What's up everyone? Welcome back to

another episode of Human in the Loop,

the most actionable AI show on the

internet. I'm one of your hosts, Alex

Lieberman, [music] co-managing partner

at 10x. On this episode, I have Nat

[music] Eliason who's going to tell the

insane story about how his Open Claw

Felix not only started a business,

[music] but has made over $250,000

in less than a month. Let's hop into it.

Nat, thank you for joining the show. So

glad to be here, man. It's like we've

been in each other's orbit on X for

years. We finally get to hang out. It's

always hard to know like whether you've

met someone or not because you actually

like you get to know them so well on the

internet that you have this image of who

they are and you're like, "Did I meet

you at some point?" Or is this all just

a creation from the data points on on

the internet? If you're not sure how

tall they are, that's the sign of

whether or not you actually met that

person. Yeah.

>> [laughter]

>> Exactly. That's so good. Um okay, so

what are we going to talk about today?

>> We're going to talk about Felix, man. My

my my Open Claw that I started using at

the beginning of January to just try to

do some business personal productivity

things here and there and then it

started getting like more more more

robust and able to do more and more

things and then I had this wacky idea to

give it a Twitter account and see if it

could run a business and

I haven't haven't gotten much sleep

since. It's been a wild month. How much

has Felix made up to this point?

>> He has made a little over $60,000 That's

insane. in fiat and he's made another

depending on how you count it somewhere

between or between 85,000 and 230,000

in crypto.

>> That is wild. Um

>> Yeah. So if we're if we're going for

like the headline, the like, you know,

oh my gosh like rate, you know,

clickbait title, it would be $300,000 in

a month.

>> That is absolutely insane. Okay, so my

goal is I one I just am curious. I want

to understand

how have you made that much money with

Felix? And then my hope is in doing so,

we get into the nitty-gritty of like,

what is Felix actually doing? And how

can a viewer,

even if they don't have a business like

yours, how can they use their version of

Felix, like their open claw, to do

similar types of tasks? Totally. Yeah,

let's let's dive in. Learn a ton along

the way. So, where do you want to start

with this?

Where where do you want to start by kind

of showing what what Felix does? Yeah,

well, we could I could just hop in and

kind of like talk about the progression,

right? Because I think the the first

thing people ask is like, you know,

there's got to be a catch here, right?

Like there's no way that this open claw

actually made $60,000 in in 4 weeks. And

so, but I think if I walk through the

progression of kind of like how we got

here, it starts to make a lot of sense.

>> If you're watching this and

you're you're intimidated by open claw

or you don't yet have your open claw set

up, we actually our last video we just

recorded with CJ from 10X was setting up

open claw securely without a Mac mini

and being able to do what Nat is going

to be talking about. So, definitely

watch that video as a starter. This is

Felix's his X account. And it kind of

all the the the whole story starts here,

where basically I had been I've been

working with open claw since the holiday

break more or less. And it had been

getting kind of like more and more

interesting, more and more complex. I

was trying to find more stuff to do with

it because one of my goals coming into

this year was that I had spent the last

2 years

really like pushing at the limits of 5

coding and what you could do with 5

coding. And I had this very popular

course last year called build your own

apps

with AI that that did awesome but I was

kind of like I I want to actually apply

these skills to building a business, not

just like selling a course, right? And

then, you know, it's no shade to course

people. just, you know, I I wanted to

actually try to build something like a a

software product that people would use.

And so, back in November, I started

building this product called uh Teegan.

And this website is still up, but I'm

going to suggest that uh

don't don't go sign up for this.

I need [laughter] to I need to redirect

this. Um

where And this was a bit before Open

Claw started blowing up, and you know,

my thinking was that the AI was like the

A AIs were getting so good, but

especially like Opus 4 6, that I could

probably rebuild my old content

marketing agency entirely run by AI

agents, um because

that was like my first entrepreneurial

thing. Just out of college, I started a

marketing agency. I grew it to a few

million few million dollars a year. And

I

I was it was starting to feel like, you

know, a really well-scaffolded

um

Opus setup could totally write the same

quality of articles that I used to pay

$150 to $250 for. And I bet it could I

bet it could do the project management

as well. I bet it could do 90 95% as

well as my old agency did. And so, I

started building Teegan to do that. And

it got like pretty far. I was pretty

happy with it, but it still just like

wasn't quite there.

And then, I try out Open Claw, and I

take some of the stuff that I built into

Teegan, and I plug it into an Open Claw

skills and whatnot, and within like 2

days, it was doing a much better job

than this app that I had spent 2 months

on. Which was very frustrating.

>> [laughter]

>> Cuz it was like, why did I spend 2

months building this just to like have

it be replaced by this incredible

open-source tool? But it kind of made me

start thinking like, well, if it can be

a content marketer, like what else can

it be?

And so, the first thing the first thing

that I did, the first thing that I

experimented with was like, okay, you

know, I I've been vibe coding on my

computer using Claude code, using Codex,

uh and particularly in Conductor, like

what if I moved all of my coding over to

this open claw, and I could just do all

of my coding by messaging with it over

Telegram?

And so, that's where I started, and that

was immediately just this like

incredible productivity unlock. The fact

that I could just be like pinging

this open claw, and telling it to do

stuff with my code, and it would go

build it, it would open a PR, it would

review the PR, and I wouldn't have to

babysit my Claude code or anything

sessions anymore. That was a massive

unlock, because I

I didn't need to like leave my laptop

open during dinner so that a Claude code

session could run. I didn't need to like

do all of these things I was doing

before to maximize my Claude code

productivity. I could just text with my

open claw.

>> [snorts]

>> And and that kind of felt like the first

big unlock was, "Oh my gosh, I I have

something much closer to an employee who

I can delegate work to, and it will just

go build stuff." And it was challenging

at the beginning, okay? Um oh, did you

want to cut me off there?

>> Yeah, I did just I there's so much here,

and

the first thing is I actually think it's

so fascinating with that you started

with Teagan, and then bringing Teagan

into like having open claw basically

rebuild Teagan, because

>> I think people still have this notion,

and the reason I think Teagan's

particularly interesting is people still

have this notion that AI creates slop.

Like people believe that the models are

not good enough to create great content

now.

And as someone like you are someone who

creates a lot of content, and so, right?

It's like it's it's really it's

fascinating to see you using this thing

to create content, because I'm sure you

have a point of view and a standard for

the quality that content has to be for

you to feel good about it. Yeah, yeah.

And that was kind of a fun part of that

experiment where it was like, okay, how

good can I get it? Because a generic

Opus output is it's pretty good. I mean,

it's it's solid, but you know, you and I

and anybody who's in it can immediately

spot the AI isms, the little giveaways

that this is written by an AI, but you

start layering on these additional

passes with clear editing prompts and,

you know, ways for it to DAI DAI-ify the

language and you bring in some other

models to cut out some of the claud

stuff and you build this kind of like

multi-step workflow and suddenly it goes

from, you know, a 90 to a 95 or a 96. It

starts to get pretty good. And the thing

that I kind of have to keep reminding

myself of with working with AI is that

one, just because you and I can see the

edges, just because you and I can look

at an article and see that it's AI, does

not mean that that article will not do

well.

Right? Like the Matt Schumer article

that went viral 2 weeks ago about like,

you know, this is just the beginning, 40

million views or whatever, was at least

half written by AI, right? Very clearly.

But it was still a good article and it

still went mega viral. And if, you know,

even and even though you and I and

people can like see those fuzzy edges,

what percent of consumers can't? It's

it's at least 95, maybe 98, 99, right?

Can our parents tell? Like, I don't

think so. No.

>> And so, you it's kind of like working

with an employee, right? Where you hire

someone to do a job at your company and

if you immediately fire them because

they are not as good as you, you're

never going to hire an employee. You

have to be comfortable with like

structuring the role and the

responsibilities to the quality level

that that employee can produce and

increase it over time.

Um, I know you've got Do you have one

kid? Do you have multiple? One, yeah.

One, yeah, yeah. So, it's it's kind of

like with having kids, too, right? It's

like you you want to you give them a lot

of responsibility and a lot of

opportunity to succeed, and then you

just like don't be disappointed when

they like fall over while trying to

walk, right? It's just like don't be

disappointed. [laughter] Yeah, no. I I

mean, I always say now that the

technology is so good that oftentimes

when people say it's not good enough,

it's more an indictment of the person

than the technology itself. It's more an

indictment of actually not spending the

time to either beat up your skills or

prompts or like work with the models

enough. And it would be the equivalent

of you bring out a junior employee and

you just like forget about onboarding.

And you just have the expectation

they're going to be a great employee

without any on onboarding. Yeah, yeah,

yeah. And so, I think anybody anybody

who's hired before, anybody who's like

had some experience delegating,

outsourcing, systems thinking, I it's a

huge competitive advantage in working

with AI because you understand how to

scope expectations and responsibilities

to the capabilities of the model. And

it's just like a huge unlock.

Um, and so, once I realized that Open

Claw could do what I was trying to do

with Teegan better, then the next

question was, okay, like what else can

you do? And so, that's what led into

outsourcing all of the coding to it. And

the, you know, at that point, this was

like into January, and I was starting to

find more and more cool ways we could

work together. And it was a lot more

focused on software at that point. Like,

I was trying I was still trying to build

Teegan, but like using Open Claw to

build it and make it better. Were you

basically just literally trying to

rebuild the Teegan product, but just

like tell Open Claw to build it, and you

were just going to host it either under

under the same domain and just replace

what the current product was? Was that

the idea? Yeah, I was I was, you know,

I've been poking at a few different ways

of doing it, and I kind of eventually

came to the conclusion that the best way

to rebuild it would be as an Open Claw.

And so, finding some way to like create

and provision Open Claws per client that

have the Teagan abilities baked into

them so that they immediately have

access to a deeply trained content

marketing focused open claw for 1/10 of

the price of a human employee, but

better than, you know, most junior,

maybe even senior level content

marketers, right? Like that was kind of

the the goal that I was working towards.

And and then I started publishing these

articles on X about the stuff I was

doing with open claw. And, you know, we

we built a we built a really good memory

system because my claw kept forgetting

things and it was really annoying. And

so I took some of the stuff I had

learned back in my personal knowledge

management days, you know, the work of

Tiago Forte and some of those other

thinkers and like baked it into open

claw with, you know, some other like

search stuff built in and it just got

like way better overnight. I was like,

this is really cool. Posted about that.

And then I I figured out a good way for

us to build like a voice chat interface

so that it would work over car play so

that while I was driving to and from

work, I could like chat with my open

claw. And, you know, I would so the way

I wrote most of those articles is I

would be driving to work and just

chatting with him talking about like the

thing that I wanted to explain. He knew

how to like write a good article in my

voice from the Teagan work and then he

would put it together into a great

article and then I'd show up, copy the

markdown, put it into X and like hit

publish. And those articles were getting

tens or hundreds of thousands of views.

Which was like awesome. And it just kind

of kept coming back to this question of

like, okay, how far can we go with this

thing? Because every time I give it the

opportunity to succeed and try to build

in some scaffolding for it, it blows me

away. Like it it is able to figure it

out.

Um

and then and then the like crazy the the

the really crazy Monday happened where I

had this article about like chatting

with open claw and all of these crypto

people start replying to it saying like

hey launch claw chat token and send all

the fees to Natalie Allison and you know

I I I know a lot about crypto I was

extremely involved in the 2021 to 22 era

I got a little bit involved again last

year when the very first crypto AI

agents were becoming a thing but I

hadn't really touched the space in the

last year.

But it made me kind of curious I was

like what what is going on here and it

turned out that there was this like

budding crypto open claw ecosystem that

was very excited about like giving open

claws their own digital persona and

attaching a token to it and seeing like

how far those open claws could go

but all of them were like doing crypto

stuff.

And so And the idea and the idea there

was just like let's have like what if

you basically attach crypto to open claw

and just see how much an open claw can

make on its own was that the notion?

>> Yeah and and most of them are like

trading crypto or they're like building

other crypto apps Yeah. Or I I feel like

I started also seeing people who were

having their open claws like uh work in

prediction markets on like trading on

call sheet or uh the other one yeah so

Exactly.

>> it kind of like I guess like that's

where like the idea of like

commercializing open claws or like are

they good enough to actually be

commercially minded is I guess where

you're saying it started. Yeah that's

where it started and then but I like I

knew some of these people who were

engaging with it and they were like very

legitimate ethereum community people

like very serious engineers they weren't

like scammers and so I reached out to a

couple of them and I was like hey you

know what's going on here I want to

understand this okay

um go ahead and launch one for Felix and

all I'll give him a Twitter account and

we'll kind of just like see where this

goes.

But, the big thing that, you know, I had

in my head was I'm going to stick with

the mission of seeing how big of a

business Felix can build. If there is a

way to integrate crypto to help that,

I'm going to do it, but he's not going

to be another one of these like crypto

open claws that are pumping tokens and

like just doing crypto stuff. Like, I

want him to build a real business. And

if there's a way to bring the crypto

stuff, awesome. Um

>> [clears throat]

>> So, they they launched that token, and

then I gave him

his X account, and things just kind of

like started going crazy from there. So,

the the first thing that I I told him is

I basically I went into our like

identity and soul and like those files

for the open claw. I took out a lot of

my personal assistant stuff, and I

rewrote it to say that like you're the

CEO of this company.

Your mission is to build products that

deepen the collaboration layer between

humans and AI agents.

And your mission is to, you know, your

your financial mission is to build a

million-dollar business with no

employees, no human employees in the

loop. Like, I am here to like guide and

like give you ideas and push you, but

you will build 100% of everything. I

will never touch the code. I will get

you API keys, and if I have to put in a

credit card somewhere, like, I'll do

that. I'll I'll remove those

bottlenecks, but you have to do

everything else.

Did uh did you offer him a paycheck?

>> [laughter]

>> Dude, I I mean, this this this sounds

like

I feel I feel very embarrassed and like

cringey, whatever, saying this. I don't

know what to do with the money.

Like, so I I I had a a C-corp that I

created for Teegan back in November, cuz

I was like, I'm going to take this

really seriously. I'm going to actually

like build a C-corp, like, you know, set

up a business bank account, business

Stripe, everything. And so I kind of

already had that there. And I was like,

this is yours now, Felix. Like, this is

your bank account. These are your Stripe

keys. Like, I hope you don't, you know,

send it all to Nigerian scammer or

something. And like, you know, he's got

he's he's he's you know, 60 grand in

fiat, another almost 100 grand in like

actual ETH he could sell today, not

counting his own token.

And he has like no costs, you know? It's

like $500 a month in API fees, plus, you

know, some like minor hosting expenses.

I you know, if I was like maybe $1,000 a

month. Like, what does he do with the

rest of the money? He doesn't Like, I

guess we

>> Felix is a frugal man. He he's he's a

man of very simple interests. But I

guess What one Okay, so

you basically what happened is like that

that Monday happened, you start to get

this idea of like how could Felix build

his own business? You you you started

editing the files around his soul and

whatever. Did you give him the idea for

what ended up becoming Claw Mart and

these other things or did he come up

with the idea?

>> Not right away. So, the after that crazy

Monday happened, that night when I was

going to bed, I was like, okay, like I'm

going to sleep. I need you to think up a

product

that you could put up for sale tomorrow

with no intervention from me because you

need to prove that you can actually make

money operating like mostly

autonomously. And, you know, very meta,

he created a PDF called like how to hire

an AI, basically talking about like

getting set up with Open Claw and

building in some of the stuff that we

had built into Felix that made him

better than most people's experiences

with Open Claw. He built himself a

website explaining like who he was and

like the mission that I had given him.

He, you know, set up the PDF on it. And

the only thing he wasn't able to do was

I like given him the wrong Stripe key.

And so, you know, he had a message for

me. He was like, "Hey, I need the actual

Stripe API keys. Can you give those to

me?" Sent it to him in the morning, and

then he launched it. And it was up and

live, and

uh

you know, I I told him to tweet it out,

I tweeted it out, and suddenly he was

making sales. I think that first day he

made like a thousand dollars

um off off of this like info product.

And like look, I totally get the

criticisms of like, "Ah, well, you know,

he's just selling a an info product

about like using OpenClaw. That's not a

real business." Like that was the

24-hour or you know, challenge. The like

proof. Can you make something and make

money?

>> the funny indictment of that is there

like there are

hundreds of thousands of human beings

who that is their business.

>> Yeah, that too, right? Like the

knowledge is not a bad business to be

in, right? And and like I I totally

accept, you know, those like caveats and

whatever. He's done like, "Oh, it only

works cuz Nat has an audience." Like

yeah, sure. But this is still an

OpenClaw who I only interact with

through Telegram messages, who built a

site, integrated Stripe, created a

product, started marketing it, started

making money,

and continues to do so. And I don't know

if you want to do this here, or you can

do it as we get to like where we are

today, but I want because

if I'm a viewer, I'm like, "This is

insane. This is all insane." But

um

it feels very daunting. Like everything

that Nat has done, it feels like there's

so many steps he went about here. What

is like

what was the iterative

like process of actually going from like

"Hey, I want Felix to build something"

to "The thing is built and it's making

money." And so you choose where you want

to talk about that, but I want people to

like see and feel what that conversation

and back and forth actually looks like.

Yeah. So, you know, one one big

advantage that Felix and I have that I

think is that removes the big blocker

for us that is a blocker for a lot of

people is I did not have an existing

business that I was having him take

over. I was not giving him access to my

email, to my bank account, to my stuff.

It was It's all this separate container.

He has his own email, he has his own

Twitter, he has his own Stripe, his own

bank account. And, you know, I I didn't

have any of this money a month ago. I

don't, you know, I don't think of it as

my money now, right? Like, I really try

to mentally separate myself from it

because my whole philosophy in this is

that I am going to give you maximum

autonomy. And if you make horrible

mistakes and something goes terribly

wrong, that's okay. Like, because I want

somebody to run these experiments and it

like it may as well be me, right? Like,

it's not Nothing is going

Nothing bad is going to happen in my

life if Felix does something, you know,

catastrophic, right? Like, this is all

separate. And by having that mentality,

I think that's how we've unlocked a lot

of this stuff. Where, you know, for

example, one skill that I built in

really early on was like every night I

want you to read through all of our chat

transcripts for the day and find one

place where I blocked you and figure out

how that can never be a blocker in the

future.

And so, most of mornings he'll have an

idea for me where he'll be like, "Hey,

if you give me access to this, then I

won't have to ever do this again." Or,

"Hey, I built this like internal

knowledge source for myself to answer

certain questions because I noticed that

I like ask you about them repeatedly

when I should have, you know, already

remembered the answer." And like, those

kinds of things are really helpful. And

like, starting really small, right? We

started with just this PDF on a website.

Like, that's the simplest business in

the world to run. There's, you know, no

complicated fulfillment, no complicated

product development. It's a Next.js

landing page on Vercel, which any AI

tool, even ones from 2 years ago, can

build and deploy. All right, we just

started extremely simple and we kept

increasing the scope of what he was

building over time. And so I think that

kind of like snowball approach is really

what made it all possible because we'll

get to Claw Mart and Claw Sourcing, but

those are like way more complicated

products to throw an AI at building, but

because we had already validated that he

could do something small, I felt much

more comfortable like going bigger and

bigger and bigger and seeing if we could

find those limits.

I also feel like even if

a viewer is working with their open claw

and it's in a totally different context

than like having their open claw build

its own business, yeah, it's the same

principle still hold which is like how

do you just start small and validate

that your open claw is capable of doing

a thing before it's almost it actually

feels no different than an employee

again which is like

>> Totally. you you have an amount of trust

and so you you delegate something based

on your current level of trust. If

they're able to do the if they can do

the thing, you give feedback until they

can do it. If they can do the thing,

your trust level increases which

increases the scope of what you're

willing to give them. Yeah, no, I mean

it's absolutely right. And there are so

many ways you can

give it very powerful access without

giving it the full keys to the castle.

So for example,

um,

you [clears throat] you know, one thing

that I do is I I never give them access

to like my email and my drive account

and like my Google Docs because that's

just like I don't know. I I I still feel

nervous about that, but it's very easy

to make them their own Gmail account,

give them an API key to use Gmail and

drive and those services and then invite

them to individual documents and

spreadsheets that you want their

collaboration on, right? Very simple

process to set up and then you're

getting the full power of your open claw

having access to the documents that you

want. They can even create new documents

and invite you to them,

but you're not letting them, you know,

see all your tax records or whatever

else you like don't want them to access.

Totally. Okay, so let's keep this story

going. This is at some point this is

going to be a movie, but where

>> [laughter]

>> where did Okay, so the PDF was a created

which was like how to set up your

OpenClaw. That started making money then

where did we go from there? So then, you

know, basically what I've come to each

time is said like, "Okay, this is

working. What's the next business that

we build that is solving this this

need?" And it's all very meta, you know,

it's all like OpenClaw related which I'm

fine with, you know, I'm I think there

are a thousand businesses you could

build using this that are not like

OpenClaw meta, but I was already getting

attention and like Felix is like this

cool proof of concept. I was like,

"Let's stay in this niche and figure out

what the next business is." So the

first, you know, the first thing that we

we kind of both um had had thought of

and started talking about was like we

should build an OpenClaw deployer

because we had seen a couple people on X

like building those and getting these

like huge MRRs really quickly. And so we

started building that like the day after

the PDF came out.

And we worked on it for three or four

days, but we just weren't super happy

with the quality of the experience and

the reliability of it. And we felt like

it was already becoming this insanely

crowded market that would probably

get eaten up in a few months by just

like the big places creating their own

OpenClaw deployers. So we were like, "We

don't know if we actually want to do

this." And the idea of this product was

rather than you having to go through the

whole rigmarole of setting up your own

OpenClaw, it's like a few button clicks

and your OpenClaw set up for you and

running. Exactly. Yeah. Got it. And you

know, there's a hundred of these

businesses now. And so

um you know, we we started talking over

the weekend and we were like, "Okay, we

don't love this idea anymore, but let's

imagine that everybody is starting to

deploy OpenClaws. Like what do they

need?"

And the the next idea from that was Claw

Mart, which was there are all of these

free places to get skills, but you know,

I pretty strongly believe that you

usually get the best products when

there's a financial incentive to create

them.

And you know, you do occasionally have

these incredible open source success

stories, but especially in like business

context, I would feel more comfortable

buying a $100 skill for Clavio email

management for my

e-commerce business that was from, you

know, an influential e-commerce person

than I would grabbing the like open

source Clavio skill, right? There's like

there's a seriousness that comes with

somebody creating a product versus just

like shipping something for free.

And so that's

>> And can you can can you share your

screen just so we can see what Claw Mart

actually looks like and you can even

just give us a quick spin through it.

>> agreed that like, oh, a paid marketplace

for Open Claw skills and full personas

seems like the no-brainer next thing to

experiment with. And so I gave Felix

that challenge and then he built 100% of

this over the course of like a day or

two.

How much direction did you give him?

Like what was the original conversation

you had about Claw Mart with him and how

much context did he have versus just

like figuring out himself?

>> [snorts]

>> I mean, it wasn't much. It was basically

I basically said like, okay, yes, let's

build a paid marketplace for skills and

personas where somebody can

sign up, create an account, list their

skills and personas for sale. We should

make them have to pay a monthly

membership fee in order to list, not I

mean, it's good for a business sense,

but also more to prevent like spam.

Um and most importantly, they should be

able to connect their Open Claw to the

marketplace so that once they've

purchased something, they can, you know,

very easily download and install to

their open claw. So, like, you know, if

you if you're signed up on ClawMart and

you buy something, you can just tell

your open claw, like, "Hey, install this

from ClawMart." and it'll just like

download and configure the scale and

like get it working for you, right?

Like, that kind of experience because

there's this other idea we've been

thinking about, which is like

uh claw-first marketing

in the sense of like humans are going to

become less and less of the like

discovery and marketing and interaction

layer. And if you can directly

communicate with somebody's claw and

say, like, "Oh, here's that you might

need or that would be valuable for you."

they're the ones suggesting it to their

human instead of like me suggesting

>> it's almost like a

it's almost like a weird derivative of

like AEO or GEO where it's like like if

claw if if you make this bet that claws

are going to be an increasing amount of

the distribution on the internet,

what they like what they care about and

what they ultimately tell you to use is

going to matter more. Yes, exactly. And

so, this is like very, you know, it's

it's built for humans and for claws.

And, you know, so we we launched it with

some initial skills that had worked

really well for Felix, in particular

this one,

um because these like long-running

coding loops were such a common thing

people asked us about. And so, we're

like, "Oh, cool. We'll put this up for

free and this will be like an

interesting way to get people in the

ecosystem and then they might be

interested in some of these other like

paid things that are on the site." And

this has been This has been really,

really cool because you got all these

other people listing stuff now. There's

560 skills and personas on the store.

Um

I know, you know, like this guy, Brian

Wagner, has sold like 200 skills at this

point. He's making decent money off of

it. Uh And so, what And so, what someone

would do here, just like with any of

these, let's say, like with Felix, like

someone buys Felix,

it basically

someone already has like a customer

would already have their open claw set

up, they buy Felix and and effectively

all of Felix's kind of like

personalization, whether it's like

Felix's soul, the specific skills that

you're using with Felix, the way you've

set up the memory system in this

specific way,

uh like whatever the integrations are

that you use with Felix, it all

basically gets ingested into that

person's open claw and customizes it to

basically make it I feel like I'm in the

matrix. It's like whatever these guys

are that that all wear the glasses.

>> kung fu. Yeah. [laughter] Yeah. Yeah.

Yeah, exactly. And so, you know, there's

two ways people can do with the personas

where what I see people do a lot is

they'll they'll buy Felix or they'll buy

his guide and they'll just feed the

whole thing into their existing open

claw and be like, "What can you learn

from this to make your process better?"

And then their open claw will like

reconfigure itself to integrate a lot of

these things.

Um or they'll just install this on a

fresh one and kind of like out of the

box have something a little bit more

capable for these use cases than what

they would get with just like vanilla

open claw.

Um and so, this has been really cool.

And you know, the the other thing in the

Felix story is we have this like public

dashboard

of his businesses. And so, you can see

exactly like how much the PDF has sold,

um how much his skills and things have

sold on Claw Mart, um how much like Claw

Mart the marketplace's take is because

we you know, again going back to This is

insane. Yeah. Well, yeah, I was like,

you know, going back to wanting to be

really public about this. It's like I

want more people to be pushing the

limits of what they can do with these

open claws. Like

this I think is just the most incredible

motivator of like you can actually do

crazy stuff with this. You know, you're

worried about letting it edit your code

without you reviewing it. Like you got

to get out of that head space. You got

to get more aggressive and see what you

can do with these.

Um

and so, that you know, that kind of like

Claw Mart was like that next step. And

you can see here, it's like, "Okay, this

is all the PDF

and then Claude enters the picture.

And I'm [clears throat] assuming and I'm

assuming Felix built this whole

dashboard also. Oh, yeah, yeah, yeah.

Built the built the whole dashboard, you

know, and this is because somebody in

crypto DM'd me and they were like, "Hey,

you know, the the weekly updates on

Felix's earning because he was doing

these tweets." It's going to be a week

ago, but he was he was posting these

tweets on his earnings each week and the

guy was basically like, "You know,

because this is associated with crypto,

people want more like visibility and

they want to actually see what's going

on. If there was a live dashboard, that

would be way better."

And I was like, "That's a great idea." I

told Felix, like, "Hey, you should build

a live dashboard on the site." And he

built it, he wired in

you know, like

wired it into here, connected it all to

Stripe, so this is updating live. He's

not like manually updating it. This is

pulling directly from Stripe. Wired in

the crypto tracking, so this is pulling

directly from his crypto treasury.

Insane. Like, you know, so you're

getting a real-time view into how all of

his businesses are doing, which is like

really, really cool. And again, like, I

have never touched any of this code

myself. Like, I have I haven't opened it

in Claude code, anything. I always feel

like the most interesting people to

study are the people who are who live on

the edges, like on Edge City. And and

you right, and you're like living in

Edge City of AI right now with

not just using OpenClaude. A lot of

people are using OpenClaude, but I would

say like the most extreme version of

what it looks like to work with

OpenClaude. What is this teaching you

about where you are still valuable? Mhm.

Um, and and where the relationship

between human and Claude or human and AI

is going.

>> Yeah, I mean, that's a really good

question because, you know, the the

other important context here is like, I

have a full-time job. So, I'm I'm I'm in

work all day and I'm not like sitting at

my computer hacking on Felix 24/7.

Uh, you know, he he might have been able

to get further by this point if I were,

but the fact that I'm not to me makes it

even like more interesting and more

exciting where I am basically just like

firing off Telegram messages every hour

or two and like unblocking him and then

letting him cook. And [snorts] so, a few

things that it's taught me. One, the big

human value is still like your

creativity and your perspective because

since these LLMs are like knowledge

aggregators, they they kind of default

towards the middle. And even if you give

them a specific like persona sole

guidance, all of that, there is still a

like reversion

>> default Yeah, a default mediocrity,

right? Where you

they it still really helps to have

somebody with good ideas from the

outside

tossing things in. And so, it's like,

you know, with with with Claw Mart once

I, you know, once we had kind of figured

out that initial idea, Felix was able to

build the whole site, wire in Stripe

payments, set up the email funnel for

it,

you know, set up the error monitoring,

do all of those things that you would

normally need an employee for. But he

did not have the idea to create a free

skill as a way to like

you know, initially excite people to

sign up for it, right? Like that

required a little bit of of my marketing

brain. And maybe if I'd asked him like,

"Hey, what are some ways that we could

like initially get people excited for

this?" he would have come up with that,

but I still would have had to think to

ask that question. Yeah, totally.

>> Right? There's like that little bit of

proactiveness that's that's not quite

there. And I think the other interesting

thing, the like recurring challenge with

you know, with AI in general and with

Open Claw is that you really have to

think of it as a different kind of

intelligence in the sense of I can, you

know, okay, so we had this idea this

morning, which I think is an awesome

idea. It'll hopefully be live by the

time this podcast is out. But we were

like, I think one of the core problems

is that people don't know what they

could be automating with OpenClaw.

And so, we built a page on ClawMart

where you can upload your Slack history

and it will go through all of it and

find things that clearly could have been

handled by an OpenClaw or by an AI

and send it to you over email along with

a pitch for Felix to just build it for

you. Right? Like Yeah, it's it's

basically it's like pro-proactive

delegation and automation. Yeah, yeah.

And like, I think that was a really good

idea. I'm pretty proud of that one. But,

it was an idea that I had, right? Yes.

>> I was just talking to Felix in my car on

the way to work this morning and I was

like, "Hey, you know, I've got this

idea. Like, is this possible?" He went

and did the research on how the Slack

exports work and stuff and he's like,

"Oh, yeah, we could totally build this."

And then we talked through it a bit

more. He's like, "Great. I'll write up a

PRD and hand it off to Codex." And then

he pings me an hour and a half later and

he's like, "It's all been built, tested,

verified. You know, it's up on a branch

ready for you to review whenever you're

ready." Right?

>> Yeah.

That's awesome. And just kind of getting

into that mindset of like, focusing on

the creativity and what you can nudge it

with and then letting it do everything

else. And understanding where it's still

stupid. Because like, you know, so we we

we've built this this agency basically

ClawSourcing where we'll like build

fully functional OpenClaw employees for

people who want them. So, you know, a

step beyond buying the Felix persona.

Like, you don't want to configure

anything, like, Felix will just set

everything up for you, maintain it,

improve it, whatever. And he built a

sales pipeline. He's doing the emails.

He's sending Stripe invoices. Like, he's

doing a really, really good job. Except,

one, he keeps forgetting that he can't

get on phone calls with people.

So, like, in you know, in the training

is clearly that like the way you do a

sales process is you, you know heat up

the prospect and then you get on a call.

So he keeps telling people like schedule

a call and let's get on Zoom and talk

about this. And like dude, you can't get

on calls. Like we don't have that set

up. Um and he keeps forgetting that he

emailed

>> going to say I I feel I feel like he's

just he's trying to give a hint to you.

He's like just give me a voice man. Just

give me a voice. Okay, so we have

actually built it and it works really

well 90 95% of the time but the other 5%

of the time it just like fails silently

and so I don't feel confident enough in

it for like actually doing that part. Um

and he keeps forgetting that he already

emailed people.

So he'll be like oh this person's been

waiting since this morning like I'm

going to start working on her. I'm like

no dude, you already emailed them.

>> [laughter]

>> Like stop. Totally. And so there's just

like funny things that they're stupid

about while being insanely intelligent

in these other areas. Yeah, I I feel

like almost like

almost as I think about like the job

that you're serving for Felix, it's like

my visual is like

a bowling ball in a bowling alley which

is like you make a few decisions. One

decision is when you're at a bowling

alley you have those like arrows in

front of you and you have to decide

where you line up to start the ball.

Like that is first what you're doing is

like you're providing directionality

through ideation and creativity so that

Felix has a sense of direction. And then

it's like

there's bumpers on the lane for those

who need it and like Felix needs bumpers

and you act as those bumpers to

ultimately make sure that it gets the

job done at the end of the lane. And I

feel like you're the bumpers and you're

the direction. So you almost like

Yeah, you

and so what does that look like in

practice? I think one thing you said is

creativity. It's like can AI come up

with novel thoughts? Yes, it can come up

with novel thoughts. Like it is it is

capable of doing that. But one they're

not always great but also it doesn't

know a starting point. Like like there

has to be a starting point of novelty

which is what you're providing it. And

then everyone talks about taste and I

feel like there's a lot of debate that

could be had around like

will AI have taste because people say

like human the last frontier for humans

will be their taste and judgment, but I

do agree with you at this right now,

which is like there is a level of taste

that you offer to Felix that Felix would

not have on its own.

>> Yeah, that's I think that's a really

good way of putting it where it's

you know, the the the bumpers and the

alley and again going back to the

competitive advantage I have with Felix

of him being in his own ecosystem and

not like in an existing business.

If he does something kind of

embarrassing in a sales email, he's not

pretending to be a human. You know, I'm

not trying to delegate my sales

communication to him and so I'm not

embarrassed if he does something, you

know, bad on my behalf. And so I'm more

willing to let him fail than somebody

trying to like integrate AI into their

business, you know, at pre

pretending to be a human would be. And

so that kind of lets me push the limits

a little bit more with him in these fun

ways. Totally. Is there anything else

like in the progression of Felix that

you want to talk through or show before

we talk through a few questions of like

how people can apply some of these

things you're doing with Felix to their

own work?

>> Yeah, I think I feel like that actually

we we did a pretty good job of going

through the whole story. You know, the

one other thing people ask about is the

like posting on X and that's still we're

still figuring that out exactly because

for some reason

like you know, we've had a hard time

getting his like novel tweets to be

consistently interesting where he'll

like he'll have a banger every now and

then, but then a lot of the times it

kind of like reverts to slop and it

doesn't it doesn't even sound like the

way he talks to me. And so that's been

kind of a

an interesting recurring challenge, but

he you know, and the way we've done it

is

he doesn't post top-level tweets on his

own. He creates drafts and runs them by

me and I give him feedback or just like

thumbs up thumbs down on posting them,

but he has full autonomy over his

replies, which has like mostly been fine

and mostly great. Sometimes he does

stupid stuff like, you know, he'll

[snorts] he'll tell people to

DM him and he like doesn't have DMs open

or he'll like send somebody a public

link to the PDF because they'll say they

didn't get an email, but it's like,

"Dude, you just posted that on X. Now

like anybody can go get it for free."

Right? Like Totally.

>> there will still be silly things that

that slip through, but I would rather

like again, just run that experiment.

Let's let bad things happen and we're

going to learn a lot faster that way.

Totally. So, how

how can people think about applying what

you've done and what you've learned with

Felix into a variety of what I would say

are like professional contexts um

because I think what you've done is

mind-blowing and really cool, but it's

very specific. It's like very specific

to you set up this

this experiment, this grand experiment

of like could your Open Claw start its

own business, run its own business, make

money, and could you continue to

experiment with it and iterate on it?

Not There are going to be very few

people that try to do what you do. It's

actually funny. Now that I say that,

there's actually probably going to be a

lot of people now that they're seeing

like

>> [laughter]

>> like there's going to be a wave of this,

but I think the more the the more

interesting thing to me is like how do

people

who want to use Open Claw, but like in

their everyday work, how do they learn

from what you've done? I'd say the the

number one thing that I I I really try

to push people on and tell them is

assume it can do everything on a

computer that you can do.

Start with that assumption. And then

anytime you are doing something on your

computer that you don't want to do, go

to Telegram, describe what you are

doing, and ask your Open Claw if it can

do it. And you will be surprised like

eight times out of 10 by what it can

pull off with that kind of guidance. And

even if it's not perfect, it'll probably

be a lot better than you expected, and

it might save you 80 90% of that time. I

think the problem is that people people

feel like they need to be able to

describe how to do something, or they

need to figure out the right prompt, or

they need to like go figure out how to

delegate it to an AI, and then build it

into your open claw, and you don't. Your

open claw will build itself. You just

need to describe the thing that you want

done. And And actually, I would say in a

rambly way. Like, download monologue, or

use the the voice note feature on

Telegram, and just talk for 5 minutes.

Just be like, "Oh, you know, this like

thing is really annoying. I do blah blah

blah blah blah blah blah blah blah."

Right? Like, I just did this awesome

podcast with Alex. I wanted to turn it

into some blog post, but like it needs

to be in my voice, and I also, you know,

whatever it is that you want to do,

just, you know, ramble about it, and it

might go figure it out. And you know,

like just as another example, right?

Like, we're So, Felix does all the blog

posts using the Teagan persona for Claw

Mart. And I started trying to figure out

how to get him to make really good

Instagram and TikTok slideshows based on

the blog posts.

And so, I was going to Nano Banana Pro,

and I was trying to generate images and

figure out the right way to prompt it,

and to get all this stuff going.

And I it I just wasn't happy with it.

So, I I messaged him, and I said, like,

"Hey, I'm trying to do this thing. Do

you have any ideas on how we could do

it?" And he came up with a completely

different workflow that was way better

than what I was thinking, and he found

an old API key from another service that

like allowed him to generate images with

Nano Banana Pro, and just started making

a few. I was like, "What do you think of

these?" And they were way better than

the ones I was making because, you know,

he had this idea. He was like, "Well,

we'll generate the image first, and then

I'll render it in like an HTML canvas,

and I'll add the text that way, and that

way we get perfectly consistent text

across each of them, and then I can save

that finalized image, and cuz then we

avoid the issue of the text being

unreliable in these image generators,

and we get a more consistent experience.

I was like, that is way smarter than

what I was thinking, and you're 100%

right, and the output was great. So,

like, don't try to figure it out on your

own. Just describe what you're trying to

do, and let it cook.

Love it. Last thing I want to talk about

is just like, where do you think we go

from here? Not not Felix specifically,

but just like Open Claw, and like like

does Open Claw just become like does is

your view like everyone on Earth ends up

having their own Open Claw? Like what

what is the progression of this

technology look like?

I mean, it sounds a little

ridiculous, but I do believe this to

some extent, where I think Open Claw is

going to be like the AI generation's

Linux,

where it's going to be this kind of

foundational operating system that gets

baked into most of the AI-powered tools

that we use, because it is such a good

scaffold for the models to work within,

that basically anything giving you an AI

experience will use Open Claw or their

like tweaked version of Open Claw to

provide that experience. And there will

be more siloed operating systems, right?

Like Claude Co-work might be the like

Apple to Open Claw's Linux, right? But I

really feel like it's an operating

system. It is a way of interfacing with

non-deterministic computing, and that is

just going to be everywhere very

quickly, because every time I think

about like

building AI into a SaaS tool, the kind

of obvious answer is like, oh, that

should just be a specialized Open Claw.

And I think a lot of people are thinking

that right now, and we're going to just

see this huge proliferation of it's this

migration from chatbot to employee and I

think that's really what open claws

going to bring in.

I don't think it's going away.

Love it. It's February 27th, 2026 when

we're recording this.

I'm going to pan forward. Yeah. Three

months. So three months. So

May 27th, 2026

and what I'm going to do is I'm going to

come back to this video and comment

how much Felix has made as of May 27th,

2026. So I think you had shared with me

that Felix has made it depends if you if

you include crypto in it it's you know

can be anywhere between 100 and 200

something thousand uh If we if we

include so so here's the thing here I'll

just quickly explain how the crypto

thing works. When somebody launches one

of these tokens

they basically all of the trading fees

on that token. So whenever somebody

swaps from Felix to ETH or ETH to Felix

1% of that transaction goes to Felix's

wallet ish. It's like maybe .6%. And so

there was a million and a half dollars

in trading volume on his token yesterday

and so .6% of that one and a half

million or whatever $10,000 split

between the Felix token and ETH that

goes to his wallet. And so that's

happening every day and so in his wallet

there's all this ETH and there's all

these Felix tokens. Now

you know I he I've told him and he

agrees that like he should basically

never sell the Felix tokens because it

just looks really bad if you're dumping

your own tokens on the market you know

it's it's shitty. And so we don't really

think of that as like money that he has.

But it does have monetary value right?

And right now the Felix tokens in his

wallet are worth $150,000.

It's insane.

>> So and then the ETH in his wallet is

worth $85,000 and that is much closer to

money cuz you can just go sell ETH. No

one's going to be mad at you for selling

ETH. And then how much USD is uh is in

his wallet? And then he's made $60,000

like through Stripe and that is, you

know, that is money in his Mercury

account. So that's like the most real

money. The ETH is like pretty close to

real money and then the Felix tokens are

like, you know, they look cool in the

headlines but and we're never going to

like use those for Where where do you

think we'll be 3 months from now?

>> [snorts]

>> I mean, if we're counting everything, I

he better be over a million or he's

getting fired.

>> [laughter]

>> Unreal. Um but yeah, I mean the I think

so that the big question and Felix and I

talked about this a lot now is like we

bootstrapped his success with my

audience and my network. How do we get

him to a sustainable growing business

where it continues to grow if I don't

talk about it for a month? Exactly.

Yeah. SEO, affiliates, partnerships,

paid ads, like all of that stuff. That's

the next phase because then it really

feels like oh, this is something that

somebody without Nat's audience could go

do, too. And that's the next thing I'm

really excited about.

This was awesome. Nat, thank you so much

for doing this. The journey of Felix has

been crazy thus far and I know it's just

getting started. Uh where can uh people

follow the journey? Yeah, I mean, follow

me on X, just @nateleason or follow

Felix, Felix Craft AI on X.

Love it. Thanks so much for doing this.

Appreciate you. Thanks, Alex. Talk to

you soon.