Creative Flux

Bilal Tahir (00:00)
hello.

Pierson Marks (00:01)
Hey, hey, episode 17.

Bilal Tahir (00:04)
Woo! 17.

Pierson Marks (00:06)
Wait,

I was talking to somebody at this event on Tuesday and they have a podcast too. They're on episode like 14. So there a few episodes in front of them. It's like, hey, we got to get to episode 20. That's the key. Get there. But yeah.

Bilal Tahir (00:18)
Yeah, it's

those milestones 20 and then probably 50, 100,

Pierson Marks (00:25)
Right, right, totally.

But I know we were just talking about this. You were at the 11 Labs event at Oracle Park.

Bilal Tahir (00:32)
Yeah, it was a really cool event. They had a booth in Oracle Park so you had the view of the whole stadium, and this is at 6 o'clock so it's empty, except for this private equity firm was having their own baller.

up on the grounds and they had like tables laid out for dinner and like you know barbecue like full service going on and I was like wow that's like the suits you know right there and these are the developers although I mean we were doing fine too you know open bar so all of a sudden I met a bunch of people from the level app so there were a bunch of people there I met the the go-to-market lead Ben Hubbell he was pretty cool he was like the I guess the

Pierson Marks (01:00)
that's fine. Wait, did you get to meet anybody from 11Labs?

Bilal Tahir (01:12)
on leading the event and ⁓ I'm forgetting the names which is saying

Pierson Marks (01:15)
Is he like a short hair white guy?

Bilal Tahir (01:17)
And maybe, I don't know, he was wearing a really nice suit. I will say that. I was like, go to Margate. That's like your uniform. Looks good. But he was still. And ⁓ I talked to a bunch of other people too. was just like, because they had a bunch of them there. they were just like very cool people. it's funny, I kept asking them about the culture, about 11 Labs and internally. And they all had great things to say, which says something. Because they've grown a lot. I think they're around 300 people.

Pierson Marks (01:24)
All right.

Right.

Bilal Tahir (01:44)
One

girl told me there were 70 people at the start of the year and now there are 300. So they've grown like or some close to that. So they've grown a lot. And it's interesting because they're going through the same challenges we are about getting buttoned up about enterprise and business. And so I was like, yeah, you guys better button up because we're buttoning up and you you're one of the providers. ⁓

Pierson Marks (01:51)
That's crazy.

Did anyone

that you talked to like know of Jelly Pod anyway? Really? That is cool. I remember when I went to, yeah. It's fun because it's like.

Bilal Tahir (02:10)
Yeah, some of them have heard about us, you know, just cool, you know, I guess. And yeah.

Pierson Marks (02:17)
One of the guys, our account manager who signed our enterprise deal with them. I mean, he was like, yeah, you guys are like the AI podcast people. I was like, yeah, there's not that many people doing podcasting. I mean, that's a great use of AI voice. I'm like, yeah, it is. mean, you kidding me? If you don't have a podcast in your business, like, Kava, Kava launched a podcast this week about like, what are you going to listen to? Like they're talking about Mediterranean bowls.

Bilal Tahir (02:42)
Really?

Yeah, like how many bowls are you gonna it's gonna be a different bowl like podcasts every week I mean, but I keep telling everyone it's funny I like You know basically pitch the I'll have a conversation and then I just end up pitching a of jelly party I'm like, hey, you should have an AI podcast. You should do this This is like podcasts are new blog blah blah blah, know, so but it's true, you know, I feel like there's something but ⁓ you know, these events are interesting because you see like ⁓ You know all kinds of people like people who are just like in the industry adjacently others are in it. I met this one

Pierson Marks (02:51)
Yeah.

You

Is

Bilal Tahir (03:16)
person, she was like a professor and she had invented like evals for voice AI and her specific goal was to come to like try to sell it her whole like like the services to 11 labs and I was like that's very cool because you literally built something for them right and you're going to this event so probably you know we'll have a good shot.

Pierson Marks (03:33)
Interesting.

Would the evals then be at test time? are you testing for naturalness and quality?

Bilal Tahir (03:41)
That's what I asked

her. was like, cause evals are hard for voice. And she was like, yeah, all those things, naturalness, accuracy, blah, blah. I'm guessing she, you know, she's an academic, I'm guessing she's, she's written papers about this. So she was really like, I guess, come in, come up with like good benchmarks about voice. Yeah. But, you know, it's like one of those things, especially a jelly pod, you know, ⁓ you know, we've tried, you know, we try to give a very natural, you know, sounding voice with emotions and it's hard to kind of quantify that. Like, what does that actually mean? And it means different things for different people. You know, if you're doing like a healthcare

podcast maybe you want a more professional sober voice but if you're doing like a reptilicon or whatever right you maybe want someone more emotional you know with a fun voice so yeah

Pierson Marks (04:21)
Right.

No, totally. I was just listening to some stuff, some podcasts today. I mean, like AI, it's just like, some people just, the podcasts suck anyway. So I mean, like AI is going to be chill. Like it's, like it's better than their existing podcasts. So, um, yeah. And it was really cool to see, I know, like we always talk about this, uh, like, like Noel reaching in the top 100 on Apple.

Bilal Tahir (04:38)
Right, yeah.

Pierson Marks (04:47)
for business podcasts, like tens of thousands of listeners. Like it's just cool. Like it's a, there is a growing demand and a growing like, Hey, if you can create good stuff, independent of image, video, music, podcasts, if it's good, it's good. Nobody cares. It's like, if it's bad, it's bad. No one's going to listen. What if it's good, it's good. And it doesn't matter about the medium of creation. It's just that

Bilal Tahir (04:48)
Right.

Right.

Pierson Marks (05:14)
Some may take more less time, effort, and it's, yeah, so.

Bilal Tahir (05:20)
Yeah, and that's what AI Podcasts can give you. You don't have to spend a shit ton of time and find out it didn't go anywhere. You can just put it out there and see if it gains traction, then invest accordingly. So I think that that's why a lot of people like it.

Pierson Marks (05:30)
Right. Totally.

Yeah. And so I mean, this week was, is SF Tech Week like every other week, but this one's kind of more formal as people would call it. I don't know.

Bilal Tahir (05:44)
Yeah, I mean, a

lot of things I guess you would assume happening, although I mean, in terms of launches, maybe not as many. Although, I mean, there was Dev Day, which I guess is like the big, you know, the big news, I guess, you know, the week so far.

Pierson Marks (05:56)
Right. Yeah, so

I mean, I think it would be interesting to recap that. In the generative media, I, Sora 2, available via API. I think it's right.

Bilal Tahir (06:03)
Right.

Yeah, and Sora 2 Pro. For some reason, I

didn't catch that the first time, but there are two Soras. So Sora 2 is like the base model and Sora 2 Pro, guess, is better. So.

Pierson Marks (06:18)
When we say better, like is it better in quality or is it better in speed or is it like, what is it?

Bilal Tahir (06:23)
Yeah, I think

it's more about prompt interference and stuff like, you know, maybe the object will actually follow gravity better and stuff. think it just for me, because maybe the slightly quality as well. But I think that's like the big difference between a bigger model and the, you know, the quantized version of that model, basically.

Pierson Marks (06:42)
Right. What's the price? Do you know the pricing difference? it like?

Bilal Tahir (06:44)
Yeah, it's

like 10 cents per second for the Sora 2 and 30 cents per second for Sora 2 Pro. Which is, it's a little steep, but still, mean, compared to Vio3, you know, Vio3 started at 75 cents a second and now it's down to 40 cents. you know, it's good. And I'm guessing they're subsidizing it because we've talked about this before. OpenAI always, I feel like never has cheap models and they, you know, their inference always costs a lot of money. So I do think they're taking a loss on this probably.

Pierson Marks (06:59)
Yeah.

interesting.

Bilal Tahir (07:12)
And especially with the app, because the app is like, can do 10, 50 generations like for free, which is, that's a lot of compute.

Pierson Marks (07:19)
So

did you generate some stuff on the app so far? Yeah.

Bilal Tahir (07:23)
I did a couple, feel like, I don't

know, I'm more into the API side, so I was waiting for that. So maybe now I'll play around with it more.

Pierson Marks (07:32)
Yeah,

it's interesting. wonder how like they're training these like, where do all these companies get their video data? Like I'm generally

Bilal Tahir (07:39)
Well, now you know,

I mean, that's one of the places where I feel like the reason they're letting you create all this video is so they have training data for the next iteration. So it's like a virtual cycle. So does me.

Pierson Marks (07:51)
So

how does, but how does that work? Like the synthetic data, I totally get, like some people, think, I think most people now coming around to the idea that synthetic data is a fine way to do it. But in this case in video, you, like what's the quantitative traits that you can actually say, Hey, this was a bad generation. Is it going to be like the user generates into the host?

Bilal Tahir (07:55)
Yes.

I think

really basically I think, and this is why I've been thinking about this, the social app idea makes more sense in that, because then you get stuff like likes, comments, remixes, shares. Those are great signals on what video is doing well or not, right? So I think that's actually why they made it into a social app, because they basically get a reward function for free. And that's why they're letting you create these videos.

Pierson Marks (08:41)
through the terms of service so I wonder if so I remember I saw this on Reddit your check like you essentially just have an open AI account the same one that you use.

Bilal Tahir (08:49)
Yeah but apparently you can't delete Sora,

you have to delete your Chargifty account or something like that. Or you're banned from it or something. I don't know if that was true or not, some guy said that.

Pierson Marks (08:53)
Right, right. ⁓

Yeah, I saw that too. I don't know if it's true. If like you delete your Sora account, you delete your ChachiBT account and you're never able to use it again. It makes sense if it's just like your open AI account and you delete your Sora and it's kind of like all like co-mingled like that, it delete your ChachiBT account, which I could get. I don't understand the part where if you delete it, you can't re-sign up. That doesn't make sense to me. But like the interesting thing on like the, what, two things. So,

Bilal Tahir (09:17)
Right, yeah.

Pierson Marks (09:24)
If users are generating content, the signals that a social app

in AI's world may give you, are those the right signals to know whether or not something is good? what will, like what is OpenAI's goals with Sora? Is it to be a realistic model that helps them understand the world, to achieve AGI, to like, hey, if there's a video where Sam Altman's on a skateboard stealing GPUs and it runs through a wall, like that's like not a good thing and it probably won't get posted or probably like it's just weird and we want to get rid of that out of the training data.

But I'm just curious because if you ever go into the dark corners of the internet and you see something weird.

things that are at least before today like they weren't AI generated and there's like weird stuff like Shrek is love Shrek is life if you remember that video that was like a weird one and it was like animation it was just weird and those got virality those had a lot of signal of like being popular is popularity on a social app like Sora 2 on the mobile app is that a good signal for

Bilal Tahir (10:09)
Hmm.

Mm-hmm.

Hmm.

Pierson Marks (10:32)
them to like refine the model or is it like might get into like a weird space where all this stuff is weird.

Bilal Tahir (10:34)
Yeah.

Yeah, I mean, that's a good point because if you think I if your objective function is like, want to just make realistic videos and stuff, but like a cat video of doing Olympics like, you know, pops up because it's funny, like that.

Like how do you distinguish between that versus like, no, this is the actual good generation. You know, we want to use that training data. So you're right. Cause all the most popular videos I've seen are just ridiculous, like fat people, lump eggs or something like that. But so yeah, I'm, I don't know. mean, you may, maybe there, maybe there, there is a way for them to filter through, but it's a good question. At the same time, I'm, it could also be something as simple as they're like, let's just get the data and then we'll figure it out. So.

Pierson Marks (11:20)
Right.

Cause it also could be like, rather than is the output good, could be like, how many iterations did the user take to get to the thing that they thought was okay to publish? Like prompting and like, Hey, they, they wanted this banana to be green, but we kept showing them a yellow banana and they could took a few times and they had like prompt. ⁓

Bilal Tahir (11:30)
Yeah.

No.

Yeah, that's smart. Yeah, right. I mean, that's

a great reward signal. We've talked about that for Jellypot as well. Like, know, people edit transcripts is a great, you know, ⁓ know, signal to follow. But yeah, that was an it's an interesting feature. I just saw it on the API as well. It's called Remix basically. But it you can't just upload any video. It only takes Sora two generations and then it can remix it. And I do think that's actually pretty a very powerful tool because it was one of the biggest growth hacks at TickTock.

Pierson Marks (11:47)
Right.

Bilal Tahir (12:07)
because people would take a TikTok video and add a reaction to it or something. And it would get ridiculous where people would react to the reaction to the reaction to the reaction to the reaction to the video. It is just like a chain. So I imagine like it would be very cool if they have that functionality where you can see how a source radio started here and then users remixed it to a certain state later.

Pierson Marks (12:17)
Alright.

Right, no totally. mean, wait so was that in the API too?

Bilal Tahir (12:33)
That's only it's called remix. It just got added. So yeah.

Pierson Marks (12:38)
So, I mean, so

they do watermark them.

Bilal Tahir (12:41)
Yeah, although people have already figured

out how to remove the watermark, which I thought would be the watermark actually, I thought was kind of, you know, smart, because I've only seen static watermarks, but this one just went like this. And I thought, oh, you know, that that'll be hard to remove. then somebody figured it out. And I actually talked to the guy who implemented it Dev Day after party. I was talking about it and he's like, yeah, I did that. I'm like smart, but like, why didn't you randomize it? You know, because I mean, it's, you know, you should just write math.random that shit. And so that's just a little harder.

Pierson Marks (13:09)
What do you

say?

Bilal Tahir (13:11)
He was like, I think maybe he got annoyed by that or something. I can't recall. But he just didn't give me a reply. Or maybe it was too noisy.

Pierson Marks (13:21)
That's funny. So how does the API know that you're remixing a solar video? Is it the steganography, kind of like this watermark inside the video?

Bilal Tahir (13:29)
I'm sure there's a signature.

These videos all have a signature, invisible signature inside that they can probably distinguish. It's the same with ⁓ fake videos and stuff. All these videos will have a signature or invisible watermark at least.

Pierson Marks (13:37)
That makes sense.

Bilal Tahir (13:45)
That's how they'll distinguish. Hopefully that's that'll be much harder to remove I've also seen papers where it's like it's not as simple as something a warm-up where it's literally built into the like the pixels themselves like like it's inside the way and that that makes it like way harder to Delete it and I think that's important to have you know, because you want something to add ⁓ You know to know it's fake but speaking of that Theo Theo the YouTube influencer, you know, he

Pierson Marks (14:01)
Right.

Totally.

Bilal Tahir (14:13)
released this avatars on it was called add Sora a Sora water maker and basically what it does is it actually adds the Sora watermark to any video and that was the kind of a interesting you know use case because I'm like well why would you do that but then you know it it kind of is funny because you can take real videos and add the watermark and make people you know think it's fake which is another interesting kind of worms

Pierson Marks (14:36)
That is

interesting. implications of that, you're like, no, I didn't say that. Look, it was an AI generated video.

Bilal Tahir (14:42)
Yeah, well, like the joke is a

lot of like people like, you know, they have Sam Altman has like compromising videos and that's why he did Sora because he wanted his avatar to be used so that he could deny it.

Pierson Marks (14:56)
Yeah,

no, that's interesting. I always love the cryptography, of putting the information inside the pixels. You kind of change the value slightly. Or it's even sometimes invisible, where it's just not even noticeable to the human eye. The data has changed slightly to insert kind of signature.

Bilal Tahir (15:17)
Right, right, yeah.

Yeah, we're adding some noise and stuff just to confuse the LLM, you know, like artifacts.

Pierson Marks (15:25)
Totally. I know

we talked about this, but I chance to talk to the fall founder, I his name, but asked a quick question about the world models and his answer surprised me. we were talking about Genie 3 quickly and how...

they were able to achieve persistence in the world and like painting on the wall and stuff. Whereas this world labs and their marble, the publicly available like marble world, that was like the, what was it? The hobbit hole one that we were talking about a few weekends ago. World labs. Yeah, five, of these. Right. And.

Bilal Tahir (15:54)
Mm-hmm.

right, right. Fire, fire, please. War labs, yeah.

Pierson Marks (16:07)
There are completely different architectures, I guess, where world labs is like pre-rendered. And so it generates the world once. And then you like walk around in that world. then like Genie 3 is actually like a world model. So I think at the surface, they look kind of very similar, but like Genie 3 is just so many more epics ahead in terms of like, it's like actually generating the pixels at every frame where it's world labs and marbles. Like they generate the whole world once, but.

Bilal Tahir (16:14)
Mm-hmm.

Interesting.

me.

Pierson Marks (16:36)
That's how they can achieve, that you can walk around and everything kind of stays the same. But yeah.

Bilal Tahir (16:40)
Hmm.

Interesting,

Although I mean, I imagine like it wouldn't be too hard to go from that to just using that as a context for the next iteration, just kind of generate more of the world. But it is crazy how fast they do it though. I mean, both, like we've talked about how Genie 3 is so fast and same with the world labs. Meanwhile video where Sora is like 30 cents a second still. So, you know, they're still like that ceiling.

Pierson Marks (17:05)
Right. Totally.

It's super interesting. Well, OK, cool. So last week we talked about SOAR 2. We've talked about, did you know that Sesame got acquired by Apple?

Bilal Tahir (17:21)
What? When did the... really? did this happen?

Pierson Marks (17:22)
This has to be voice people. Somebody said that to me

the other day, and I was just like, wait, what? So maybe Sesame is going to be in Siri. That crazy good voice model.

Bilal Tahir (17:32)
Interesting.

I mean they kind of like dropped out the map like they had this really cool release when they came out and then yeah went nowhere

But it's interesting because we were talking about, I was talking to some people about the competitors to 11 Labs and they're all kind of like dropping off like Fives. Like Cartagia, nobody really talks about them anymore. They were trying to do more just the fast real time voice because they had, their model were SSM based Flash and the founder is like, he invented Flash attention. they had all this.

theoretical knowledge, but some just couldn't find PMF, think, or at least that I know of. Same with Sesame, guess, you know, I mean, you know, guess they found out like they couldn't really survive on their own, so might as well get acquired, which, you know, they probably got a good deal for that.

Pierson Marks (18:20)
I mean like this is this is like the thesis I think Too many people compete at a foundationally hard level and if you can win you win big if you become the next AMD or Nvidia and you build this really hard thing and and chip like or you become like you're competing at like your entropic or open AI and you're building like the language models

Bilal Tahir (18:29)
Hmm.

Pierson Marks (18:45)
There's all this non-differentiated work that is expensive to get to creating a model, to create a chip, to create like these amazing technologies where the opportunity, if you do win, it's very large. It's a winner takes all or winner takes most market. But I think that, and those are the things that seem the most attractive to most people. If you're an engineer, you're like, I want to work on these hard problems, which is totally fair. It just is like,

Bilal Tahir (18:53)
Yeah.

Hmm.

Yeah.

Pierson Marks (19:15)
There's so much more value that can be created with a higher likelihood of success at the application layer on top of these platforms. Like OpenAI, they're going to build the most amazing model and you'll have Anthropic there, whatever, like what happened to Mistral or Mistral or these like, they can differentiate themselves through like compliance, like Mistral is doing where, you know, we're going to be the EU, we're going to make sure that everything is like compliant to EU standards and we're going to go down that, but it's not like a technical, it might be some technical

Bilal Tahir (19:20)
Hmm.

Right.

Great, great.

Pierson Marks (19:45)
challenges for sure but like it's this non differentiated kind of research work that the payoff's huge but it's gonna be hard to compete because people are gonna fall off same with voice 11 labs they raised what 300 million dollars or something recently and if you're a Cartesia you're someone else like what do you what dimensions are you competing on you're competing on realism

Bilal Tahir (19:58)
Thanks.

Yeah.

Yep.

Pierson Marks (20:09)
naturalist, you're competing on the number of language supported, which is a finite number. So it's like, you can't like always compete. Once you can support every language, you support every language and everybody else has to catch up. Latency, like you just like they're not trade offs really, like you could still have a very natural, well, there probably is, but like you could see that you could have a very fast model that's very natural that speaks all languages. And the first person to get there wins the market.

Bilal Tahir (20:14)
Mm-hmm.

Pierson Marks (20:36)
Like I think what's another person going to do? Like, mean, they got ⁓

Bilal Tahir (20:37)
Right. Yeah. Yeah. Yeah. Well, provided they can

execute on the relationships, which is what they're doing. Like it's all about enterprise relationships and, you know, just building the reputation that we are the company, we handle your voice data. It's all secure. We're all compliant, blah, blah. So those things just become, you know, the what matter. Because you're right. I mean, it becomes a commoditized feature, basically. Same with the music. mean, Suno was first out of the door. mean, UDO and others are falling, but it's hard for them. I mean.

So maybe it's like winner take all there or it'll be like they get acquired, know, which I think is a, there's a big chance that Suno or Udo just get acquired because they realize they can't just exist without them because they are getting sued out of existence and all that. And so they go to Spotify or Disney and they're like, all right, you you understand studios, you have the relationships, just take us, you know.

Pierson Marks (21:30)
Absolutely. mean, I think we talked about this earlier week, like Spotify, the CEO stepped down, CEO and founder, Daniel and the two, the now co-CEOs, think, or maybe it will be in 2026, I think, not immediately, but like, I think Spotify and the musical landscape is going to completely change. Like I think that Spotify is a hundred billion dollar company. They've crushed

Bilal Tahir (21:46)
Right.

Mm-hmm.

Pierson Marks (21:58)
They completely changed music industry forever. Went from 99 cent songs and albums to streaming anything at any time. And I think that they're going to move towards, we did the distribution. We get music to people's ears. Now, like we talked about in past, we're going to help people create more and go direct. Hey, if you're an artist and you're using cool tools and you just want to go distribute directly through Spotify versus the label.

Bilal Tahir (22:04)
Right.

Right.

Yeah. Yeah.

Pierson Marks (22:27)
come to us, don't go to the record labels, we're going to get you money, more money than the record, we're not going to take as much as the record labels, you're going be in the algorithm, you're going to be promoted, and I think they're going to buy up the AI tools. I genuinely think that you could see Spotify saying, we're going to invest in creating an amazing ecosystem for more people to create music, because why wouldn't they do that? They want more people to create music.

Bilal Tahir (22:53)
Well,

I and I think that's the big question, because I'm with you. I'm both Spotify. think they are, it's theirs to lose. But I think the big, big question is like they have the, their biggest thing is like their relationship with the artists and they are dependent on the artists currently. Right. And if they're so they are, because they're such a big elephant in the room, artists have done this for their stage walkouts, Taylor Swift's taken a music off or whatever. Right. I mean, if they start, if they become like the face of AI music and stuff and all the artists repel,

Pierson Marks (23:21)
early.

Bilal Tahir (23:22)
easier to rebel against Spotify, they can like go get into this weird place where they know the next stage is AI music and they want to do it but they don't want to piss off their existing you know artists and so they don't move as fast and that that could be the opportunity. I mean that's that is like the one question I feel like I have on them and where maybe a new company can come in and have and you know take advantage of their innovators dilemma. It's not really an innovators dilemma because I don't think they they don't see the writing

Pierson Marks (23:34)
Good luck.

Bilal Tahir (23:52)
the wall it's more that they have to proceed with caution and that's why we talked about lofi i think lofi was a great first

like that should just be a gender because there's no big lo-fi music. mean, nobody listens to lo-fi music. Like I actually do have a couple of artists I follow, for the most part, people don't follow an artist for lo-fi music. And so even if the artists get pissed off and leave Spotify, they're like, ⁓ whatever, you know, we have enough lo-fi tracks, but Taylor Swift leaving, that's huge, right? I if they started doing like pop songs and stuff, know, then, you know, that's like a little dicey territory. they, and I think that's where I think they might do something like a hybrid where they

And I actually saw this 50 Cent of all people tweeted this out. He took a song he hadn't released and made ⁓ like a 1930s jazz version of it. It was like a hip hop song, but it's like jazz version. And I'm like, you know, I've always thought of like fans doing that, but like, why can't the artists do that? So maybe that's like another way they can get in where the artists come in, they have their tracks and they're like, okay, we'll take this old track you have, remix it, A-B test it, find out what works.

Pierson Marks (24:57)
All right.

Bilal Tahir (24:57)
do a new release, we'll like streamline that for you, new revenue, et cetera, et cetera.

Pierson Marks (25:01)
Yeah, the rem, I mean, and that goes back to like the remixing of TikTok and like the growth of like.

Bilal Tahir (25:05)
Exactly remixing

is powerful. You know, there's so many interesting use cases We haven't even think about like of remake. We've done remixing images video Music just shows, you know, I've think I've talked about this before, you know it Call when content basically becomes as cheap to produce as code, you know, you can do stuff like I'll fork Game of Thrones and make a new season eight, you know Why not because I hate it I like the ending and then I'll put it out there if it gets more votes Maybe it'll get merged to main, you

Pierson Marks (25:36)
Right, no, totally. I love, there's some Star Wars fan films that I've, I was like, damn, this is good. There's this one 10 minute long clip, it's Darth Maul, and it's his last challenge by the Emperor, essentially Darth Sidious, before he becomes like a.

Sith Lord or whatever. And it's like a completely fan film, like videos, like 10 minutes long, and he just slaughters all these Jedi's. And it's just this crazy, like awesome sequencing of like lightsaber battles. And I'm like, this is so sick. Like this is way better than the new trilogy or not the new trilogy, like the, yeah, the new trilogy. So.

Bilal Tahir (26:00)
Right.

yeah.

Right. Yeah.

Well, then, yeah,

the new newtory because I think of the first three films. yeah, mean, that's like, but yeah, you're right. I mean, it's such a huge universe and Dartmo. I mean, he was such a cool character. only got like that one cool fight scene with Obi-Wan and Qui-Gon. And so, you know, we want to see more of him. So why not?

Pierson Marks (26:22)
and right.

Yeah. Hey, I mean, he was

in the animated Clone Wars series. And that's like a great.

Bilal Tahir (26:38)
I haven't watched that one. It's been on my

list for a long time. I didn't watch Clone Wars and I feel like I missed out a lot of Star Wars lore because of that.

Pierson Marks (26:46)
or the Clone Wars is so sick.

I've seen every Star Wars out there except for there's like this new show that they got like something. No, I saw Andor. There's like squad something. I don't know, there's like more children's like non-animated.

Bilal Tahir (26:54)
and or.

I

know what you're talking about. Yeah, I think I saw that one. But yeah, mean, maybe that's a way we replenish these old IPs that are making, they're just pumping out bad films, so let the fans take over. But also, mean, it's an interesting time for new artists who just come up and they just are AI first and they're just producing music using AI and they don't care as long as they're going to be able to monetize. So it'll be interesting. It'll be very interesting.

Pierson Marks (27:07)
Yeah,

Yeah.

Right. Right.

For sure.

One of the things that I thought would be interesting to talk about, because we went to the SuperBase conference on Friday. We got to hear some speakers, people that were there. So SuperBase for anybody that's listening, because I know it's a database plus all this other nice stuff on top of it. Makes it easier for developers who are looking to create their first application to have a database authentication. And it's pretty cool. They raised 100.

million dollars at a five billion dollar valuation. That's what they announced on Friday. Congrats to them. They had some awesome speakers there. They had the founder and CEO of Figma. They had Patrick Colson, the founder, co-founder and CEO of Stripe. And they just had some amazing, I thought they had some pretty cool talks. The Figma, so Dylan,

He was talking about design. He's talking about design in the age of AI, how it tastes as your moat and all these cool things. Figma has been around for a while. It's a publicly traded company. It was the first time I ever heard him speak. He's much more sorry to him. He kind of feels like a little teddy bear, kind of. And Gary Sossfoken, I would love to see him under like high pressure. I haven't ever seen an interview where he's like grilled. It would be really cool because...

Bilal Tahir (28:40)
Yeah.

Yeah.

Pierson Marks (28:48)
It seems like he's

a designer. He's a smart, he's a software engineer and designer. ⁓

Bilal Tahir (28:51)
Right.

Pierson Marks (28:52)
And he was talking about this like future where like design, how important this design is. And I was like, oh, I really resonated. It was a 30 minute talk. I resonated with him and his vision. Monday morning, I invested some money. Monday morning, right when the stock went up, I was like, okay, cool. He sold it to me. They were also major investors or investors who were based in Series E. So Monday morning when the markets opened, invested some money. And then a few hours later,

Bilal Tahir (29:07)
wow. So he sold it to you.

Pierson Marks (29:22)
it popped up by 20%. It's like, happened? And then the next day, it popped up another 20%. This week it's up like 45%, right? Crazy. Because on Monday, OpenAI announces major partnership with Figma that you can now create Figma designs in ChatGPT and work with ChatGPT and port them over everything from flow diagrams and all this stuff. It's like, crazy.

Bilal Tahir (29:43)
Yeah. Aw, man.

I just like,

I feel like it's a game. I feel like it's one of those games where I keep missing or at least I don't know. I'm like, I feel like opening eyes all these years, they're like, they do a deal with Oracle. Oracle shot and stock goes up, AMD boom. And I'm like, I just want one hit. Like just give me a little bit of alpha. I moved to SF for this. Come on. Just give me a little bit of tiny whisper in the ear at a bar. I can, you know, just, you know, go all in. But it sounds like you're, you know, you're doing well, you know, that's awesome.

Pierson Marks (30:04)
Thank

I

I just was like,

I thought it was so funny because on the car ride home, I was talking to my parents like, ⁓ I really had this fun time at this event with you and just meeting these other people at SuperBase. And I was like, well, I mean, I can't invest in SuperBase. I like the vision. This is new generation of like the new cloud, the Neo clouds. These are the people that we'll look back on, I think 10, 15 years from now, like, wow, like a lot of infrastructure is built on those because nobody really wants to go to AWS and learn all that stuff.

Bilal Tahir (30:13)
He

Pierson Marks (30:40)
It's powerful and it's cheaper, but you're not going to start there. And I was talking to my parents about it. like, I'm just going to put some money in and whatever. I was like, Monday morning. And then something completely unrelated came on Monday with OpenAI. was like, oh, wow. And then AMD and that major partnership, they came out and just like.

Bilal Tahir (30:41)
Right, right.

Right, right.

That's great, yeah.

Some guy made, put in $10,000 in AMD calls and you made $10 million. I like, God damn it, that's insane. Yeah, I mean, that's a lot of money to put in calls. I mean, just for one, I mean, you must know. I don't know. But I mean, at the same time, in hindsight, hindsight is 20-20. You look back, I'm like, yeah, of course the hyperscalers are, you know, they're more compute, more demand, you know, we're building chips. So at the very least, I mean, I look back and I'm like, wow.

Pierson Marks (31:05)
How do you make those like...

Yeah.

Bilal Tahir (31:25)
Nvidia was obviously gonna go up. I mean, it still might go up. Like it might go to 250, 300. Like even now it's like 180. And so the question is like, I feel like a lot of some, know, like someone said, investing is less about the brain and more about the gut. You know, I mean, it's like about the real, just the ball, the cone is to just do it. And I, you know, I mean, so sometimes it's just about taking those big swings.

Pierson Marks (31:49)
Totally. Investing is, it's, you know, like just long-term investing. That's kind of where I stay. It's honestly, I feel like the alpha that we have in this environment here in SF, yes, it may be a bubble, like every bubble.

usually leads to like some cool growth afterwards because you have a lot of investors losing their money funding stuff that's not going to be profitable for a while but like at least for us it feels like sometimes we have a foggy crystal ball into the future like today i saw figure three did you see the figure three uh video it's cool i mean like is it ready no is it like inspiring of course

Bilal Tahir (32:19)
Absolutely.

Yeah, yeah. What did you think about it?

Yeah, there's definitely a lot of marketing. I mean, I'm excited. I was excited

for it, but I feel like they just keep doing these cherry pick videos. Like guys, it's like ultra slow, like tomato pick. I'm like, all right, cool. But this is like the hundredth time I'm seeing this.

Pierson Marks (32:35)
Yeah, yeah, yeah.

you

100%. And

it's just to get some like they're hopeful that the next video and the next one gets a little bit more coverage and you could raise more money and invest and just keep that going for a long time because it's a hardware like that's also I think an area where it will be a winner takes most because it's very expensive. You have to get the data, the robotics data to like train the models. That's just very hard. You're competing against Tesla who has cameras on every car.

Bilal Tahir (32:59)
Mm-hmm.

Right.

Pierson Marks (33:08)
They're gonna so it's like you're gonna have to get a lot of public you're gonna have to raise a lot of money You got a lot of public support. You're gonna have the people Want and be excited for these things you have to show that there's like actual like excitement For this stuff, but it feels like we have a crystal or a foggy crystal ball sometimes until

Bilal Tahir (33:25)
Well, yeah, mean, on

that note and similar, I mean, the other robot that's doing really well that we both have used is Waymo. I mean, it's amazing. And now it's funny, I get into Waymo, I don't even look up, it's just so normal for me, right? It's crazy, you just adapt, but it's...

amazing technology is getting rolled out and I mean people don't think of Google as like you know they didn't call robotaxi and figure but like Google's up there I feel like the just that one division is gonna be like a huge trillion dollar opportunity for them you know so yeah you know you can use that and then make an investing decision which you know no financial advice I mean I personally did a little bit in Google so we'll see you know how that goes

Pierson Marks (34:02)
Good night.

It's really

interesting because the whole argument is the Elon versus Sundar argument where Tesla is not going to put LIDAR radars on the cars for self-driving. Google Waymo did. Google Waymo's been around for longer in terms of their self-driving abilities. And I just wonder if the LIDAR, if like,

Bilal Tahir (34:14)
Right.

Pierson Marks (34:25)
think it kind of makes sense. get the Tesla argument like if humans can do it with their eyes, cars can do it with just their eyes, especially if you can communicate car to car and they kind of have a network mesh like what's going on. But I wonder on the Waymo argument if you have, they have cameras, they have LiDAR, they were able to penetrate the market earlier, have that upfront cost to be like, hey, we're going to, you know, gain the trust of people earlier. We're going to get brand recognition.

Bilal Tahir (34:36)
Hmm.

Pierson Marks (34:52)
but then we're also going to use the LiDAR data in addition to the cameras to make a better initial world model so that future iterations of Waymo can be cheaper to produce that only use cameras because we already have all this LiDAR data and that you pair that with the camera data and so that like you're going to have more nuance and like higher fidelity information when training the models.

Bilal Tahir (35:16)
Yeah, yeah.

Right, for sure. And I think it's one of those things I feel like it's, I don't think we, I mean, especially when I, I can't even wrap my head around the second or third order effects of what happens when you have autonomous vehicles.

on at scale, because what's hard to imagine is how the behavior changes, like small things. Like somebody actually posted about how on the highway, the safest place is behind a Waymo and in front of a Waymo because it just never crashes. ⁓

Pierson Marks (35:45)
All right.

Bilal Tahir (35:45)
On the other side, I've heard actually my brother was saying this is funny that, ⁓ he was like crossing, ⁓ you know, he was about to cross the street and he hesitated because it was red. But then he saw it was a Waymo. He's like, fuck it. And I'm like, well, people do that. Like on the edge, they make they're making these decisions based on the fact that it's just more safe. And how is that going to span when you have thousands of these vehicles? Like, how's that going to work? You know, and some of it will be worse. Like these other cases, others will be better because Waymo is one of those things where I think it's like a

Pierson Marks (35:58)
Yeah.

Bilal Tahir (36:15)
network

effect, the more way most they are, safer it is because the more predictable their environment gets because they're the ones on the road now. And at some point, you know, maybe we'll fundamentally just think about traffic lights. Like why do we, like we have these inefficient traffic stops and stuff just because that's what humans, you know, like are used to, but maybe there are ways where, you know, autonomous robots can just keep going and just stop, you know, when they need to because they're so safe. So.

Pierson Marks (36:38)
Right. Have

you ever seen that video as a fake video of like a crazy intersection and all the cars like this?

Bilal Tahir (36:44)
yeah, people are walking around.

I was like, even if I knew that it was safe, I would still not like do it. I wonder how they actually filmed that. I mean, this is like before Sora. So I mean, that was definitely actors walking around. Maybe they photoshopped it and made it closer or something.

Pierson Marks (36:53)
Right.

I

think the one thing with all the troubles and problems that this country and the world has, when you have issues and...

There's such optimism. think I take the optimistic approach. Like, hey, when you're here, there's a lot of room to grow. And when you talk about cities, and you talk about building and infrastructure and autonomous vehicles, you can imagine a future that if we all come together and are committed to improving society in some ways, you might have different opinions on how to do it. But like,

Bilal Tahir (37:18)
Mm-hmm.

Pierson Marks (37:34)
autonomous vehicles, the information that those gather, lights and signals, like who hates, like everybody hates waiting at a red light when there's nobody around. And there's all these things that just make the world less efficient because it has to overcompensate for the few percentage of people that will take advantage, like tragedy of the commons, like they'll take advantage of their environment at.

Bilal Tahir (37:55)
Bye.

Pierson Marks (37:57)
you know, but damaging everybody else around them, prioritizing themselves, self-motivated individuals. And then like, you could just kind of, um, make a more, like just like a more efficient world. And I think that when everybody is so concerned about how AI, like what am I going to do when AI takes our jobs or like all this stuff? Like, first of all, I the premise is wrong on like, what am I going to do when AI takes our jobs? But it's also like,

Bilal Tahir (38:10)
Right.

Right. Yeah.

Pierson Marks (38:25)
As long as there's inefficiencies in the world, there's things to do. There's businesses to be built. Like unless the world is perfectly efficient, which will never be, there'll always be the opportunity to build something and to do something better and make something better. And that's like the argument I always have about like all that stuff is like, well, I don't think AI is going to make this world perfectly efficient for everybody. So. Right. Totally. Well.

Bilal Tahir (38:36)
Right.

Just gotta find those cracks, you That's awesome. Awesome.

Pierson Marks (38:54)
Episode 17, Creative Flux. On that note, yeah. We'll end it here. Sweet. Yeah. Totally. OK, I'll talk next week.

Bilal Tahir (38:56)
Yeah, on that note, on that optimistic note, know, like, let's, yeah, let's get out there, find those cracks, fill them up.

Talk to you. Take care. Bye.