Latest / 1 to 100 / E45: Swiping Left on Ads Like Tinder with John Gargiulo

Transcript

E45: Swiping Left on Ads Like Tinder with John Gargiulo · 1 to 100

0:00
So we have an API that you could create an MCP based on.
0:04
We are going to have like a better, yeah, like just
0:08
out-of-the-box MCP. That you can like so awesome.
0:10
Work, yeah. Yeah, I think we're everyone's
0:13
thinking about that right now, including Airpost.
0:15
Yeah, actually, John, maybe that's an awesome place to start
0:18
because I've been obsessed personally with this for the
0:22
past four days. First of all, great to meet you.
0:26
We don't know each other, but you too.
0:30
I love, I love just like sort of diving, diving right in just to
0:35
make sure I'm calibrated. How would you describe what Air
0:40
OPS does? Air OPS is an awesome company.
0:42
That's not us, but Air Post. Air post I'm sorry.
0:49
Yeah, no, don't worry. No, it's a failure of.
0:52
All right. No, no, no, no.
0:53
I'm I'm going to redo that, OK, just to make sure I'm
0:56
calibrated. Just to make sure I'm
0:58
calibrated. Tell me what Airpost does in
1:00
your own works. Yes.
1:02
So the goal of Airpost is to arm enterprise advertisers spending
1:08
a million or more a month on paid social with fresh ads every
1:13
week, mainly video ads. And I think what's different
1:17
about us is it's not an AI ad making tool.
1:20
It is people in the machine. So it is creative strategists
1:25
led by Christelle Selinger, who led creative strategy at
1:28
Narrative for a long time with Iron Man suits on that are doing
1:31
90% of the work internally at Air Post.
1:34
And for the clients, we're really busy.
1:36
We work with Calm, Doctor Squatch, reading.com, they're
1:39
just getting a slack every week. Hey, your ads are ready, opening
1:42
up the platform. They can tweak clips, they can
1:45
tweak copy. OK, so.
1:47
Upload and go. Sorry, they can open up Slack.
1:51
You said they're interacting with you on Slack.
1:54
How are they interacting with you?
1:55
A. Little bit yeah, I mean it's
1:58
mostly the their Rep says your ads are live.
2:01
They get new ads every week. And and I think the the key
2:04
point here is these ads are not auto generated.
2:06
They're not even majority AI footage.
2:09
We use their footage. We have our own library we think
2:12
is the biggest in the world of UGC vertical video footage that
2:14
we built up at our agency ready set.
2:17
And then we have a team that makes AI Jenny I footage
2:19
specifically for them to fill gaps.
2:21
And the engine orchestrates ads. The biggest piece is actually
2:26
the learning loop. So we have analytics.
2:27
We're like tapped into what's happening with the ads.
2:30
This whole taxonomy. It's a whole longer story.
2:32
But what we are building the vision is a whole the full loop.
2:36
Like, you know, there's a lot of companies that do the analytics
2:39
piece, but not the like, oh, you really?
2:41
I had a client tell me the big fintech in Europe.
2:43
Yeah, our motion AI tagging is cool and it tells me her words
2:47
like we need more people with brown shoes in the ads.
2:49
Like what am I going to do with that, right.
2:51
But if you could actually make the ads that make same day, I
2:54
think that's where it gets. OK.
2:55
Wait, so we think Europeans like brown shoes.
2:57
Is that is that like what's happening?
2:59
Yeah. Fairly.
3:01
Wait, was it was it a shoe? Was it a shoe company?
3:03
Or is it just like, were you saying that facetiously as if
3:05
like, oh, like this is what they've come up with?
3:07
Like it's like a it's an uncalibrated AI insight is what
3:11
you're. Saying, well, that's a good
3:13
point, That's a good word. Like unstructured, uncalibrated.
3:15
It's like there's just so much data out there.
3:17
It's been there for years. Big Mob had a thing where it's
3:19
like, oh, you ought to do more of this.
3:21
Or like when you see blue in the ad, but you know, a does, is it
3:26
BS or not? And even if it's not, what can
3:29
people do with that insight? Creative is really hard to
3:31
produce. So we're trying to build both
3:33
sides. There's this big thing right
3:37
now, I think in like let's call it SAS land for lack of a better
3:41
framing of like how the workflow actually happens and who's doing
3:45
what. So in, in your system for the
3:50
workflow, how is the customer interacting with your, let's
3:54
call it a service for simplicity.
3:56
How's, how's the customer interacting with your service?
3:58
Is it through Slack? Like what are they doing today?
4:01
There is a platform they're logging into air post, the
4:04
client portal and there are three taps.
4:08
There's the brief, which is like a living brief.
4:10
That's like the value props that you know, ideal customer
4:13
profiles, things like that. It's a living brief.
4:15
If they adjust it, it adjusts the ads they're getting.
4:18
There's an assets tab, that's the three baskets of assets,
4:21
their footage, 100,300 thousand clips we own and AI footage that
4:26
the engine's working with to make the soup.
4:30
And then there's the ads tab. And then once a week, 1020,
4:34
sometimes 40 new net, new Andromeda friendly ads appear.
4:39
And then they can go into those ads and they just do the two
4:41
things that 90% of people want to do, edit the words that are
4:44
being said on the screen or in the voiceover or edit the clips.
4:48
I don't like the way that guy sits on the couch do.
4:50
You think they want to log in? And then they're off and running
4:53
from there. That's a good question.
4:55
So I'll be honest, we have clients who do log in and do
4:59
make a lot of tweaks and edits and say, oh, you know, can I do
5:01
add an emoji here? We have a lot of clients of shop
5:04
who are like, they don't log in, they don't log in for weeks.
5:08
There's like, yeah, whatever. I I go in to get the ads, I
5:12
upload them. They work like performance has
5:15
been awesome. We could talk about that.
5:17
That's the larger point of all this.
5:19
But you know, and we both know the question that that begs.
5:22
It's like, do we need these interfaces?
5:23
I don't. Think anyone wants to log in to
5:25
anything? That's my personal view.
5:28
I hate logging in. I can't even remember my
5:30
password half the time. Forget about the interface once
5:33
I log in. The act of logging in is
5:36
irritating. I don't know Jeremiah, Zach,
5:38
what you guys think, but like over this past week, I have
5:41
never. I mean, I think did you just say
5:44
that you did you just admit to not using?
5:46
A password, first of all, I do not.
5:48
Because I would be that would be.
5:49
That would be insane. I do not use a password manager.
5:53
I can't believe I'm saying that publicly.
5:56
Oh wow. That's.
5:58
Insane. That's insane.
6:02
Holy crap, that is the most insane thing I've ever heard.
6:04
So that's true. The second thing that I think
6:08
would be kind of surprising to you guys is I, I, I don't know
6:13
what is in the air, but over the last five days, I just don't
6:17
believe in, in, in. I have so little confidence in
6:21
any interface right now. So much so that for the first
6:25
time ever, I actually like downloaded cloud code both in VS
6:29
Code and in Terminal to interact with things because I got I've
6:33
got I'm just like, this is insane that I am confined by
6:37
what I can do based on someone's decision to put someone else's
6:42
decision about what should be in the interface.
6:43
That is an insane constraint. And I'm like realizing this in a
6:48
very visceral way this week. Yeah, I'm, yeah.
6:51
I'm curious what you guys think. I think I think that's really
6:54
interesting what you're describing because I think if
6:58
we're talking about like a if, if I need to review 2040 ads and
7:03
determine what I want to put into my ad account or not in
7:08
that context, I don't actually want to do that Slack or via
7:11
like a messaging app or something like that.
7:14
I need it to be a more visual layout.
7:15
But what you just described a shot in terms of like could you
7:18
just have? Your own interface do whatever
7:20
you want it to. Do in that context.
7:21
That becomes interesting, but I still feel like you need some
7:24
sort of interface for that type of work outside of a chat
7:27
window. I think that the biggest truth
7:30
around all of these conversations is that like our
7:34
opinion, meaning the people on this call is so muddied by our
7:40
like past of building software in which now we have opinions
7:44
and taste about some of these things.
7:46
And the expectations that I think the average person is
7:50
going to have opinions and tastes about these things is so
7:52
much less than ourselves. I think that could that be
7:56
something that humans adapt to over time as they start using
8:00
more interfaces. Like for me, like the most
8:02
beautiful thing about Claude code and all the is that I
8:06
actually can now make more interfaces for things that I
8:08
want. It's not that I want less
8:10
interfaces. Like it's like I'm actually a
8:12
heavily visual person in which like for me to not have
8:16
interfaces for certain things and not have proper
8:18
visualization, I can't actually consume the information.
8:21
And so like I could see it evolving over time, But I, I, I
8:23
don't think that this is like a, a widespread thing that like
8:26
people don't want to log into software.
8:29
Like I, I, I think it's just like people want to log into
8:31
good software is like. I think there's, there's
8:34
something in there, there's a, there's a through line that is
8:37
truth here. I'm not sure if you guys saw
8:40
Nick Shackleford's tweet recently that went viral.
8:43
But like imagine on the topic of interfaces, imagine every time
8:47
you needed to check the time, you needed to ask AI what the
8:50
time was. It's actually way nicer to have
8:52
an interface that tells you the time so you can get the answer
8:55
without having to ask for it, right?
8:57
So you just look at your phone and you have the time, right?
9:00
I'm not sure if you saw this tweet but Nick Shackleford said
9:05
99.9% of you using AI, you know daily number crunching are
9:09
probably so fucked with your data.
9:13
There is some like Midwest on Reddit is saying that they've
9:17
relied on AI, That's their expectation from the AI to give
9:22
them their numbers and the numbers that they've been
9:24
operating off the information they've been using to educate
9:26
the decision making is all wrong.
9:29
And everyone in the conversation is saying the exact same thing,
9:31
right? They're saying yes, this is the
9:33
problem. AI is notoriously bad at math.
9:35
Catch up eventually. Such a bad take, right?
9:37
But like. The idea here is you're trying
9:39
to get the truth. You're trying to get to time to
9:41
maybe maybe right, But maybe also the the best way to get
9:45
time to insight isn't on the other side of having to re
9:48
prompt the same answer over and over again, right?
9:51
Like what is the time? I just want to see the time.
9:53
Oh, you don't want an interface that tells me what the time is.
9:55
I mean, I mean, maybe. Maybe you do, Jack.
9:58
Maybe you do, but you want to know the time for a reason.
10:02
Yeah, it's an extra step of friction otherwise.
10:03
Like, imagine I had to ask Claude, right?
10:06
Like imagine I had to ask Claude to build me a widget to put on
10:10
my IO on my home screen of my iPhone every time I wanted to
10:14
know the time. I don't want that, right?
10:16
I want it to tell me the time the second I look.
10:19
Although you ask it once and then it's always there like it's
10:22
not. Yeah, or I don't even have to
10:24
ask for it. It came on the phone.
10:26
I think that don't having to ask for it super interesting thing
10:28
on my mind right now because what we're thinking about is
10:30
like y'all probably played with gamma or these like really
10:33
agentic things where you're just talking to it and saying make me
10:36
a deck based on this call. I don't like slide 5.
10:39
Also the photos are a little weird and it's pretty magical
10:42
and also a little exhausting. And so as I think about, should
10:47
we agentify airpost instead of like a traditional sort of SAS
10:52
thing? You log in and you see stuff and
10:53
you click buttons. Should you be able to go in and
10:56
just be like, make me an ad that hits this ICP, you know, based
11:01
on what's working XYZ and have a come up.
11:03
Then I started thinking this morning, why do I even want to
11:05
type that? Like I want something that knows
11:08
that that should be the thing I do and suggests it and suggests
11:10
it mean it already makes the ads.
11:12
It already tells the time 10 times on the screen.
11:15
And then to Zach's point, ads are visual.
11:17
I want to glance at them. I want to scrub them.
11:19
OK, that's the one I want. So it's like we're all talking
11:22
about the same thing. It's like where you move the the
11:25
curtain either like you make people do it all themselves,
11:28
which is cool and powerful if you're a person who wants to
11:31
type a bunch of things in, or you just identically make times
11:37
for them. So go look at it, yeah.
11:39
I totally agree that. Makes sense.
11:41
Yeah, I think it makes sense. I'm just not someone who wants
11:43
to reinvent the wheel. I don't want to reinvent the
11:45
wheel at all. Like if it already exists and
11:47
it's there and there's no friction to getting my answer,
11:49
I'd rather that than rebuilding. So, Jack.
11:52
Can I ask you a few questions? I, I, I promise I'm not debating
11:56
for the sake of it, but I, I, yeah, like, I have had such a
12:01
big mindset shift this week And, and I, I, I'm just curious how
12:04
you would think about this. When you go work out, what do
12:08
you want the clock to say? Like when you go, when you're,
12:10
when you're working out, like whatever, however you like.
12:13
Do you, do you do weight training?
12:14
Like what do you do? Do you run?
12:15
What do you? What do you?
12:17
Do I don't want to look at the clock?
12:19
Yeah, Yeah, I do like body weight workouts and I run OK.
12:23
Amazing. What do you want your time
12:25
keeping device to be telling you when you're doing that?
12:28
Nothing. I don't look at the time.
12:29
I'm sorry, this is probably not going.
12:32
I don't look at the time. You don't?
12:35
Oh, no. OK, OK, OK, All right, let me,
12:38
I'll just give my example, OK? Because I'm just going to
12:41
simplify this. So when I go workout, I'm trying
12:44
to do sets every 3 minutes, OK, to make sure that I'm resting
12:48
appropriately. So, and I and I know the global,
12:51
like I know the global time is, I want to get the workout done
12:53
in an hour. But like actually the most
12:55
relevant thing for me is like sets of three minutes.
12:57
OK, so wait, my body gets enough rust, yada, yada.
13:00
And I'm not trying to get hurt. So in that context, I don't want
13:04
to know the time. Like I want to know where I am
13:08
in the three minute interval. If I'm running, I want to know
13:11
my pace and I want to know how much I have run.
13:15
And so like I think Zach's point is the is the insight, which is
13:21
yes, we all need a visual interface, but I actually think
13:25
we each know in any given moment and any given context what that
13:29
interface needs to be for us in order to do the job.
13:33
And so like my provocation for like air post.
13:37
For example, see. See we we do.
13:40
That's where I'm saying we do like for example, one of the
13:44
most annoying things that I've been working on recently is I am
13:48
trying to rebuild our entire road map.
13:51
OK, I hate project management tools, which is why we seemingly
13:55
switch project management tools. We switch project management
13:57
frameworks like, and like I've been talking now with Claude Co
14:00
or Claude Cohort to like come up with our new project management
14:03
framework and like blah, blah, blah, blah, blah, blah.
14:05
But there's no still like really good project management
14:07
software. It's like, this is what I want
14:09
to accomplish in the next 90 days.
14:11
I know nothing about project management like practices and
14:15
things like that. Here are the resources that we
14:17
have. This is how great like I want to
14:19
get. Here's an example of how
14:20
granular I want this project to be broken down.
14:23
Do it like that still doesn't exist.
14:25
And so like, and I, but like I, I don't, I still don't know
14:29
exactly the fundamental structures around some of these
14:34
problems that I'm trying to solve.
14:36
And so therefore I actually don't know the interface yet.
14:38
And this is where I'm saying like there needs to be, this is
14:40
where interfaces come in. It's like when you aren't so far
14:45
down that path in which you have opinions about it, and I think a
14:48
lot of software users fall into that category is like, we're
14:52
not. I'm sure John has people on that
14:54
log into air post that like they haven't been obsessing about ads
14:58
for four to five years. Maybe they previously only ran
15:03
display at wherever they were working before and now they're
15:05
getting into social. It's like they actually don't
15:07
have that much of an opinion and they are looking for tools to
15:10
give them some opinion on the way that it should be doing
15:13
things. Yeah, it reminds me of the
15:15
question of thinking about this visual versus not visual.
15:17
Put your slide your feet into the shoes of like AVP of growth
15:20
at you know pure ology or you know, Doctor Squatch, whatever.
15:25
If you could type into a platform, if you could sign a
15:29
partner and say make me ads that work, you know, making the ads
15:34
that make me more money. And it's like, got you, it's
15:36
2027. Here's what I'm going to do to
15:38
make a bunch of ads. You're not going to see them.
15:40
Is legal cool with it? Oh, no, here's everything legal
15:42
says we can say and not saying. And also, please check the
15:44
latest FTC guidelines. Yeah, we got that.
15:46
Well, we won't put anything that doesn't do that.
15:48
And we're going to put cost caps on.
15:49
We're going to run 100 things this week and and you know,
15:51
we'll take care of it and you pay us 3% of the outcome and we
15:55
got you. And then it just goes next video
15:57
ads. Do you care?
16:00
I'm being a little extreme just to make the point, but it's
16:05
rush. Totally.
16:07
I think Zuck has said in earnings calls, 3 earnings calls
16:10
in a row that that's where he wants to take meta, right?
16:13
Or at least last year here. And the pushback you get from
16:16
the brand folks. And I'm a recovering brand guy.
16:18
I was in Airbnb for three years. I, you know, went to Miami high
16:21
School to be a creative and I love creative and I love brand.
16:24
And do you think it's important is I think a lot of a lot of
16:29
wishful thinking to make it go away, like, oh, wait, what do I
16:32
do now? Etcetera, which is
16:33
understandable. And I think everyone's going to
16:35
be just fine in the new world. You know, and it's funny, the
16:39
vision for air post few years ago was Acmo with a coffee in
16:44
her hand, walking to the Manhattan to work, swiping left
16:47
and right, like Tinder swiping. Yeah, that looks good.
16:49
That ad. No, yes, no, yes, no.
16:51
And I would even advance it now and say, and that's just
16:54
training. And then eventually she doesn't
16:55
need to watch the ads anymore. And eventually we're all
16:58
laughing about this podcast in five years because like, every
17:01
company's running hundreds, thousands of ads, they're all
17:04
finding their niche Icps. And, you know, creative is the
17:06
targeting. And of course, no one's watching
17:08
them all. And of course, no one's
17:09
prompting anything. Ha ha, ha.
17:11
Like, I do think that's where we're headed.
17:13
Totally. I like that, you know, it's, it
17:16
reminds me of agents, you know, operating agentically, having
17:19
agency, right? It's bypass permissions in core
17:22
code. It's I trust you.
17:24
Go do it, right? It's, it's, it's delegation.
17:27
It's the reason that you hire people.
17:29
Right, and I think, Jack, that tweet you showed is the brake
17:32
pumping that some of the scaredy cats are doing where I thought
17:35
scaredy cats just like, I just sigh when I see that like Rashad
17:38
did, because it's like, you know, I you, you did put the
17:40
caveat in that people in that thread are like, well, it's, we
17:43
know it's going to get better. I don't think people, I think
17:45
people, it'll actually say that. I think they're feel they're
17:48
emotionally pumping brakes like cartoon character, like with
17:51
their feet in the sand going forward.
17:54
But come on, like, we're not even talking about
17:56
hallucinations fit as much as we were a year ago.
18:00
All the things they're complaining about are going to
18:01
go away in what, maybe six months, Two years?
18:04
Who cares? Like I think it's much more.
18:05
Easy. They're already gone.
18:06
Actually, John, that's the crazy thing.
18:09
The hallucinations people are talking about are already gone.
18:13
OK, this is a user error. And if 100 people utilize their
18:18
sales that are like that guy, probably 95 of them would have
18:22
had no problems. It's just a viral tweet, you
18:24
know what I mean? The big issue though, ultimately
18:28
the big issue, I think it comes down to an emotional reaction
18:32
that many of us have, myself included in that like, OK, do I
18:38
actually want a world where AI is this capable?
18:42
Because ultimately then you like what you're talking about the,
18:45
the brand person, right? The problem there isn't that the
18:48
AI can't do it. It's that you want to have
18:50
control and you want to put your own stamp on something.
18:54
And if AI takes that away from you, what does that mean for us
18:58
as humans? And I think that's like, that's
18:59
the question, right? And I think this is the thing
19:01
that I'm trying to work on with. I mean, I'm, I don't know what
19:05
it's what it's like for you all, but a lot of the people who work
19:07
for me are scared of AI. And I myself again, included, I,
19:11
I kind of fall into like having some capacity.
19:13
I look at like what the future could be and I actually don't
19:15
love the idea of that future, but I can't stop it from coming.
19:19
So what does it look like to actually fully embrace being the
19:23
most human version of myself that I can be and enabling that
19:28
for the people who work for me? So that's how I've been kind of
19:30
trying to think about it. And I actually don't know the
19:32
answer to that, but I think that's, I know we've talked
19:34
about taste and some of those kinds of things.
19:36
And so where like where do we as humans fit into that world?
19:39
And I think that's kind of the big thing that scares people.
19:42
And that's why we see the the pumping of brakes like you're
19:45
talking about. It's it's that a lot of us don't
19:47
want that future. So.
19:49
I think, I think there's two, there's, there's two types.
19:52
There's two types of pumping the brakes and there's two avatars
19:55
of it. One is the thing that I'm doing
19:58
right now that is paying for the food that I eat tonight and the,
20:01
and the roof over my head is going to go away.
20:03
And so therefore I'm pumping the brakes.
20:05
The other of all I'm pumping the brakes is where I fit and it's
20:08
that like. The what you realize when you
20:11
start using AI tools a lot is that you realize that actually
20:16
all of your work becomes important work and you no longer
20:19
have the escapism of bullshit work.
20:22
And so a lot of us cope with things like our e-mail inbox and
20:27
a document that our lawyers sent, and we pretend like we're
20:31
doing work, but we're actually taking a break.
20:33
The truth is, is that when you actually adopt AI tools, 100% of
20:37
your time is really shrewd decision making and it's makes
20:41
your day much harder. You're like, I, I try like it is
20:45
an exhausting reality of the world and like, it's actually
20:48
not the fact that we're going to go away.
20:50
And there was this study done about companies that had full
20:53
adoption of AI across the entire workforce and hours worked and
20:58
intensity of work went up, not down on the companies that that.
21:02
So it's like, that's where I'm like fucked.
21:04
Like I don't know if I want to work this fucking hard.
21:07
No, because the problem you want to know what the problem is.
21:09
You want to know the problem is, is because you become the
21:12
bottleneck and and what's it never used to be that.
21:16
I mean, it used to be that way if you're like the CEO or
21:19
founder, because everybody's asking you for shit all the
21:21
time, but everyone is now the bottleneck to their agents.
21:27
OK, I have like explicitly told our company, if you don't have
21:32
workflows where agents are asking you to do things because
21:36
they don't currently have the tool set to do it themselves,
21:40
you're doing it incorrectly, right?
21:44
Like if you're still like a you doing stuff like that is the
21:47
wrong workflow. Like the workflow is you're
21:51
being told like like outcomes, an action list and the agent
21:56
tries really hard to do it themselves.
21:57
And then there's a set of things that they simply cannot do, and
22:00
you're doing it and and now you're a bottleneck, which is
22:02
why it's so exhausting. It's so exhausting.
22:05
And it's funny, it's so true. And I think Sam Allman, the
22:09
Denise, what's the name said this recently.
22:11
It's like, it's not that the technology won't be there.
22:13
It's not that diseases won't be solved in maybe two or three
22:15
years technically. But the people he went through
22:18
the whole timeline of like the COVID vaccine took like a year
22:20
and a half to go through all the truck.
22:22
Like people that I talk to at businesses.
22:24
By the way, guys, when I go to like a New York ad week or like
22:28
a conference of my old traditional advertising
22:32
industry, no one's talking about any of these things.
22:34
No one has definitely tried quant code or Co worker knows
22:37
what it is yet and they're quite terrified.
22:39
I think. Jeremiah, your point and thank
22:41
you for sharing. That is extremely spot on.
22:44
Like it's like when you someone's being argumentative or
22:46
something, you're like, hey, what's what's really going on
22:47
for you? Like my most vulnerable fancy
22:50
senior friends from traditional advertising are like, I'm scared
22:53
and I'm pissed and I don't want that world to happen.
22:56
So that's why I'm saying this is not good enough.
22:58
This is not good enough. So I think it's a really good
22:59
point. You just think like something
23:02
has tipped in the last, I don't know, over since it started this
23:07
year basically something has tipped like I, I really feel it.
23:13
I'll give you like a very simple example.
23:15
So we just, we were talking about MCPS earlier.
23:18
So we just launched our MCP 2 two days ago.
23:22
It first of all, it took an afternoon to deploy OK, like
23:27
almost no time. And I'm already using clod to
23:34
interact with our system and start up our own UI and it's
23:39
better. OK.
23:41
Which is like, do I like seeing that clod is better than our UI?
23:47
I mean, I don't love it, but it is better and it's like it's not
23:52
easy to make good agents. What is the like the value of
23:56
our system, thank God, is unrelated to the UI, right?
24:02
It has more to do with like generation and and data.
24:05
It's not particularly relevant to the conversation, but like
24:09
the fact that I can use it in flawed and then ask questions
24:12
like, OK, given this and then now pull in all this data from
24:18
all of these other places and go make decisions.
24:21
And then by the way, now I instruct it to go make an to go
24:24
do the actions back inside the Fermat product, all without ever
24:30
going into our product is just better and I work here.
24:35
It's the same for us by the way. Yeah, I I just this just
24:38
happened the last couple days too for me yourself, which is to
24:41
be a creative strategist working behind that curtain in the back
24:44
end of air post to make these ads is really you got to be
24:47
trained. There was an 80 minute call with
24:50
our head of operations describing to someone new.
24:52
Hey, there's this this fits this janky.
24:54
You got to press that button, wait for it to spin.
24:56
I gave it all to Co work and I'm very excitedly last night sent
25:00
my team like guys, it made ads like and they're good.
25:04
I remember what I was going to say earlier about the being
25:08
exhausting Zach, like I did feel exhausted the last couple of
25:11
days I was describing to my friends.
25:12
So I as I have them running now, I have like 6 or 10 things in Co
25:16
work going all the time, building all the time, and they
25:20
keep getting stuck in bottlenecked and usually I'm the
25:22
bottleneck. I just had to figure out things
25:24
I didn't think I was capable of because I'm not a technical,
25:26
technical person. Like it was like, Oh, if you
25:28
give me an opening eye key, I can go make that video thing in
25:30
Sora and I'm like, how do you do that?
25:32
It's like just go to this links and click make key.
25:34
I did it all, but it's like I give the analogy.
25:36
It's like I'm cooking all these stews like these like slow
25:39
cooker things around my kitchen and I'm running around and this
25:42
one I started to chop and I totally like failed and there's
25:44
just some onions sitting there. These ones are going and
25:47
sometimes it takes me all day to like bake all these things.
25:50
And it the, the, the, the context windows get insanely
25:53
long. But by the end of the day, I've
25:57
been able to make like 8 or 10 stews by myself.
26:01
And it's, it's exhausting, yes. But also I feel like incredibly
26:05
powerful. I won't take the time, but I
26:07
could give you guys so many examples of things that like I
26:08
would need a whole team to do that I'm able to just have.
26:11
And I encourage everyone to have as many Co work windows going as
26:14
possible. And my job now is just dancing
26:16
around making sure nobody got stuck.
26:19
Co work yesterday saved me like 65 to $75,000 from Deloitte.
26:25
So we are catering company. We do shred which is a tax
26:30
credit type thing. Last year our tax credit I think
26:33
was like $600,000 and we paid Deloitte 12 1/2% to do our
26:37
thing. They messaged me and it's the
26:41
worst process ever. The Deloitte consultants that we
26:43
get, because we're not that big of a company are ESL horrific to
26:48
talk to. They don't understand our
26:50
product whatsoever. And so you need to layman's term
26:53
explain it to them so that they they can layman's term explain
26:56
it to the CRA. And we have three years now of
27:00
shred claims. And so I uploaded to cowork all
27:03
three years of our shred claims. All of our shred claims this
27:05
year are continuation of projects.
27:08
I gave it, I exported our, our task database from Notion and it
27:14
made the most, it looks like it makes Deloitte's previous claims
27:19
for our shred look like a 2 year old made them.
27:22
It's, it is incredible. And I don't have to pay the 12
27:27
1/2 percent. I'm like what's going to be
27:28
$700,000 in tax credit And it's like it, it took less than half
27:33
a day to, to do all of that. And it's like, I, I, I think I'm
27:39
like, and that's, that was actually the biggest tipping
27:42
point for me. I was just like, oh wow, like
27:44
this, just this was so much better than a well paid
27:47
consultant. It's like so much better.
27:49
Dude, you know what? You guys want to know what?
27:50
What we just did. I can't believe I'm going to say
27:53
this out loud, OK? You want to know what we just
27:55
did? We took because we're going
27:58
after large enterprise. We said, here's our target list.
28:04
OK, Then we had a bunch of bots go figure out who are the
28:08
relevant personas in each of those.
28:11
Then against those relevant personas, we asked to go do
28:14
research against the types of things that we care about,
28:17
whether they say publicly or not.
28:19
Then against every one of our marketing streams, we asked
28:23
which marketing stream is relevant for which person inside
28:26
of each of our target accounts. OK, Like literally what is going
28:29
to happen is at the person level across all sorts of different
28:37
ways that you can engage a person, we're going to know what
28:41
to what is the right way to engage them.
28:44
And all of this is going to be at 0 effort.
28:48
To us, it's just a bunch of agents doing a bunch work.
28:52
And guess what? It's basically, I mean, not only
28:56
is it basically correct because you can like just spot check,
28:58
like you can go to these people's LinkedIn and whatever,
29:00
whatever, whatever. Dude, the insights that it pulls
29:04
out about what people care about is ridiculous.
29:08
Ridiculous. Yeah.
29:10
How do you guys, how does this affect the shop and everybody
29:13
how you hire? So I find it because look, this
29:17
is if we were having this thing four weeks ago, we may not have
29:19
a lot of these insights and Co work and all this stuff or two
29:22
months ago, right? So I have empathy for people and
29:25
if I'm interviewing someone, I just had an interview an hour
29:27
ago and they were like, yeah, oh, no, I use ChatGPT and it
29:31
helps with the project management of this gaming
29:32
company. And I mock up a couple things
29:34
and basically it's people who you just say ChatGPT and then
29:38
they're still chatting with it. They're not doing any of these
29:40
things we're talking about. And the dream is that everyone
29:42
you hire is like, you know, I picture a running race, like
29:45
they're running ahead of you going down a path and you're
29:47
like, cool, let me catch up. Not like no, come on.
29:50
Have you not tried Co work yet? So I don't know.
29:52
How has this affected the way you guys source and hire people?
29:55
We haven't hired anybody for a year.
29:59
I was going to say, yeah, when people leave, we're hardly ever
30:02
backfilling positions at this point.
30:03
And. So the truth is like, and I also
30:07
don't like if you were to say, OK, someone that's like like a
30:10
clawed weapon, I'm like, well, how did you get to a, how did
30:15
you get to be a clawed weapon and want a job?
30:19
Because like in a weird way, it's like, I think anybody with
30:23
quad could just place any sort of income like really quickly,
30:27
like very quickly and like not have to have me tell them what
30:32
to do. So I think like it's really just
30:36
like you. I think it's more of like, you
30:37
want someone that's like. I mean, not a Luddite.
30:40
But. I think yeah, yeah, this is
30:43
going to be provocative. And I know there are people at
30:46
my company who listen to this, so I have to be careful how I
30:50
say this. But the, I think the end state
30:54
reality is either you're one of the people who embraces it or
30:59
you're not. And I think the entire tech
31:01
industry is only going to be filled with the former.
31:05
And we can sit here and debate over what period of time.
31:09
I'm not sure how relevant that debate is.
31:11
Like is it by the end of this year or is it in two years?
31:14
And then if you say that that's true for the tech industry, you
31:17
can say that that's going to be true broadly for anyone in
31:21
couldn't quote IT functions at other industries, right?
31:25
And so then we're talking about like a rate of propagation.
31:29
Five years feels like an incredibly long amount of time.
31:32
Like it's for sure not 5 and like 6 months is too short
31:36
because like, like humans just can't, I mean, like we're just
31:39
not going to change that fast. So it's like some some number in
31:41
between those two at our company.
31:45
I mean, what I am literally telling people is I'm just the
31:49
messenger. Like this is happening whether
31:53
you like it or not. And I'm trying my best to be a
31:55
good messenger and tell you what the expectation is going to be.
32:00
And you and I can disagree at what period of time that that
32:03
expectation will change for everyone.
32:05
But given that I'm the messenger and given that I'm giving you
32:08
the opportunity to make yourself in Camp 1, you should take it
32:14
because if you don't, it's going to be much worse.
32:17
And that's like the best way that I can push people to where
32:20
they need to get to. It's because unless they
32:24
understand the inevitability, which is what I'm trying to
32:26
characterize for them, it's like, guys, this is not me
32:28
asking. This is the market changing what
32:31
it's expecting. And I'm simply here to
32:33
articulate to you that that's true.
32:36
And the the technology doesn't even have to get better for that
32:40
to happen from here like we are. We already have the technology
32:43
to replace, I don't know, 80% of the work that's happening
32:46
manually in our, at least in my business.
32:49
I don't know if that's the case for you all as well, but that's
32:52
and maybe more than that, but that already exists today.
32:54
It's actually just about inserting into processes,
32:57
getting at the right data, training it to do the tasks the
33:00
way you need them done, all of that kind of stuff.
33:02
So yeah, I'm, it's, yeah, it doesn't have to get better.
33:05
And presumably it will get better.
33:07
I would dovetail one thing on that though, Jeremiah, which is
33:10
it's also AI can get you to 80% of stuff.
33:12
It also gets you 80% of the way there really fast.
33:16
But that last I call, the last five miles of AI is just really
33:20
hard. Like I was even talking a big
33:22
game about gamma. Yeah, I've used gamma a few
33:24
times. I've put some call on this.
33:25
I was really impressed. I actually haven't used it in a
33:27
few weeks. I don't because the last miles,
33:30
like it's annoying, right? So that's what that's what we're
33:33
obsessed with at Airpose. That's why we proudly have
33:35
humans in the machine is I think things are getting more
33:39
automatic. Like I was blown away the last
33:41
couple days what Co worker could do in one shot really deep
33:45
complex 3 hour things. But mostly most stuffs not all
33:48
the way there yet. I think the really scary
33:51
exciting binary moment is going to come and tip.
33:54
And I just got AI just had a call byte dance this morning.
33:57
They were showing me what C dance can do.
33:59
Like there's going to be a tipping point and it keeps
34:02
feeling like it's happening, but it's going to happen in the next
34:04
year where whoosh. Like it does get to 100% in a
34:09
lot of use cases from a lot of companies and everything starts
34:12
changing really fast. I think the, I think the one
34:18
thing to think about for everybody though, is that like
34:22
what we're essentially doing is we're getting a bunch of really,
34:24
really good violin players, a couple cellists.
34:29
And at the end of the day, like each one of those individual
34:32
like agents or whatever they can do, it's like, yeah, they can
34:35
bust on the corner and make like 20 to $30 a day.
34:39
The real value is like to put together an orchestra and like
34:42
put on the Symphony and sell tickets.
34:44
And so like the the the truth that like the last 10% is the
34:48
hardest part has been true before AI, It will be true after
34:52
AI. The difference between greatness
34:54
and good is like you don't stop at 100%.
34:56
You actually go 10% less or more than 100%.
34:59
And it's totally, it's always, always is like small little
35:02
differences. The ingredients are all the
35:03
same, but like the way that you put them into the market to then
35:08
extract value. Yeah, like the right person can
35:11
take a bunch of St. musicians and make hundreds of thousands
35:15
of dollars in a show, but each one individually can't.
35:18
So it's like really thinking, I think it's like you think
35:21
yourself, OK, I'm going to, I'm going to create an orchestra.
35:23
How do I actually turn these things into things that fill
35:26
seats and things like that is like, I think the way to think
35:28
about. It and that's the difference
35:29
between the one guy who just makes an income messing around
35:32
and claw doing one thing really well or any of our companies,
35:35
frankly, doing one violin really well and and an orchestra.
35:39
So like what we're talking about now is like, OK, we don't just
35:43
make content. We also like have this complex
35:46
taxonomy that goes really deep and wide against angles, ISP S
35:49
blah, blah and understands what's working for real and then
35:51
actually make something based on that's going to be the company,
35:55
right? Oh, you know, it's not just meta
35:58
by next year. We're also doing TikTok and snap
36:00
and what's going on CTV with vibe.
36:01
But if we don't build that orchestra, we'll just get
36:04
drowned out by some kid who builds an air post in a year.
36:07
Yeah. I mean, I think like it's funny
36:09
that you said like the framing of the orchestra.
36:15
Some of the people on my team have gotten so good at this that
36:19
they just produce things at like this ridiculous rate.
36:23
And some of the things that they produce are are not that useful,
36:27
right, which is like a not a surprising thing to say, right.
36:30
And so it's like, how much do I2 hard questions.
36:34
Once people get really good at this at the production side 1 is
36:39
how much do I even want to discourage that?
36:40
Because like if you're batting 300 but you're swinging a shit
36:44
load, what the fuck do I care? Honestly, right, Like just keep
36:49
swing gay like you, you go to town.
36:51
The other side of it is you have this like human problem where
36:55
other people perceive you doing work that's not valuable.
37:00
And so you have to like think about how do I frame how this
37:02
person operates to the rest of the company and, or how much
37:06
feedback do I give them to just like take a second to think
37:09
about whether or not like how do they approach the problem
37:13
statement and framing Like what is actually going to matter for
37:15
people's? That way they like put their
37:17
effort into the right thing. And like, even if it doesn't
37:21
work, which like, I candidly don't care if it doesn't work,
37:24
there's a way to articulate why they've worked on the problem in
37:26
the 1st place, right? If for no other reason than to
37:29
like manage everybody else's psychology.
37:33
And I think that's actually becoming way harder because we
37:36
have like 30% of the company now operating this way.
37:42
I want to say I think like we're probably going to be able to hit
37:45
70 by end of April based on like some of the actions we're taking
37:50
internally. But do those 30%, the rate at
37:53
which they produce things is insane.
37:56
Like it is, there is zero question the difference in
38:01
output between that like them today versus them three months
38:06
ago. It's like a 5X difference.
38:09
Tell me what you guys think about this.
38:10
I think what we're seeing is a sort of increase in the velocity
38:13
of execution and and software businesses, right, where we've
38:17
already seen this in different markets and did you see it was
38:19
drop shipping, it was the availability of products
38:21
overseas. So every you know so many
38:23
brands, so many brands out of the world work selling the exact
38:27
same products. We have this velocity of the
38:29
ability to execute, but sometimes really poor direction
38:33
doesn't matter that you can execute, you're executing in the
38:34
wrong direction. It's not something that anyone
38:36
wants to buy. I think that's maybe what we're
38:38
seeing. You know, I'm not sure if you
38:40
guys have heard of 1080 ten, the idea that you have 10% at the
38:43
beginning, which is direction, 80% which is leverage performed
38:46
by humans or Asians and then 10% at the end, which is evaluation,
38:50
a review. That kind of decision of the 2nd
38:54
order effects cybernetics, right?
38:57
You're looking at what's happened and you're deciding
38:58
where to go next direction. Again, maybe that just becomes,
39:03
I'd be curious to hear what you guys think. 198 one.
39:06
And the most important thing like Zach was saying the other
39:09
week is the product manager, the person who can make bets, who
39:13
can make executive decisions and place entrepreneurial judgment
39:19
in the right direction. I think the leverage is not go
39:22
Fusion is that? It's a really interesting model.
39:24
I think it is going to go to 98. I think the one will be really
39:27
critical. I'm not worried for humans.
39:29
I think that directing where to point the rocket before you put
39:33
the AI rocket fuel light is going to be more important than
39:35
ever. I just had a call before this
39:38
with a design leader at Amazon who was saying like, you know,
39:42
taste making is the most important thing now.
39:45
It was actually interesting. When he talks with Claude Code,
39:47
he's like, hey, look at the top ten agencies and frog and
39:49
whatever and like, that's your bar and gives a lot of inputs.
39:54
Yeah, I think that's a really interesting model.
39:56
I think the 80s is just going to grow and grow and.
39:58
Grow I I do think like at first when you say you're not worried
40:01
for I am worried for a very specific type of human and that
40:05
is white collar worker that isn't that is.
40:10
Between the ages of 25 and 35, that isn't technically that
40:15
curious. The younger than 25, you could.
40:19
There's enough sort of like plasticity in the brain that
40:23
this will just be the way that they're doing it.
40:24
They're going to be learning your jobs in this way.
40:27
But 25 to like, yeah, 3540, maybe they're stuck a little
40:31
bit. They have a lot of life left.
40:34
They have a lot of wealth to accumulate.
40:35
There's going to be a few of those cohorts that just get
40:39
absolutely rinsed. And it will be interesting to
40:43
see the economic impact of those people not, I think,
40:47
contributing to society and therefore having a payment for
40:50
their contribution to society. Can I tell can I tell you the
40:53
easiest way to see this issue? It's you look at the market cap
40:59
of Anthropic and the number of people who work there or Open AI
41:04
and the number of people who work there.
41:06
And you look at the market cap of every other company and the
41:08
number of people who work there. And and it's like the most sort
41:12
of visceral way to fully understand that the total number
41:17
of people who will actually use these tools well in a way in
41:22
which is at the leverage point that is going to be expected by
41:25
the market is a fraction of the current workforce in technology.
41:30
I want to be clear in technology, I'm not worried for
41:34
the like, I'm not worried because humans are really good
41:37
at figuring out what to do. But I think it is under.
41:41
Technology, it's not, it's not even as bad because those
41:43
people, the, the technology market's going to grow so much.
41:47
I think the issue is like totally white collar worker.
41:50
I understand, I understand. I understand you notice
41:53
something like not in technology like that is like way worse.
41:57
Yes, yes, I guess what I'm trying to, yes, I agree with
42:00
that. And what I'm sharing is I don't
42:04
think you need to speculate a lot because you can look at the
42:07
revenue, you can look at the market cap and you can look at
42:09
the number of people at companies like Entropic, Open
42:13
AI, Gamma, whatever it is. And then you can look at those
42:16
same metrics at quote, UN quote traditional companies.
42:20
And that is telling you what the dis UN quote display like, you
42:27
know, short term displacement could look like.
42:30
And literally every I mean, John, I don't know what your
42:34
hiring plans look like, but Zach and Jeremiah literally just said
42:37
they haven't hired anyone. Amazon, Shopify, Microsoft
42:41
publicly or show me their headcount looks like this,
42:45
right? Yeah, every company that is very
42:48
largest saying that their headcount is going to do this.
42:51
And so then it'd be, yeah, it begets a non trivial question
42:56
right now. Again, I think humans are really
42:58
good at figuring out how to add value.
43:00
So I'm not stress free humans per SE, but I think for any
43:03
given company I think it's a non trivial thing.
43:07
I'm not worried about humans. I'm not worried about humans.
43:09
I think that if this is just going to create so much surplus,
43:13
so many humans are going to be able to enjoy doing exactly what
43:16
they want to do in the day, being relatively unchallenged.
43:21
That's not where I want to be, but that's, I think where a lot
43:24
of people are going to find themselves.
43:25
We have this miracle in the United States and England and
43:28
Canada and Australia where people will sit around all day
43:32
unemployed and not die, right? They're totally unchallenged and
43:37
they're actually enjoying life, right?
43:39
They're watching Netflix. They're eating, by all
43:42
geographical and historical standards, good food.
43:45
France probably too. Yeah, the whole of Europe,
43:47
right. And I think that's uneven
43:49
accelerate. It is interesting though, like I
43:53
don't know, I mean, do you, do you all think there's a world in
43:56
which it's like we have like a 10% employment rate and it's
44:01
basically like you've got everybody?
44:03
You don't think so? No, I think this normalizes over
44:05
time. There's actually two other like
44:07
things just like 1, like the consumer debt service in
44:10
America. Like I'm not saying that people
44:13
won't be happy to sit around and do nothing, but like if they
44:16
lose their car, I think they would or something.
44:19
And just if you just look at where that's at in America right
44:24
now, it's actually, it doesn't really matter like long term
44:26
what happens. Like actually small faults can
44:28
create like really big problems specifically in that.
44:31
And so I actually think that long term, yeah, not worry about
44:36
humans, we'll figure out a way to create value and like
44:38
prosper. But like specifically like in
44:41
America and like other Western countries that have like a lot
44:45
of consumer debt, I think that right now like any sort of like
44:50
pull back in employment because of changes in market like
44:53
actually can make like a really big impact on people in a
44:56
negative way. Can I circle back to a SAS thing
44:59
which by the way I want to make a prediction.
45:01
I bet this will not be called SAS operators a period of time
45:06
we call agent, but business model.
45:10
So for those of you that charge by seat or it's not really
45:15
directly outcome based or usage based.
45:18
If the world goes agentic, if it's not just you know few of
45:21
us, but it's millions of people in a year and a half who's or
45:25
using their agents to go do stuff in our products and
45:28
they're not never looking at the interfaces.
45:29
They don't need more than one agent to do it all.
45:32
Has that gotten you guys thinking about how you?
45:34
Charge totally. I mean I made a public comment
45:36
on it and got in trouble for it I think.
45:40
I think per seat is really hard to continue to charge for
45:43
because every customer will demand an MCP at some point in
45:47
time, at which point the per seat thing by definition is dead
45:51
right. And I think like the, yeah, I
45:57
think if you're not associated to usage or outcomes, it's going
46:00
to be really hard for someone to rationalize how to think about
46:03
you because they're going to think about everything else in
46:07
usage and outcomes. And so you have to make your way
46:11
to usage and outcomes and speed with which you make your way
46:15
there. I don't know.
46:17
But John, I mean, you like your customers must, must be like
46:23
asking for this, right? Like there's, there's no greater
46:26
thing that someone wants to pay on a usage basis than us, right?
46:31
Like I want to pay for the ones that I use.
46:34
They we charge, we charge a minimum three, we most of our
46:38
clients, it's a percent of spend.
46:39
So it's like they do well, we do well.
46:41
We don't have a $0.00 minimum yet, but we're flirting with it
46:44
because it's like, I do think the world's going outcome based
46:47
and the product's good enough and we're getting the
46:49
performance for everybody. So if we say, hey, look, it's
46:51
free to try and we can get out there, get huge distribution,
46:55
get all the aggregate data that helps us make better ads.
46:58
I think that's the the mode over time.
47:00
I also wonder, do you think this whole new agentic world would
47:02
cheapen some of our products? Like Foreplay is a really
47:06
beautiful, awesome interface. How would you feel, Zach, if in
47:09
the year 90% of your customers were never seeing it?
47:14
I wouldn't care as long as the checks were still hitting the
47:16
bank account. I don't like, I'm not that.
47:19
Yeah, emotionally tax. Yeah.
47:20
Or maybe a better question is would they pay the same amount
47:24
right 4th. Place is very fucking cheap and
47:28
so like I, I like don't think all that much about like pricing
47:33
and stuff and like I appreciate you saying that about like the
47:38
UI and stuff like that and it's like, yeah, it's like cheap and
47:41
like I think our product quality is actually very good.
47:43
I don't like to we never position ourselves as like the
47:46
cheap option. We're just like our pricing
47:48
model like isn't attached to spend is actually like our, our
47:51
like angle. But that being said, we are
47:53
building make more AI features in which like we're going to be
47:56
incurring like different types of costs.
47:58
And so like going to some sort of usage on AI tokens is what
48:02
people seemingly are happy to pay for now.
48:03
Like it's kind of funny, Like even myself, I hate paying for
48:06
seats. We charge seats, but I'm like
48:08
willing to just like, you know, take a dump truck of cash and
48:14
throw it on Anthropic if like Co work just keeps working, you
48:16
know, like it's like, so there is like this weird little like
48:19
difference in the way people feel about stuff and like, I
48:22
think I I feel it. So, you know, there will be a
48:24
change. I just like, I don't think that
48:27
rash decisions are ever good actually.
48:30
So being like, oh, the world's going to like, you know, usage
48:33
basic. If I converted out like 4 play
48:35
to usage base right now, I could probably cut the business an app
48:37
or something, probably more than that.
48:39
So we will adapt over time through added features, not just
48:46
like changing the pricing model you.
48:47
Know what you just made me think of?
48:49
Not saying foreplay, but like there's a lot of SAS products I
48:52
probably pay for that I haven't logged into in seven months.
48:54
And if it was usage based, I wouldn't be paying that right?
48:57
Which would be more efficient for the customer.
48:59
And Zach, don't you think that this is the, this is the point
49:01
of growth? Like every new customer, if you
49:04
bring on with these new features that are usage based, you don't
49:08
have to do something dramatic, but you'll wake up a year from
49:11
now and your business model will be different.
49:15
Like like that's just sort of how this thing works.
49:17
Yeah. Yeah, yeah, yeah, yeah.
49:22
I think there's also, as it relates to foreplay, two ways to
49:26
think about the future of building the product.
49:28
And I'm sure that this is a big part of what Zach's thinking
49:31
about. You can build for that like 98%
49:34
of execution that's going to happen agentically probably very
49:38
soon, maybe already, or you can build for that 1%, which is
49:42
taste and accountability. That's handled by humans, right?
49:47
And I think that, you know, the user interface is the, the
49:50
visual part of the product, kind of the actual ad is right, the
49:54
video that they watch and they decide like, is this appropriate
49:58
for my brand? And I, I just, I don't know if
50:02
that's going to go away. I think that person that applies
50:05
their bias, that forecasts consumer demand, yeah, I don't
50:10
think that's going away. I think that's, I think that's
50:12
going to become more important. I also think that Foreplay's one
50:16
of the like luckiest companies actually of all time.
50:19
It's actually the luckiest company I've ever like been a
50:22
part of where it's like we, we accidentally timed the whole you
50:27
need to make more ads and creative is the only thing that
50:29
matters. You know, it kind of just like
50:32
stumbled into that little, little like wave.
50:35
And then so the way that I feel like we're like being kind of
50:38
like luckily like blessed with right now is the idea that like,
50:41
yeah, prompting is, is great. But like, what if you don't know
50:44
what to prompt? And like one of our long term
50:47
thesis is was like creative people worked like flow state is
50:52
triggered by inspiration. And so when I even just said
50:56
we're like, oh, now you need to be doing multiple things at one
50:58
time. Like, if anything, like
51:00
inspiration as like as a concept is going to be drained out of
51:03
humans at a faster rate. And so I just need to be
51:07
thinking about how do we replenish humans that have taste
51:10
to work in flow. I I agree that like there can be
51:13
agents that are going on the researching and finding angles
51:15
and USPS and all that kind of stuff, but there still is going
51:18
to be like human taste, like where is this company going?
51:23
And I think if we can just continue to help people
51:25
replenish that inspiration over time in different ways.
51:29
Like I actually know that's like the kind of like core of our
51:32
companies just put creative people in flow state.
51:36
You know what's nuts? I remember I was on a call with
51:40
a customer and screen sharing foreplay and we were showing
51:43
them as his big brand, right? And we press play on one ad.
51:46
It was by True Classic and they watched it.
51:50
I'm not sure if you guys have seen his ad, but it's like the
51:52
guy is like making a bunch of innuendo, a double entendre
51:55
jokes while he's like showing off his shirt.
51:58
And the the tastemakers at this brand were like falling apart.
52:04
They loved it. They loved this video that
52:06
they've muted, you know, but they were talking to each other.
52:08
They suddenly stopped paying attention and they were
52:10
inspired, right? And I just, I think that person,
52:14
the person that the brand, whether it's the founder, it's,
52:17
you know, the tastemakers at on the team, they're going to exist
52:21
and and they live for this, right?
52:24
This is this is what they love. This is what they're building,
52:27
right? They don't want an AI agent to
52:29
do that decision making for them.
52:31
I. Mean good news for John and for
52:33
you guys, right? Like that's sort of that's sort
52:36
of the core of the of the thing, right?
52:40
Is like at some point humanity comes back into the flow.
52:44
So, but yeah, I mean, what a crazy time to be building
52:47
software. It's so fun.
52:51
I'm right where I want to be. And what you just said yourself,
52:53
that's that's the such a big question.
52:56
Like I'm almost scared to say it because I love brand and taste
53:00
and design great coffee table books, but like what if tastes
53:04
gets taken over? Like what if it can amalgamate
53:08
and look to net new ideas in such a way?
53:10
Like already you see it on Sora. I mean, people are prompting
53:12
that thing. There's already automated
53:14
creative things happening that people are gravitating towards,
53:18
they're finding value from. I would hate to see the spark of
53:21
humanity and creativity go away. I still believe humans would
53:25
find lots of interesting things to do.
53:28
But we just, I just want to say the unspoken, it could be in
53:31
five years or 10 years, these things come up with better
53:34
creative ideas than I could at Miami, at school or, you know,
53:36
at an agency. Yeah, I mean, what If however we
53:39
read if this podcast is done by an EI agent?
53:42
It might have just happened before this call.
53:44
Actually, it's kind of funny you said that we had a, we had our
53:46
first SAS operator strategy call and yeah, we're renaming it, but
53:54
and we use Claude to rename it so.
53:56
So I was on something with the renaming.
53:58
Yeah, Yeah. We had the call quite literally
54:00
before you got on. Yeah.
54:01
That's. Funny.
54:02
Amazing. This was awesome, John.
54:03
It was awesome meeting you man. Cool, this is great.
54:08
Yeah, you too. This is fun.
54:09
Thank you for having me.