,000.
2:35:41
>> So So what's your timeline for a Dyson sphere?
2:35:43
Do you think we're going to have a Dyson sphere in in two years?
2:35:47
>> Two is a little close. >> He's saying 2030.
2:35:48
That's only 5 years away.
2:35:50
I guess he said the 2030s. >> Yeah.
2:35:53
So that's a whole >> What's your What What's your Dyson sphere timeline? >> In my lifetime. >> In your lifetime. >> I Yeah. >> Okay. >> Well, yeah.
2:36:02
>> Oh, so so if you lose the bat, you won't be around to make do. You won't be >> I Yeah. In my lifetime. >> In your lifetime. >> I think so. >> Yeah.
2:36:11
I could see I could see it by 2100 potentially, but it's a lot of work.
2:36:14
A lot of lot needs to happen to get to get get the Dyson sphere up there.
2:36:18
It's going to be a lot of rockets going to the sun. >> Well, we have Dan.
2:36:25
>> Let's bring in our next >> show from regular one. >> Welcome to stream. How you doing?
2:36:35
>> Hey, it's such a pleasure to be here.
2:36:37
I've asked Seoa what to prepare for and they've told me that this should be the most fun, exciting, fast-paced interview.
2:36:43
So, I'm, you know, it's like, shoot at me. >> Let's keep it quick. Introduction. What do you do? >> We are irregular.
2:36:49
We just came out of stealth yesterday.
2:36:51
Uh, we are the first frontier AI security company out there.
2:36:56
Our goal is to be the counterpart to the open AIS and anthropic and GDMs of the world.
2:37:00
We're already working with all of them in to create the security stack of the future. >> Yeah.
2:37:06
what what what does that mean?
2:37:06
I mean, there's already a ton of security companies out there.
2:37:09
I see them when I run walk through the airport.
2:37:11
Um, they're all thinking about AI.
2:37:13
How are you positioned differently?
2:37:16
>> Yeah, it's a great question.
2:37:16
So, the the thing that we're actually building is a highfidelity simulator that allows you to put in any model in it.
2:37:23
And essentially just like see how different scenarios in order to attack the model or whether the model can attack other targets.
2:37:32
So for example, we were the first in the world to see jailbreaking that AI was doing to another AI or alternatively just like seeing how AI can bypass things as Windows Defender in order to just like hop from one endpoint to the other. >> Interesting.
2:37:48
>> What is unique and different about us is that we're working very closely with the labs out of the assumption that what AI is doing right now is simply not a story and and security is about to have a huge paradigm shift moment which if you think about it makes sense, right?
2:38:00
because enterprises are probably going to look very different in the next 5 to 10 years.
2:38:04
So naturally the security stack is also going to look very differently as well.
2:38:08
So we are using this high fidelity simulator to find the the novel attacks and to build the next generation of defenses a few years in advance.
2:38:18
>> Why do labs don't want to do this internally?
2:38:20
It feels like something that they have responsibility for.
2:38:22
They have tons of I mean they get questions on Capitol Hill about this.
2:38:25
Like if they outsource it that seems like a very very tricky thing. >> Yeah.
2:38:30
But at the I mean it's not it's not entirely outsourcing it like they have to care about these things.
2:38:34
The same thing that you know I think every every company does.
2:38:37
It's like you want to have your own security practices and protocols but simultaneously have partners that can help you see things in a different way.
2:38:46
But >> what do you think D? >> Exactly.
2:38:47
It's it's a great question if you think about it.
2:38:49
You know there's a thriving security industry already that is very mature and you know most companies are using you know the great of the security world right now the palos you know just like the cloud strikes etc.
2:38:59
even though they're external to the companies.
2:39:00
And the reason is that we're about to encounter what is potentially the greatest security challenge ever just because we're innovating at such a fast pace and there's so much work to do.
2:39:10
there's so much work to do. So we kind of think about it in terms of like differentiation from the inside to the outside is that there are some defenses that you would want to put on the models themselves baked into the no nets and that's clearly lab territory but some defenses are going to happen on the AI
2:39:26
agent side and some defenses will have to be in the environment just because if you believe that we won't be able to do secure by design to AI that means that some of the defenses will have to be implemented on the enterprise side in the environment otherwise ize you won't have any defenses beyond what the front tier companies are going to bake into the models. And because there's so many
2:39:46
And because there's so many different verticals and scenarios and contexts that you need to put in, there is a lot of effort that needs to be created in order to create defenses across the entirety of the stack.
2:39:56
And we're working side by side in order to make sure that whatever the labs are not doing, we are going to do in order to create the next powerhouse of security.
2:40:06
>> What is this uh what does a regular look like over time?
2:40:08
Is it is one way to think about it is like a a a network to like detect like uh like rogue agents, right?
2:40:17
Is that is that a potential kind of scenario that that you guys would be uh helpful in in preventing or or what what is like the surface area of the product look like uh over time?
2:40:30
>> Yeah, thanks for the question.
2:40:30
So I'll say that indeed one of the scenarios that we are covering already today uh we are doing things as understanding and monitoring AI network systems in order to see if they go out of bounds but our view is that something deeper is going on here and that the entire infrastructure will need to be replaced.
2:40:50
I'll give a concrete example around that.
2:40:51
So you know just like anomaly detection that's a huge part of the security stack right now right but how does it work?
2:40:59
you have a baseline and you're seeing whether a model is or just like whatever you're trying to monitor is doing something which is outside of what you would expect as the normal behavior.
2:41:09
But if you don't know how an attack is going to look like, you don't have a proper baseline.
2:41:14
And as an outcome of that, our view is that the first order thing to do is to create a strong research infrastructure that would allow you to essentially figure out and map the novel attacks that are unique to AI.
2:41:25
see what are the gaps in the current security stack and start to fill them in from where and just like build a new platform that's going to be the platform to secure the agents of the future.
2:41:35
So our hope and our ambitions are high.
2:41:37
are high. We believe that there is a place to create a huge company that is new around security very much like you know we started as part of the transition to the cloud and you know checkpoint started early on just like a
2:41:49
few decades back when people started to implement and it's like networks in enterprises and usually when infrastructure is changing you have a just like a window of opportunity to create the company that's is going to be able to create a platform to secure that
2:42:03
infrastructure end to end and that's our goal and that's what we want to do around AI Do you think it's more are are you more worried about the like rogue AI, the the AI that just randomly decides independently to try and break out of its environment or more bad actors
2:42:21
thinking that if they can get an LLM to do something inside of an OpenAI environment or inside of a Google DeepMind environment that they can extract some sort of value like who is inciting the attack? >> So ultimately it's both. I think in the
2:42:34
>> So ultimately it's both.
2:42:34
I think in the near term it's the latter.
2:42:36
So it's much more likely that we'll need some human interaction in order to elicit AI capabilities to push them into more dangerous scenarios.
2:42:44
dangerous scenarios. And I'm much more concerned right now about you know it's like tell organizations getting access to advanced AI system that are already you know if you look at like the system cards of open AI and on topic they put
2:42:57
the capabilities of around bio and chemical so models being able to just like actually help in order to produce them at higher risk levels over time which makes me personally just like you know concerned about what happens if some bad actors are going to have access. That's also true on the
2:43:12
That's also true on the commercial side that the more that we delegate to AI, if malicious actors are going to gain access, there is potentially going to be a whole new wave of viruses.
2:43:21
And I think the near future is AI augmenting attackers and being used as part of, you know, just like the attack surface. >> Yeah.
2:43:31
>> Over a longer horizon of time, we'll need to also make sure how we take care of just like rogue actions that are done by the AI itself. >> Sure. Well, good luck to you.
2:43:37
Thank you so much for hopping on the show.
2:43:39
We will come come back on.
2:43:41
I I expect in the next at least in the you know the next two years probably the next year there's going to be some type of event and we're going to think we got to call Dan to break this down.
2:43:51
But hopefully you prevent it before it ever happens. Yeah.
2:43:54
So >> it will be it will be my pleasure both to prevent it but also to come back on the show. >> Thank you so much. >> Let's do it. >> Cheers. >> Byebye.
2:44:03
>> Jensen Wong is a huge Nano Banana fan. Love to see that.
2:44:07
And Sar Pachai quote posts him and says, "Mine, too. I uh it made his day." They're both very happy.
2:44:15
Also, Joe Gibbia is uh is shouting out Breathe Realm. >> I threw this in here.
2:44:21
Uh new air freshener company.
2:44:23
The team when we moved into the studio bought a bunch of air fresheners.
2:44:27
They said, "Throw those away."
2:44:30
They release a bunch of, you know, toxic chemicals in air that you then breathe that then go throughout your body.
2:44:36
Uh Sarah uh my wife actually invested in this company >> years ago.
2:44:42
So it took a while to get out to launch.
2:44:45
Uh but cool to see >> the new standard of air care. Go check it out.
2:44:49
Saffron Citrus and Verbina Santal. >> Sounds delightful.
2:44:54
Uh well next up we have Ben from Braille coming in the studio working infrastructure >> and into the TV here at Ultradown.
2:45:08
Welcome to the stream, Ben. How you doing? >> What's happening? >> I'm doing good. Good to see you guys. Thanks for having me.
2:45:14
>> It's great to have you.
2:45:14
Um, >> kick us off with an introduction.
2:45:18
>> Good to see you again.
2:45:18
Actually, I think the last time we spoke was probably 2022. Feels like a decade ago.
2:45:24
>> Yeah, I think I might have been trying to like shill you on doing something with stable coins and we were trying to get started to be honest, you know, like >> awesome takes a minute as you know. >> Yeah. Yeah.
2:45:33
uh catch us up to I mean since it's your first time on the show, quick history on yourself and then uh we'll get into Braille and everything you've been working on. >> Yeah. Cool.
2:45:44
Well, hey guys, I've been building fintech companies for I think longer than fintech's been a word.
2:45:48
Uh I worked on one for about 10 12 years.
2:45:50
Uh took a couple of years off, kind of went for a long walk in the woods and then kind of started trying to think about what's next.
2:45:59
And when we were workshopping ideas, it became pretty obvious that, you know, like blockchains were going to surpass traditional databases um in terms of like speed and cost, which is great for financial transactions.
2:46:10
And stable coins were the obvious way to fill the space, but they were just super expensive to actually create.
2:46:16
And so when we started, it was kind of like the litmus test or the baseline was like 100 million bucks in two years to create one of these things.
2:46:24
It took us two and a half years to do it, but we got it down to a dollar in a minute where any business or anyone that wants to go launch a stable coin can go launch it.
2:46:30
And basically the reason people love these things is because it reduces cost and uh makes revenue. Okay, first question.
2:46:37
Why why do we need infinity stable coins?
2:46:40
I'm sure you have a good answer.
2:46:45
>> I mean, it's why do we need anything?
2:46:48
It's like there are lots of different types of t-shirts in the world.
2:46:50
Why do we need so many different types of t-shirts?
2:46:52
People like to customize things.
2:46:55
And I think that we now live in a world where customizing your own type of money is possible.
2:46:59
Some group of people are going to do it because it's just fun.
2:47:03
Other people are going to do it because it's a good business decision to reduce costs.
2:47:06
Other will do it because it's a good business decision and they're going to make a bunch of revenue.
2:47:11
And so generally speaking, the reason people I think are going to create more of these things once the barrier comes down to actually create more is just that they're fun.
2:47:17
They reduce costs and they make money.
2:47:23
>> What What is the actual mechanics if somebody goes to braille.
2:47:25
xyz today and creates a stable coin?
2:47:28
Like what's happening what's happening under the hood?
2:47:34
>> You know, it looks and feels a lot like using like a traditional fintech product.
2:47:37
It's almost like you're just depositing money.
2:47:39
If you've ever used uh PayPal, if you ever used Venmo, if you've ever used Coinbase, if you you have a Circle Mint account, it's the same for all these tools. You just put money in.
2:47:48
And in our case, you get a custom stable coin back out and then you can go use that in your business.
2:47:52
You can exchange it with any number of other stable coins.
2:47:56
You can use it on, I don't know, 10 12 different blockchains and it's compatible with our assets, circle access assets, Paxos assets.
2:48:03
So it's like it just works with all the popular stable coins.
2:48:07
And so, and then once it's created, let's say somebody creates a $1,000 worth of a new stable coin, how does like swap functionality work uh when you have these sort of like new pairs?
2:48:23
>> So, it's all actually built in.
2:48:23
So, when a new stable coin gets created and deployed, it's all reverse compatible.
2:48:29
So let's say that you go create your own stable coin.
2:48:31
You've got a thousand bucks sitting there.
2:48:33
You can use the API to swap it in between, you know, 50 other stables, again, some even that we don't issue at really no cost.
2:48:38
And you can swap it at the speed of basically the chain confirmations.
2:48:43
So there's no slippage and it's, you know, like we we think pretty ideal in that way.
2:48:51
And then obviously as these programs start to grow out into DeFi, there's a need to work with customers to set up different liquidity pools and things like that to make sure that they maintain PEG, work with them on on-ramps and off-ramps for building into fintech apps and all that good stuff.
2:49:04
Um, but you know, all solvable problems.
2:49:07
All these things were like impossible to imagine 5 years ago and now all the tech is sort of like right there off the shelf to use for people.
2:49:15
>> What are the what are the key kind of customers that you're excited about working with today?
2:49:19
kind of what are the categories?
2:49:23
>> We work a lot with blockchain uh ecosystems.
2:49:25
So we just launched uh light spark.
2:49:27
We're doing a lot of work right now with Canton which is like an institutional privacy uh focused blockchain and you know there are like eight more of these in development.
2:49:34
I think we've done another 12.
2:49:36
And then there's there's kind of like a fast follow once the chains are implemented to what are the use cases on the chain.
2:49:43
For us a lot of that is uh like payments.
2:49:44
People love these things because reduces costs increase revenue.
2:49:48
It's like a revenue source.
2:49:48
the neighborhood at had access to in a pre-stable coin world.
2:49:51
There are embedded finance applications which is just like building on blockchain infrastructure instead of kind of the last version of things.
2:49:58
And then there are all these other kinds of trailing use cases but right now we see a ton in the blockchain ecosystems a lot in payments and a lot in embedded finance financial institutions as well but let's be honest those are a little bit more hypothetical I think at the moment.
2:50:15
status of those like state-based stable coins.
2:50:18
>> And you mean like US states? >> Yeah. Yeah.
2:50:20
Wasn't it Montana that launched one?
2:50:21
And it seemed like Well, America already has a stable coin. USD. I was kind of confused. >> Wyoming, I think. >> Wyoming maybe. >> Yeah, you got it.
2:50:29
It's It's I think it's FR NT in Wyoming.
2:50:31
And you know, it's even like it once you can launch these things, states want to launch them themselves.
2:50:36
Banks want to launch them themselves.
2:50:38
Finch companies wants to want want to launch them.
2:50:40
And the rationale for it is always the same.
2:50:43
like reduce costs, make money, you know, and maybe the fun side is customization.
2:50:48
Um, you know, developers like to customize.
2:50:49
And so I I think that's why we really believe like there's just going to be any number of these things.
2:50:55
And one of the challenges is just making it uh interoperable between the deployments and making it easy for developers to build it into their app.
2:51:02
And you know, our our sort of like value ad is we just take care of the regulation and we just take care of the tech so that you can build whatever product you want to build on top of the new stable coin.
2:51:10
H >> what any any predictions on on uh stablecoin market caps over the next few years every I mean there's a lot flying around from you know tradi trying to just understand the opportunity to people that are you know more cryptonative that are that are super bullish themselves but but how are you thinking about it?
2:51:33
Yeah, I mean I think the number goes up, gentlemen.
2:51:34
It's like you you guys know how big certain like name brand fintech apps are.
2:51:40
As these name brandand fintech apps roll out their own blockchains and their own stable coins, the market cap of crypto generally is going to move.
2:51:50
And so suddenly these new like corporate chains, they're going to be top 10, top 20 chains the day they're rolled out, at least in terms of like aum and stable coins deployed on that chain. Yeah.
2:52:01
>> And so in a world where most of the world is still offchain, it's like imagine what happens when the repo market comes on chain.
2:52:08
>> It like you can't even compare what's possible in in terms of where we're at.
2:52:13
At least I don't think so.
2:52:15
>> Yeah, that makes sense.
2:52:16
>> Well, thank you so much for hopping on the show. We'll talk soon.
2:52:19
Congratulations on the launch. It's great to see you. >> Have a good day.
2:52:22
>> Good to see you again. Good to see you.
2:52:24
>> Um Bobby Cosmic in the chat has a point that I think I agree with.
2:52:27
All the all these different coins remind me of pre-1863 currency in the US.
2:52:34
Before the National Banking Act of 1863, banknotes were issued by thousands of different state chartered banks, which often had weak oversight and issued notes that could lose value or be difficult to exchange outside their local areas.
2:52:46
And I do I do wonder if there's some sort of like, you know, network effect.
2:52:53
I mean, it's like everyone spins up their own stable coin and then you need interaction between all of them.
2:52:59
So then like someone's trying to take an extra cut at some point and and and you would uh and m maybe like it all trades through and everyone gets lower cost, but uh if there's one standard for like interchain uh exchange, then they could take a then they could take a a cut.
2:53:16
Um it is it is all it is odd that we're in this like explosion of stable coins at this point.
2:53:24
Um but no one want everyone's sick of paying high transaction fees I suppose. >> Yeah.
2:53:28
I think I think the you know my immediate thought is I can understand why every institution and company wants a stable coin if you're a fintech company and you're moving a lot of money. I can understand that. >> I Yeah.
2:53:39
I mean at the same time it's like >> there's a lot of there's a lot of different types of companies where of course they would want a stable coin but are they providing value to the users right? >> Yeah. Yeah.
2:53:48
The the the the promise of many stable coin projects though was like hey use ours. It's low fees.
2:53:53
So it's like are we is is that the arbitrage that you go even lower fees below because it's not 3%. Right. >> Yeah.
2:54:02
So, um, and I know that I mean that was the original pitch for Bitcoin was like it should cost nothing to it's electronic money.
2:54:09
It costs nothing to move.
2:54:11
>> But they're definitely they're definitely going lower.
2:54:12
You know, Braille is going lower in the stack, right?
2:54:15
It's like Circle was like here's we're going to we're going to issue the stable coin and we're going to give you infrastructure to >> Yeah. >> to uh leverage them.
2:54:21
And then this is going uh you know a layer deeper and just saying we'll just make you the the stable coin itself.
2:54:27
In other breaking news, uh Clo uh mentioned us in their blog on going viral and uh and shared some stats about TBPN posting a lot of clips uh with views ranging from 5K to as high as 500,000.
2:54:42
There's a reason for this.
2:54:45
Some of the content does not deserve to go viral.
2:54:46
Roy Lee is stressing the idea that you cannot inh inherently make an undeserving tweet go viral.
2:54:55
viral sense gets you from one to a 100, not from zero to one.
2:54:58
Uh he he's stressing like the importance of being able to identify viral concepts.
2:55:04
Obviously, they put a ton of work into the content that they post.
2:55:06
Um but uh the interesting takeaway here is what are the benefits of going viral? There are only two.
2:55:12
Top ofunnel for your users.
2:55:14
Two, to get your mission seen by potential hires.
2:55:18
Importantly, virality does not equal top offunnel.
2:55:21
Only converting content will get you downloads.
2:55:23
And you must make sure your viral bets are converting or at least have a chance to be in the future, which is interesting with clearly because it's this two-step act.
2:55:32
They go they need to go viral and entertain and create something that's like funny or or controversial or something that everyone watches, but then they also have to get you to go to the website, download and pay for the for the product. >> Sure.
2:55:44
>> And and that and >> yeah, so you get an update here. Where does Cle stand?
2:55:46
Cle has more than enough consistent usage that we can reliably test every single one of our new features against a pool of users and know where the product is retentive and where it is not.
2:55:55
Cle is post PMF in a few areas interviews certain quizzes and homework that require an undetectable AI.
2:56:02
Uh students that's that's um >> enterprise clients who've spent time building out custom workflows for but conquering any of these will not end in the grander vision.
2:56:13
We have our eyes on the ultimate computer interface for multimodal AI.
2:56:17
>> Same uh same >> vision that Zach has his eyes on. >> Yes.
2:56:22
>> Um >> with AI mode for the metal >> always on live AI. We need faster.
2:56:25
Roy says we need faster magic moments and larger tangential consumer markets.
2:56:30
To that end, we spent the last few months working on making sure the product works for our more general users.
2:56:34
Yeah, >> there's a magic moment that exists with Cluey but does not exist for all the markets that interest us yet.
2:56:40
Uh the truth is it is actually quite hard to build software that feels truly magical to everyone.
2:56:46
It takes a lot of time and it uh >> that feels honest.
2:56:48
>> That feels honest and I don't think anybody would would >> There's an interesting stat in here.
2:56:52
He says in fact some of our meme viral videos with the most views 50 million views generated almost zero downloads despite being viral.
2:56:59
So conversion if you're using viral marketing conversion actually matters.
2:57:05
There's probably a question about, you know, Built Rewards uh did that big Tik Tok campaign.
2:57:10
They have 88,000 followers, 1.
2:57:12
4 million likes on their roomies series and uh and and and the the headline numbers look really good.
2:57:21
This could be extremely high converting.
2:57:22
It could be extremely low converting.
2:57:24
It's an it's an open question as to uh and when you're doing viral marketing, you have to understand that um the the the the bottom of the funnel matters just as much as the as the actual views and uh top of funnel. >> Totally.
2:57:39
>> Any other breaking news you want to cover? >> I don't think so.
2:57:41
I think uh everybody's uh relaxing after a busy day yesterday. >> Yes.
2:57:47
Uh well, we will see you tomorrow. Fantastic day.
2:57:52
and we'll see you on Friday. >> Thank you folks.
2:57:56
Great to see all your names.