Latest / AI for Founders with Ryan Estes / Floot helps non-coders build full-stack apps with AI

Transcript

Floot helps non-coders build full-stack apps with AI · AI for Founders with Ryan Estes

0:00
Welcome back to AI for Founders. I'm Ryan Estes.
0:02
Today we've got Huge, the cofounder of Flutecom, fresh out
0:06
of Y Combinator and not fully out yet.
0:08
He's already making moves. Flute is an AI powered platform
0:12
that helps anyone, coder or not, build real production ready web
0:16
apps just by chatting with AI. Huge calls this vibe coding and
0:20
don't we all. Not just throwing A1 shot prompt
0:23
at the machine, but actually Co building and iterating with it
0:26
until your vision comes alive. What started as a scrappy
0:29
experiment on Reddit is now one of the top products on Product
0:32
Hunt, complete with the specialized framework, built in
0:35
hosting and serious full stack muscle.
0:38
Today we'll dive into the future of Vibe code and why non coders
0:41
are leading the way and how Flute plans to stand out in a
0:44
crowded AI dev space. Let's get into it with huge from
0:48
Flute. But first, some business.
0:50
This episode is brought to you by Kit Caster.
0:52
The easiest way for founders, CEO's and experts to grow their
0:55
brand through podcast guesting. Imagine landing interviews on
0:59
the shows your customers actually listen to, without cold
1:02
pitching, without chasing hosts and without wasting hours of
1:05
your time. That's what Kick Caster does.
1:07
We've already booked over 15,000 interviews for 800 clients,
1:11
helping founders raise capital, close deals and build authority
1:14
in their industries. If you've got a story to tell or
1:17
a product to sell, Kick Caster puts you in front of the right
1:20
audiences with Done For You podcast outreach.
1:23
Ready to go on podcasts? Head over to Kick caster.com and
1:26
let's get you booked. That's Kick Caster by MO Burst.
1:30
Now, you and I know startups are built in boardrooms.
1:33
They're built in the wild during 3:00 AM.
1:35
Bug fixes, botched launches, and that one meeting where
1:38
everything almost fall apart. That's what Code Story is all
1:41
about. It's a podcast where tech
1:43
founders breakdown the make or break moments behind their
1:46
products. What went wrong, what went
1:48
right, and what nearly wrecked everything.
1:50
Hosted by Noah Labhart, Code Story goes past the highlight
1:54
reel and into the honest stuff, late pivots, brutal bugs, and
1:58
somehow coming out alive with a product people actually love.
2:01
If you're in the build stage and the scale stage or the please
2:04
Let this feature work stage, I can relate.
2:07
This show will speak to your soul.
2:08
Listen to Code Story at codestory.co or wherever you get
2:12
your podcast. Just search Code Story.
2:14
It'll be right at the top. Because great startups aren't
2:17
born, they're forged one hard moment at a time.
2:19
Code story.co AAI. Live tonight Shining Stars
2:27
computer live AAI take the flight in the globe with me on
2:35
Brian. All right, Huge.
2:39
How are you? Good.
2:40
How are you? I'm doing great.
2:42
Thanks so much for joining me on the show.
2:44
Yeah, thank you for having me. Absolutely.
2:46
I'm really excited to learn about flute.
2:48
Now I before we get going I want you to define for me what is
2:51
vibe coding. Vibe coding to me, I think it's
2:55
when you have an idea in mind and you're just like using
2:58
natural language to have the AI built out for you.
3:01
To me, I think, and a lot of people like to do like this
3:04
whole one shot prompt thing like oh, I one shot like Netflix or
3:08
Airbnb. But to me I think it's more
3:10
important that when you're vibe coding, you're actually Co
3:13
building or iterating with the AI like this.
3:15
The idea is never fully formed and you're kind of like seeing
3:18
it, seeing a text take shape as you build it out more.
3:21
So to me, I think a lot of the vibe coding NDC is like about
3:25
having the AI build up your vision and also refining a
3:28
vision. Amazing.
3:30
Now you are the founder of Flute, which actually just
3:33
launched as flute.com, floot.com and would you would it be safe
3:39
to say this is a vibe coding platform?
3:42
Can you say again? Would it be safe to say that
3:45
this is a vibe coding platform? Yes, we do have some small
3:50
discussions about ourselves like do you want to go with like with
3:52
the Best Buy coding platform or something like that.
3:55
But I think so far we've, I feel like vibe coding to me still
3:58
sounds a little bit not serious. And we really try to build this
4:02
platform as a place for you to be like serious apps.
4:05
So, you know, we don't actually have vibe coding like on our
4:07
website or like, you know, tag line.
4:10
But I think it's safe to say it's kind of, you know, I think
4:13
according to my definition earlier, I think it does fit
4:16
into the category. 100% now we're big fans of vibe coding.
4:21
The the kind of the early leaders seem to be bold,
4:24
lovable, VO, replit. How does this kind of fit in and
4:28
where do you differentiate? Yeah.
4:30
So I think if you look at the by coding platforms, they are kind
4:35
of different. I feel like they all serve
4:37
different audience in a sense. Like we have some users coming
4:40
from Replit and it was just like, oh, Replit is so
4:42
complicated because I feel like Replit is very technical.
4:46
I think both to some extent is also fairly technical and it's
4:49
but they cut many specialized more and like having your
4:52
prototype something quickly and then you kind of move on to like
4:55
closer or like whatever kind of coding tool you have.
5:01
V-0 also, you know, coming from the cell, obviously, I think
5:05
they're still like more leaning towards the like being more
5:07
technical side of things. I think the level was probably
5:09
like closer to where we are, where they're like more
5:12
non-technical focus. I'll say that our
5:15
differentiation there is that we actually provide a more serious
5:17
hosting platform once you finish building.
5:21
And I think we also invest a lot more into like the whole
5:24
building experience when you're building a food stack, like
5:27
everyone claims that they have food stack, right?
5:29
But it's actually not that easy to do it, right.
5:31
I think the fact that we have our own hosting platform and we
5:35
even deal like a very special framework.
5:37
It's kind of almost like next to this, but like for AI to write
5:40
code that's like Didi integrated into a platform, like it really
5:44
makes it a lot easier for I think.
5:47
I'm sure it's the same AI model but like they were able to do a
5:49
lot better job on our platform. Amazing and congratulations.
5:53
I mean you just launched it's fresh.
5:56
So maybe talk us through a little bit the the idea for the
6:00
platform. I know you're a graduator of YC
6:04
and then how you kind of decided through that YC cohort to
6:08
actually go ahead and launch Fluke?
6:11
Yeah, also actually wifey is wrapping up in a couple of weeks
6:15
during the process. Still in it.
6:17
Nice. Yeah, so crazy, crazy busy right
6:20
now You talk to. Are you asking me about how this
6:23
idea came around and how how wifey journey was like?
6:26
Yeah, yeah. So and also wait, I think the
6:30
idea kind of started like way earlier maybe like start of this
6:32
year, end of last year ish was mostly experiment.
6:37
It was actually more fun and focus, I think at a time it
6:41
wasn't like I'll I try and give this to like some friends, some
6:44
people in the community, and I don't think it was that well
6:47
received. People don't really have a need
6:49
for that. But thinking more of it like
6:51
cursor but with better funding integrations.
6:54
But interesting thing is that once I put it onto Reddit, I got
6:57
a lot of response from people that already code and they were
7:01
like very excited by like the fact that they can like, you
7:04
know, like use natural language for like something that looks
7:07
pretty nice. And then they're like, the
7:09
question was basically like, all right, like how do I connect the
7:12
database? How do I like launch it as like
7:14
a proper website? And back then it didn't even
7:17
have pages or whatever. It was really like very funny on
7:20
my single page and people would like try to tag around it to
7:23
like have a multi page app. And so that that way it kind of
7:26
brings me like, all right, like there's quite something we can
7:28
build here that actually has a lot of demand for.
7:31
So I think so and I worked with my Co founder on this idea.
7:34
Together we can build our whole back.
7:36
And finally we like we really upset with like making it
7:39
actually work well. And then we did like a free test
7:42
on Reddit. Funny enough, we like going to
7:45
this competitors. They just mentioned just going
7:47
to their son, like the thing if you want to test for free.
7:50
And the response was actually really strong.
7:53
I think there's like shows that like a huge demand there.
7:56
And I think one thing that's really interesting that happened
7:58
was some of them, I mean, obviously it's free, some people
8:02
try it out, some of them drop off, some of them like it,
8:04
whatever. But like some of them actually
8:05
say, all right, can I pay for this?
8:08
So like, all right, there's something there.
8:10
That's when we kind of incorporated, we start like
8:12
charted parts users. I think, I think kind of right
8:15
around the same time was like the YC application down line.
8:17
So we also decide to apply got in and that's kind of go through
8:23
the whole YC process. I think there's a lot in YC and
8:26
I can talk about you, but you know, through this process, you
8:30
know, obviously we go and launch IC and then we just try to push
8:33
it out to like Product Hunt and that's where we are now.
8:37
Fantastic. It's incredible.
8:38
And, you know, do you remember which subreddits you were kind
8:42
of communicating with folks in? Anyway, the premiums went to a
8:46
lot of competitors subreddits. I think there's both reps.
8:51
I don't know, we're getting banned right now.
8:57
Yeah. Like it's just there's like
9:01
people there are obviously very high intent and I think they're
9:05
not happy with where the these products are unless there's
9:08
still a size 1 number of people there are not super happy about
9:11
them. Yeah.
9:12
You know, you're the second person I've talked to that has
9:14
found validation on Reddit. What do you think it is about
9:17
Reddit? Yeah.
9:18
I mean, they, they the criticism of Web 2 communities that are
9:22
based in anonymity, that people are generally mean to each
9:25
other. But Reddit seems to kind of
9:27
prove the opposite. Yeah, I don't, I think red
9:29
people already are, are nice. They, they're not mean.
9:34
I think it's Reddit. I, I, it's like a nice place
9:37
where people bring community together.
9:40
You know, like you're into wire coding.
9:42
There's, I think there's actually a couple of other wire
9:43
coding Subs that we have tried to promote our product before
9:47
and you know, people there are very open to trying things out.
9:51
So I do think it's a good place to have the product out there.
9:54
Obviously, you soon need to convince people to like that,
9:58
that you're worth a try. Yeah.
10:00
Why do you think that is? I think just the sense of
10:05
community like well, I think part of that is there are always
10:08
going to be someone that's interested in trying things out.
10:11
They would about your thing, they would like comment on it.
10:14
And I think that probably help kind of brings the more hesitant
10:17
people who are sprouting around more engaged.
10:21
So there's quite like a network effect there.
10:23
Yeah, 100% very cool. So your, your launch schedule is
10:27
super aggressive. I mean you start from like a
10:30
couple of months ago with the idea you run it into YC and you
10:34
launch before demo day. What's what's the the reason for
10:38
the the launch so soon? Well, I mean, to be fair, we
10:42
have always been launching even before then, you know, like I
10:45
said, we did a free test and we start trusting people.
10:48
And then I think we still go on Reddit to like find more users.
10:52
Obviously that we're on Fri C and Fri launch is bigger scale
10:56
and we got a lot more visibility.
10:59
But I think it's just launch like you know if something best
11:03
way to validate it is to launch it to have user using it, to
11:06
have people actually waiting to pay for it and they will tell
11:09
you what you can be profile. Yeah, 100%.
11:13
It's so cool. I mean, I let's talk about like
11:17
the back end. So what model is is does Flute
11:21
use? We actually use all of the the
11:24
best models. So we we have, we have some we
11:26
use operate from open AI. We're going to look into TPT 5
11:30
soon. We use cloud a lot too.
11:33
We use the German and models from Google.
11:36
We also use a couple other small models for like more auxiliary
11:39
task in there. But our philosophy is always
11:42
been like, don't cheap our models like use the best model
11:45
to give you people the best result is.
11:47
Does it fluctuate or do you have kind of a process that it starts
11:51
with depending on the prompt or depending on the the sequence of
11:54
in chain of events, it's running through different, different
11:57
tools. Yeah, you're.
11:58
Right. So, you know, I think when
12:00
you're talking to flute, it actually has, you know, it like
12:03
kind of breaks things down into tasks.
12:06
And we have a way to kind of route some of these tasks to
12:08
like different kinds of tasks, different kinds of models
12:11
depending on, you know, things like latency, cost benefit
12:15
analysis and like whether this model is really suited for the
12:21
past. Like one example here is when
12:23
debugging, we try to use the all three models a lot because in
12:26
the thinking so far, I still think that open air for now, the
12:29
whole like RL training and thinking the best.
12:32
So in terms of more like complicated things that you need
12:35
to think through, I think operate does a pretty good job.
12:37
Yeah, amazing. So I personally absolutely love
12:42
vibe coding. It's the best.
12:45
Now there's a fantastic vibe coding group here in Denver that
12:49
meets every other week. I think Denver also has the very
12:53
first AI Co workspace that Danny Newman put together and it's an
12:58
extraordinary place. Big barrel top warehouse, you
13:01
know, creative kind of place. I mean, it's appropriate to go
13:05
in there and do some vibe coding, but also like if you
13:07
were to like rehearse your punk band, that probably would also
13:10
be pretty cool, you know? Now what I love about vibe
13:14
coding is it brings together everybody, you know, So in this
13:18
group, you have senior developers, you have junior
13:20
developers, you have product people, you have people that,
13:24
but there's a guy who brews beer, you know, there's, there's
13:27
people with ideas, there's people that are very well
13:30
versed. And the exciting thing is, is it
13:32
kind of flattens the table to say that like we all have
13:37
different skills and areas of expertise and curiosities.
13:41
And with vibe coding, at least you can kind of come together on
13:44
projects and see what you can make.
13:46
And because we're in this infancy of what this is the,
13:51
the, the capabilities are still kind of wide open, like we still
13:55
haven't hit the walls of like, what exactly can we do?
13:59
And then also that that begs the question of what are the kind of
14:03
products that are AI native that are going to really take off in
14:06
this kind of like big pipe cycle as far as AI?
14:10
Now, that being said, I love AI or I love vibe coding, but it
14:14
does have its haters. Now what?
14:18
What would you say that the vibe coding haters get right?
14:22
What does the vibe coding haters get right?
14:25
Yeah, well, so far I think the only time that we have run into
14:31
haters is when we try to launch the more like developer centric
14:35
communities. I think I do.
14:40
I do see a point from them that, you know, like the AI generated
14:46
codes are not necessarily the easiest for human to understand.
14:50
I think AI tend to write code in a, in a, in AAI way, if I might.
14:56
And it's kind of like if you are a developer, you have this
14:59
colleague who likes to write code in a way you don't like and
15:02
it's not the best, you know, you're not going to enjoy that
15:04
working relationship. So I, I don't think there's
15:09
something there. Maybe that's why this some of
15:11
these people don't like it. I do think there's also another
15:15
kind of emotion there about like potentially this would kind of
15:20
make developers less valuable. Maybe.
15:22
So there's like a whole fear. But I personally think that the
15:26
vibe coding is here to stay and it's it for anyone like, you
15:29
know, obviously for people that don't write code, it's going to
15:32
unlock a huge possibilities. But also for people who write
15:35
code, like including me personally, we also use fluid to
15:38
build flute, right? And it's a huge unlock for our
15:41
productivity too. And just like a couple of days
15:44
ago, I tried to like run some numbers using our database and
15:47
we just had flu. Like I cannot look at our
15:48
database and write a query. Honestly, I don't think like
15:53
I've done that before myself maybe, but it's going to take
15:56
way long. So yeah, I think there's some,
16:00
some something, some of some of something.
16:02
I can understand why where some of these haters are coming from,
16:06
but I also think that overall it's probably like a more
16:08
positive thing for for us than negative.
16:12
Yeah, I totally agree. And you know, I think it is
16:14
largely it feels like an existential threat, you know,
16:17
and but they also get to benefit from the tool as well to like to
16:22
your point, they can use flute and and 10X to productivity.
16:25
Now. It, it seemed like in the
16:27
beginning we were talking about more of a utopian future where,
16:32
you know, because we can 10X our productivity with AI, we, we 10X
16:36
our rest and relaxation. But it sure doesn't seem like
16:39
that's happening. It seems like the opposite.
16:42
It seems like actually now we have much more on our plate and
16:45
we're able to process that information and at a higher
16:49
level much faster. I, for one, find it very
16:54
pleasure pleasing and intoxicating.
16:57
How about you? Yeah, yeah.
16:59
Well, I mean, I think it, if you look back in time, like, you
17:04
know, 4040 years ago, people are writing games using assembly,
17:08
right? And nowadays you can buy pro a
17:11
game yourself in like 1 hour. So that's why huge change there.
17:16
But even in between there was obviously all these new
17:18
technologies that make creating games a lot easier.
17:22
And then this results in like more interesting games, but
17:25
ultimately people still kind of spend time building them and
17:29
playing them. I think it, it brings, I feel
17:34
like it's not just like it's just, it doesn't just like give
17:37
you more things. It gives you, it allows you to
17:39
do things more in depth too. And that's made why maybe like
17:42
even with all of this productivity boost, you will
17:45
still be doing kind of maybe even the same thing.
17:47
But now you get to do a lot like going a lot deeper in a shorter
17:51
time with those things. I agree.
17:53
You know, And I, I also really, when I go to the zoo, I like to
17:57
look at the primates because I really relate to them how
18:00
they're just so lazy. They just roll around and eat
18:03
lettuce and apples, you know, and just basically that's me on
18:07
Sunday, you know, so, So what I've noticed is that like I have
18:11
this ramp, you know, that like begins Monday morning kind of a
18:15
on a like a productivity hangover from the weekend
18:19
because and then ramps up through Thursday through Friday.
18:22
Friday I'm getting pretty torched and then boom, on
18:26
Saturday, Sunday, I'm hanging out with my kids, I'm going on
18:28
hikes, I'm doing a little archery, you know, kind of
18:31
whatever. But I've really found that like
18:33
it takes me like a day or two to start really ramping up to that
18:37
capacity that I'm firing on by let's say it's Tuesday afternoon
18:41
now. I feel I'm really well tuned up.
18:43
I can have, you know, 17 tabs open and do 4 projects at the
18:47
same time and I feel really invigorated by doing that.
18:51
Do you find that there's a similar hangover from just like
18:54
relaxation that you need to ramp up your skills again through the
18:57
week? Maybe before, but I see.
19:01
Honestly, I don't think there's relaxation work every day, every
19:05
day, every hour, every. Day every hour non-stop.
19:09
Well, that's good. So at least you can keep a pace
19:11
going, you know. What about your demo day?
19:14
When's that coming up? I think it's, let me actually
19:18
look at it again in about 3-4 weeks.
19:21
Yeah, now. What 3 or 4 weeks?
19:24
What is the ideal outcome for you?
19:26
Obviously you'll have offers to, you know, for, for equity share,
19:32
for investment opportunities for for flute.
19:35
What are you looking to get out of it?
19:37
Yeah. I think what Demo Day is really
19:39
about getting investments. So you know, you know, we work
19:42
with our YC partners to kind of come up with how we should go
19:47
about doing that and demo that is really like a chance for us
19:51
to like show this project to a lot of investors getting
19:55
interested. And I think that leads you into
19:58
more conversations and hopefully we close them and we get the
20:01
investment offers. Yeah, what?
20:04
What would be a good seed round for you?
20:08
I know a lot of people are talking about seed strapping.
20:10
Other folks are looking for the biggest check they can get,
20:13
particularly in a space like this where you have some pretty
20:16
heavy competitors. What would a good seed round
20:19
look like for you? Yeah.
20:21
So we're not actually looking to raise any like crazy numbers.
20:24
I think obviously there's a lot of competitors, but we also like
20:29
pretty lean and want to stay that way.
20:31
I think raising too much money has its downsides too.
20:34
Like there's obviously dilution and I think it's sort of like
20:37
sense of like losing urgency. And we we also don't like like
20:40
the idea of like we just like pumping millions of dollars into
20:43
getting users if those users don't have staying with us.
20:47
So I think to us probably raise like a pretty regular seat
20:50
around and you must get back to work as soon as we can.
20:54
Yeah. So.
20:55
The IT also seems like the good path to success is kind of
20:59
differentiating on kind of the ideal customer.
21:03
So if you had to pick one segment of people that would
21:06
really love your tool, who would those people be?
21:09
Yeah. So I think right now our focus
21:11
is already on like, founders and entrepreneurs, yeah, building,
21:15
you know, new products or building things for the existing
21:19
business. Cool.
21:21
Now maybe you can help me with some like ideas because I, I
21:26
probably build a couple dozen things and I, I kind of, I put a
21:31
task on myself months ago that I'm going to ship one thing per
21:34
week, largely so I can see what people are, are, are wanting,
21:38
but also to limit the, the size and scope of the, the products
21:43
I'm putting out. I have a kind of a hypothesis
21:47
that something very easy, simple, sleek, that just fits
21:50
into somebody's daily life in a very normal way is kind of the
21:55
stuff I want to experiment with. So by saying, OK, you got a week
21:58
to put this through it, it puts the proper parameters on the
22:02
project. Now what happens though, is I'm
22:04
starting to create very similar products.
22:08
So when you're talking about games like what is the, the
22:12
ability to, to kind of use flute to, to design games?
22:17
Yeah, I do. Games is a very challenging area
22:20
and I can see products specializing in those areas.
22:23
So far I have seen people viewing like simpler games.
22:27
I think the the difficult power games is once you get into the
22:30
more advanced, like, you know, the 3D animations and like
22:33
really complicated gaming logics, there is a bottleneck of
22:37
how you can communicate effectively with the AI on like
22:40
how exactly the game should work.
22:43
And I think that AI still has, I think there's still some
22:45
technological limitations where the AI have to deal with a
22:48
really complex logics. I don't think that there's like
22:52
interesting and simpler games that you can probably build with
22:55
today's technology if you're not like going into like the really
22:59
crazy high end. I mean, yeah.
23:01
Absolutely. What about like 8 big games?
23:03
Can I build Super Mario Brothers?
23:06
Oh, so I so kind of this is kind of for more like a technical
23:10
perspective. I think you can probably build
23:12
something that works like Super Mario Brothers.
23:15
The difficulty there and then is just designing the levels and
23:19
kind of coming out with a way to actually design that in whatever
23:21
game that you're viewing. Cool, I bet you could build like
23:25
some like Oregon Trail. Do you know about Oregon Trail?
23:28
Actually, I haven't heard of that, he goes.
23:30
Way back, this is like, you know, back in the floppy disk
23:34
days, but like you could it was a fun game where like you're
23:37
setting off on the Oregon Trail and it was kind of like a choose
23:41
your own adventure kind of word game.
23:43
But then every once in a while you go deer hunting and you get
23:45
hit the space bar and and shoot at the deer is absolutely
23:50
captivating in 1989. But it it was really cool.
23:54
So OK, so, so on a limited scale there, there could be games.
23:58
I haven't really gone into games too much, largely because I
24:01
don't know. I'm more interested in kind of
24:03
I, I would say tools. Now, one thing I've tried a
24:11
couple of times and I would love just to get some feedback or
24:13
maybe I can approach this problem.
24:15
So one of my problems I have is I just can't seem to eat enough
24:20
protein. Like if I look at the, the
24:22
recommendations, I'm supposed to be getting like 200 grams of
24:25
protein a day. That's a lot of protein.
24:30
So, so you know, I, I've tried a couple times of figuring out
24:34
either like a recommendation engine for like how to get it.
24:37
How would you approach kind of building a tool that can help
24:41
somebody who needs to get more protein?
24:44
Oh, that's an interesting idea. If I'm going to do this, I'll
24:48
first start with almost like a simple log app that you can log
24:54
in, like stuff I have eaten today.
24:57
And you know, I think AI can probably estimate roughly how
25:00
much protein it is and then have a way for AI to recommend like
25:05
you should eat this for dinner or something so you get enough
25:08
protein. That'd be a.
25:10
Good one. So yeah, could could you do
25:12
something or like it, you can upload a picture and it would
25:14
say the protein that would complement or something like
25:17
that. Yeah.
25:18
So I think picture will be another interesting to do that's
25:20
just like instead of logging just have have the AM look at a
25:23
picture and come out with a number of like proteins in that
25:27
picture. And OK, I'm, I'm, I'm going to
25:30
give it a shot right now. While we're talking, upload a
25:32
picture and have recommendations for additional protein.
25:40
OK. This episode is brought to you
25:44
by Moberst, the AI first growth engine trusted by iconic
25:48
startups like Calm, Robin Hood, Reddit, Uber and Google.
25:53
Moberst is a digital, full service, mobile first marketing
25:56
agency that helps companies to scale and become category
25:59
leaders. Whether you're finding product
26:02
market fit or ready to accelerate growth, MO Burst
26:05
delivers creative, data-driven solutions that help with scale
26:09
and efficiency and impact. At MO Burst, we're not just
26:12
guessing what works, we're predicting it.
26:15
As an AI first agency, MO Burst uses cutting edge tools to
26:19
analyze user behavior, adapt creative in real time, and
26:23
uncover hidden growth opportunities.
26:25
The result? Marketing that's faster,
26:27
smarter, and built to convert from early stage startups to
26:30
global brands. We've helped hundreds of
26:33
companies overcome complex challenges, unlock new
26:36
audiences, and optimize every metric that matters.
26:39
At Moberst, we believe that creative thinking, advanced
26:42
technology, and data are the core drivers of success at
26:46
scale. If you're serious about scaling
26:48
your business with precision and purpose, it's time to talk to
26:52
Moberst. Visit moberst.com and let's
26:54
build something category defining together.
27:00
OK, cool. Now one thing that I notice that
27:03
I'm doing more and more when I'm working on tools like this is
27:06
I'm bouncing back between GPT and in my case, I use lovable.
27:11
From now on, I'm be using flute, but what I'm doing is I'll say,
27:16
hey, GPTI want to vibe code an app in flute and I want you to
27:22
ask me questions to make it better.
27:24
So I'll go through like this whole process of like kind of
27:27
discovery and ideas through GBT and then they'll say, OK, make
27:32
the prompt. Then I take the prompt and I go
27:34
and put that over in in flute, for example, like this.
27:39
Is that a common use case for some of these platforms?
27:43
Yeah, we have seen a lot of people doing that actually.
27:47
Usually when you when you see someone have a very long prompt,
27:50
chances are it's generated by GPT.
27:52
And but we think that we can provide something that's more
27:55
built in that is go through the questions with you and then you
27:58
know, build out the thing for you.
28:00
One issue with that, I'm not sure whether it happens with
28:02
other platforms. I think it probably does is that
28:05
GPT come up with it's own opinions on like the kind of
28:09
stack that you should use and not.
28:11
It's not always like it doesn't always work on our platform.
28:15
And for example, we already have our own database.
28:18
So like, you know, GP, like I say, like use super base, which
28:21
I'm sure like bring a lot of good business to a super base,
28:24
but that's like unnecessary and like add small friction to your
28:27
viewing process. So we actually have to do
28:29
something there to reduce some of those prompting.
28:33
Yeah. But yeah, I think it's a pretty
28:34
valid use case that you you kind of have some process to have you
28:38
refined the idea and then you kind of dump that into a most
28:42
specialized agent. Yeah, you know, and maybe I'm
28:46
really looking for like new use cases too, because like I said,
28:49
I'm kind of pigeon holing myself into kind of a standard way of
28:52
I'm using it, which I think is fine, but I think it's largely
28:56
built on the tool. So you know, for example, with
28:59
Lovable you you create something, you can deploy it,
29:01
you can buy the URL. So you have a custom URL, it's
29:04
got a landing page and then the the app is housed within this
29:07
website so you can share it really easily.
29:10
But the, the more I've been thinking about some new tools,
29:13
they're not necessarily web tools.
29:15
Like for example, I wanted to make like a, a Chrome extension
29:20
for, for some, for different products.
29:23
Are there other deployment like platforms that I should be
29:27
considering when I'm thinking about building tools with these
29:30
AI designers like like Fluid? Yeah, so, well, so first of all,
29:34
I think Fluid wouldn't be able to do a Chrome extension.
29:37
We're just focusing on web apps right now.
29:41
I think for like some of this, like I'll say, I guess they do
29:44
qualify as more niche use case. Like maybe someone said, could
29:48
be like a Chrome extension by coding app thing.
29:52
But also I think you might be able to have get enough success
29:55
with cursor and what not. If you, you are, you know, if
29:59
you already think about Chrome extensions, you could have some
30:02
level of technical knowledge. So like maybe work around those
30:05
issues and typically these things don't tend to be too
30:08
complicated that you can probably do it well enough on
30:13
like most of the platforms very. Cool, so I I was working on my
30:18
protein app here on flute and it, it directed me to add an
30:21
open AIAPI. And is that basically so that
30:25
like it's pulling from like not only the the the tokens from my
30:30
API, but also so it can be informed by my personal API,
30:35
personal ChatGPT. Yeah.
30:37
So right now the way it. Works is, you know, all the
30:40
conversations with flute goes through the flute credit system.
30:44
But if you're viewing an app itself, in this case, if you do
30:47
something with open AI, then we could just let you use your own
30:51
key. We do plan to offer something
30:53
that's more building in the future, so you know, you can
30:55
skip this step. But for now that basically the
30:58
idea is whatever your app uses, you go through your own, your
31:02
own like API connections. Got it.
31:05
Amazing. So let's hear about you a little
31:07
bit. So prior to the the flute
31:10
explosion with YC and everything else, what were you working on?
31:14
Yeah, well. So I was an engineer at retool.
31:18
So it's a internal tools beauty also AYC company.
31:22
And I think I, so back at retool, I, I rebuilt the, the
31:26
whole runtime system in retool and I think that gave me a lot
31:28
of experience of like building and this like flute, flute as a
31:32
platform where the AI has access to all of this under the hood,
31:37
old infrastructure that you can like get feedback and make
31:40
changes quickly. Before retool, I was at Astana
31:44
working a lot on the web front end and like, so I should be
31:49
reviewed the Astana text editor. I think I worked on like the
31:53
Astana, you know, whether you use a sound of it, it used to be
31:56
more like a, a list kind of view.
31:58
And I think at some point Asana might kind of introduce the
32:01
whole spreadsheet view and I'm working on some of those two.
32:05
Yeah, before that I actually went to school in Singapore and
32:10
I kind of moved to the States forward.
32:14
Amazing. And then?
32:15
So what was the the what's the impetus to make the jump to to?
32:23
After retail, yeah, or or basically?
32:26
You know, those are some big companies and probably some
32:29
really fun projects. Why dive into start up world?
32:34
So it's something I always wanted to do.
32:37
I think it's just hard to find the right time and I feel like
32:41
at some point it's just right out of excuses.
32:44
I think to me it was like, I think last year I was like,
32:46
maybe I should stop exploring this idea.
32:49
And back then I didn't even know what I want to work on.
32:51
So I just decided to take some breaks and see if I can find
32:54
something interesting. Nice.
32:56
And this is certainly. Interesting, isn't it?
32:58
OK, what I want to do, you know, you're a new founder.
33:01
This is your first, first start up.
33:03
Is that correct? Yeah.
33:04
OK, So we're going to we're going to do the founder
33:07
challenge questions brought to you by kidcaster.com.
33:10
Are you ready? All right, all right, you get
33:14
there's four different categories of questions to
33:16
choose from. So whether you want tough,
33:20
explosive, risky or reflective. Let's go risky.
33:27
Let's go risky. I like your style.
33:29
All right. Have you ever made a decision
33:31
purely to spite someone? Purely to spite someone probably
33:38
did. I was not a maybe.
33:40
Or did something something something mean it back at
33:42
school. Who knows?
33:45
That's that's kind of a main question, isn't it?
33:50
Let's see what else we got here. All right.
33:54
Oh, what's a decision you? Had to make that you knew would
33:57
hurt other people, yeah. I did have to fire someone when
34:04
I was, you know, in my previous job.
34:09
I mean, I think it's a tough decision, but it's something
34:12
that I can make, Yeah, you know? Right now, how big is the team?
34:18
All right now, me and my Co. Founder that's it 2 amazing so
34:21
you're. Going to be starting to looking
34:22
at at hiring very soon, you know, maybe in the next four to
34:25
six weeks. What would your first couple of
34:27
hires be? Who will they be?
34:29
I think it would find very. Hard to find people in my
34:31
network, you know, just we want to close them early.
34:35
We want, we don't want to go through like lengthy interview
34:37
processes that will be like the top choice and barring that,
34:41
that's for the other, other channels.
34:44
So quite focusing on that and focusing more on like engineers,
34:48
engineers, more engineers. When does it make sense to like,
34:53
get into marketing the tool get into marketing?
34:57
Marketing, yeah. Oh so.
34:58
We do want to hire. Someone for marketing too.
35:01
I, I just think of it as like in terms of priority, but I, I
35:05
think marketing is very important for this tool, 4%.
35:09
Cool. Have you ever publicly praised
35:11
the tool or platform that you secretly despised?
35:16
No, no, hey, you chose the risky.
35:20
Questions. This is on you as if they they
35:24
don't sound. That bad to me.
35:25
Yeah, I I got. To tone them down, I'm going to
35:28
do a question generator on flute.
35:30
Now what flute is? flute.com is a great domain.
35:33
How'd you find the name in the domain?
35:36
Well, so. We actually had a different name
35:38
before and it was like, oh, if we change our name, we otherwise
35:41
the partner and he was like, oh, you find it asking this.
35:44
Yes, you've changed that. All right, sounds like we should
35:47
and we spend a bit of time thinking about it.
35:51
Flu was just turned out to be a name that's easy to remember
35:54
that shot and we can find out a domain for a reasonable amount
35:57
of money for so we'll go with that.
35:59
Yeah, what you pay for the domain.
36:01
Because it's what, 5 letter domain that is not cheap.
36:04
Yeah, it's nothing. Crazy.
36:06
We pay I think around 30 K, maybe 30 K 2010.
36:11
Less than 30K. 20K yeah, man, that's I mean, that's.
36:13
A great name. Yeah.
36:15
Honestly, do you have a a favorite flautist?
36:19
Flautist. Yeah, one who plays flute.
36:23
Oh, no, no. So one thing we were actually
36:25
thinking about this was like, do we want to have the associated
36:28
with like flute instrument? Look at our website design,
36:32
there's like no feeling of military instruments.
36:35
So yeah, don't, don't do it. I I mean.
36:37
I personally, I'm a jazz fan, and I really like jazz flute.
36:41
You know, I love like Herbie Mann, of course, because he had
36:44
great record covers. But I like Eric Dolphy.
36:48
He played a mean flute. Now, if I play jazz flute around
36:51
my wife, she gets visibly angry. She cannot handle jazz flute.
36:57
So yeah, you just keep the flutes out of it.
36:59
But it is such a great name. No, thanks.
37:03
Cool. I'm going to.
37:05
We're going to go into the lightning round now.
37:07
All right. If you had to return to school
37:10
as a freshman knowing what you know now, what would you study?
37:14
I'll skip school. Would you really?
37:16
Yeah, why? Well, I already know.
37:20
What I know Why do I go to school again?
37:23
So there's nothing in the in the future that you're like, wow,
37:26
maybe I should study this to get me there.
37:29
I. Wouldn't do that in the school.
37:31
I'll do that. I think it's much more fun to do
37:35
it, you know, as a way to build something.
37:39
There are certainly certain subject where you have to crack
37:42
out to school for, but I I can't really see myself starting.
37:47
But those things I'll probably just do it outside of school.
37:50
I love it, this audience. Loves business and product
37:54
ideas. What's something I could vibe
37:56
code on flute to make me 1,000,000 bucks for the next 12
37:59
months? Well, if I know that I would
38:03
have done that. I think, I think business in
38:08
general. Is is tricky, right?
38:11
It's I think you have to kind of find the right niche that you
38:14
are good at that you can find a channel to sell and you know,
38:20
that's that people actually want and that's always been hard.
38:25
I would say that probably find something that you are like
38:27
really good, like you're a domain expert on, maybe do
38:30
something around there and you know, if that thing that you
38:33
know has value, then you probably can make some money out
38:36
of it. Yeah, is.
38:38
You're an ideas guy. Is there an idea hanging out
38:40
there you wish you had time to get to?
38:46
Right now I'm my focus right is on float and I don't think I've
38:49
been like thinking about ideas lots outside of that.
38:54
I mean, I think there are a lot of interesting things you can do
38:56
nowadays, given like how much more opportunities is unlocked.
39:01
I think there's like quite a lot of very interesting way you can
39:04
look at existing big stars companies and just like make
39:08
something a lot smaller because then you can now do it that you
39:12
have a lot more connection with. Like maybe you know, you're
39:16
doing podcasts. Maybe you can do your own
39:17
podcast stars that you on a new use or maybe you can sell to
39:21
your friends but like doing similar podcasts instead of like
39:24
having them to buy? Any other expensive solutions?
39:27
Cool. Let's say you sold flute today
39:30
for a billion dollars. What do you do tomorrow?
39:36
Tomorrow I think I'll take a break before I get back to it in
39:40
a in a couple of minutes. What does the break look like?
39:42
What do you do? I spend more time with my.
39:45
Family travel a bit. Yeah I'm pretty boring outside
39:50
of work honestly. I'll just do it like regular
39:52
relaxation stuff. Yeah, that's.
39:55
Not boring at all. It's necessary considering your
39:58
last six months have been pretty much pedal to the metal.
40:01
Now that being considered, what do you know now that you wish
40:05
you knew a year ago? Well, I wish I started on this
40:09
idea way earlier. I think, you know, that's
40:11
probably going to make it a bit easier too, given how crowded it
40:14
is nowadays. Yeah, that's why that one, the
40:20
only thing that I, I mean, maybe, maybe I think, well
40:23
sometimes. This has been awesome.
40:26
I got a couple more for you here, Here for you.
40:28
Huge. What's the most?
40:31
Unsettling thing. About AI for you right now.
40:35
Most unsettling things. I think it's the listen.
40:44
I feel like AI has been progressing a lot, but I feel
40:47
like it's some of the like promises early on.
40:52
Like, you know, the costs have been going on a lot, but it's
40:55
not really going down as fast as some people want.
40:59
But imagine that when it does happen, like imagine that
41:02
whatever by calling to you're using today, it's like 100 times
41:05
cheaper, 100 times faster. Actually.
41:08
It's very far to wrap my head around like how that's going to
41:11
change the landscape of software.
41:15
Like do people still be a software in the case for the
41:19
other people? Like because it's so fast, like,
41:21
oh, maybe we just have like huge, like very complicated,
41:25
unimaginable software. Like everyone has their own like
41:28
everyone can do their own like Facebook scale software at that
41:31
point. I totally agree the.
41:35
The A is desperate need for power for compute.
41:39
It might be the, the, the biggest outcome of AI is
41:42
actually creating basically free unlimited power for everybody.
41:46
You know, because the race that is on now we're probably just a
41:51
few months away to like major investments in in nuclear to, to
41:55
fund all these data centers. So I mean, we are seem to be
41:59
like moving very quickly to very, very, very cheap energy,
42:03
which sounds pretty amazing. Yeah.
42:07
While the oddest T batter is going to be eaten up by AI
42:09
again, we're not like that. Still won't be enough.
42:14
You know, how are we going to? Go to war against EU.
42:17
F OS if we don't have all the the the the AI you know what I
42:20
mean yeah amazing. Where can people connect with
42:25
you online to learn more about you and your work?
42:28
Yes, go to Flu dot. Com we have a discord community
42:31
we're pretty active there. Keep you to join and ask
42:33
questions around you know we have Twitter accounts, fluid HQ
42:40
and we're also pretty active on agent to search for fluid FLOT
42:45
amazing. We have a lot of founders that
42:47
are listening to the show and probably a lot of soon to be
42:50
founders and getting accepted to YC is a major milestone for a
42:55
successful product. You actually crossed that
42:58
threshold. What's a step by step process
43:01
people can do to ensure a better chance at getting accepted at
43:05
YC? Yeah, I think.
43:08
Well, anything that one YC interview so far, I do think
43:12
that what my impression was that they care a lot about your
43:16
conviction. And I, I think you have to be
43:19
actually really confident in what you're doing and knowing
43:23
that you can succeed. I, I know some people try to
43:26
find some random ideas and with the hope of just getting into IC
43:29
and then figure something else out.
43:31
I think there's some, maybe you can succeed in doing that, But
43:34
like, I feel like it's pretty, it's going to make it harder.
43:36
Like rather find something that you, you, you truly believe in
43:39
that you want to spend time on. And, you know, I think that that
43:43
does show when you're interviewing with Weiss.
43:47
Amazing. Huge.
43:48
You're the man. Congratulations and and here's
43:52
to your continued success. Thank you so much.
43:54
Thanks. For having me.
43:57
Hey founders, let me ask. You this.
43:59
Are you building with AI or getting left behind by it?
44:03
If you're serious about scaling smarter, faster and bolder, you
44:07
need to check out AI for founders.co.
44:11
It's your go to resource for cutting through the noise and
44:14
actually using AI to grow your start up.
44:17
Each week, AI for Founders drops a sharp, no fluff newsletter,
44:22
dives deep with expert guests on the podcast, and hosts monthly
44:26
workshops where you can get your hands dirty with the real tools
44:29
and strategies. Whether you're precede or Series
44:33
B and beyond, launching something new or just staying
44:36
sharp, AI for Founders gives you the edge you need to win in 2025
44:41
and beyond. Hit AI for founders.co to
44:45
subscribe to the newsletter, tune into the podcast, and grab
44:48
your spot in the next workshop, and it's all free.
44:52
Build smarter, move faster, think bolder with AI for
44:56
Founders. Now back to the show.