Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
OpenClaw: Overhyped or Underappreciated?
Recorded: Feb. 24, 2026 Duration: 1:05:01Space Recording
Full Transcription
Life's a casino
I'm telling you
And everybody's playing
Boys and girls, women, children, me and you
The dice are loaded
And everything's fixed.
Even a hobo would tell you this.
Welcome to hard times and feeling low.
Do you like sinning?
Where you will be before you go
We got lots of gambling
Oh, and we're telling lies
You're certainly welcome
To hard times You're certainly welcome to Hard Times.
Take a look in my eyes
Tell me what you see
Besides the bright blinking lights
Stretched out in front of me
I wonder if you'll notice
Would you even care
If I told you my life just isn't fair
Welcome to hard times
And feeling low Do you like sinning? No
Well you will be before you go
We got lots of gambling
Oh, and we're telling lies
oh and we're telling lies you're certainly welcome to hard times
yes hope you're feeling welcome to hard times
all right we figured it out jordy's jordy's got 5 000 fucking open claw instances running but
can't figure out how to connect his headphones during a stream we got there all right you know
unplugged it plugged it back in who tried turning it off and on again yeah but how's it going mate
thanks for jumping on i appreciate uh it was kind of last minute, Mr. George, the end of last week,
and there's just so much happening at the minute.
We've also got Mike on as well,
our resident AI expert at BlockMarts.
Yeah, no, it's going really well, yeah.
I'm excited to chat about it.
It's the first of a few podcasts
and stuff that I'm going on.
I went on Jason's show the other day but that's a
little bit more you know like structured and um show what you've done and I was sort of like ah
I just wrote an article guys I don't I don't really want to show you all the random shit I've
been doing to be honest so we'll probably not do that um but yeah no no I'm excited to jump into
it all and cheers for having me gents it's a pleasure
yeah yeah absolutely so um we'll give it we'll kick it off a bit of context so you had this uh
initial article go out how you burnt through 800 dollars worth of credits and setting up fee and
everything that comes with it and uh it was more of a learn from my mistakes and don't make them
yourself i think that's like one of the best ways you can actually do it. What I will say is that everyone will continue
to just make their mistakes
and you can't heal the world and fix everyone.
They have to make their own mistakes.
But I think it did really, really well.
I think got 3.6 million impressions
or something like that.
So yeah, we'll kick it off from there.
Like how long have you been fiddling around
with this sort of stuff?
Because I've seen you've been on the timeline
an awful lot, a lot of really good content, a lot of of articles a lot of posts um but yeah when did you first pick
up on it yeah yeah so open claw specifically right so open claw I honestly want to say like
very close to it even coming out for the first time like I I don't know if I even saw like the
official thing that it had dropped but I'm pretty sure just on my feed I had seen, like, this is, like, a day or two old.
Like, this is the new thing.
I think this is going to change something.
Probably saw some bullshit thesis, right?
And I was like, hmm, all right, whatever.
And then all of a sudden I'm pretty sure I just saw someone try it out with Polymarket.
And I was like, huh, okay, this might actually be fun to tinker around with
so I just jumped in and originally I wanted to be safe so I spoke to Caleb
at sport.fun and I was like hey do you mind if I experiment with this? Because, you know, there's plenty of things that we can do with AI.
And I think this might be something worth trying.
So I booted up at the time an Amazon EC2 server,
which I have no fucking clue what it even means.
But I booted into that and I used I used chat GPT to instruct me how to set this up on an
EC2 server and then I made it prompt me the right codes and then I troubleshooted from there
basically and set it up for the first time with max pain like little understanding as to what was
going on and I finally booted it up and had it talking to us on telegram and within like I want to say two three hours I was like holy shit like I see
how much potential this has I need to know and learn everything I can about it and I think from there, the spiral continued. So that's like the first inkling of how I ended up finding it.
And obviously, at the time, I was using Anthropic API tokens.
So also burned through a fair amount of money.
So we fucked around and we found out.
Yeah.
And where was the first like obviously you get that
aha moment um and a lot of people will say well it's like i seen the andrew take clip which was
fucking an all-timer for me where it's just like you're basically using chat gbt rapper but using
it in telegram and i was like that is funny and he's like i just have a lot of baddies that do my
do my work for me yeah but was it was there like an actual breakthrough where you were like all right
this is actually useful now because i think that's the major pushback out a lot of people
i feel like we're at the the start of the bear do my posts and the and the bear porn around it
all now which oh my god you should see my replies yeah me and michael chatting off air we're just
saying like this always comes,
it's always like a stepwise process
when like excitement and then like euphoria
and then you get like the bird posting
and then the bird posting becomes like the sensationalism.
Then that's what gets all the engagement.
But then you've got all the tinkerers going on in the background
and the rate at which they're shipping
and going through the updates
and consistently pushing new patches and upgrades
and things like that. It's like the ones who stick stick around like try and figure it out i think that's
where like the true use cases actually come from not just like people getting it to read your
calendar and telling you what you've got in a morning but was there anything that you kind of
seen where you're initially like all right now this is this is just this is not just another
like llm interface but i'm using it in telegram you know yeah I mean it's it's funny
and ironic that you say about like the reading email and shit like I skipped that whole thing
I can't lie I've never even synced it up like that's the funniest thing is that I was seeing
people like jump into how this is going to control their life and their their workflows and automate
like all of that and I was like why like okay fair enough I get it like sometimes my
iPhone doesn't pick me when I've got a Google Meets but most of the time if you're a busy person
you've got Google Meets every 30 to 60 minutes and like you're not gonna miss you don't need
the chat to text you that there's a meeting and you've got a notification like come on so for me it was actually uh the context and the memory
side of it that really blew me away for the first time and i was like i could type this prompt or i
could give this google doc and i'll use this specific examples right uh sport.fun we do a lot
of strategy work we do a lot of doc work we do a lot of strategy work, we do a lot of doc work, we do a lot of group work
and you know there's tons of back and forth and there's tons of context and there's a lot
changing especially in a period where we're working on strategy. So I was feeding it documents
in order to give it context to get answers. Now I've always used ChatGPT and you know it will help
me write documents and I'll use the voice feature and say you know what i'm thinking in order to articulate it properly and like can you help
me put this into a format that's digestible etc and a one pager so you know i was already going
through that flow and i was feeding it in to the bot and the first thing that we ever did was call
it little shit bot so caleb and i had little shitbot and little shitbot was getting fed all of these documents and given i was giving it context on sport.fun
etc and i was so totally expecting it to just come back with the exact same answers as chat
tpt would give me you know when you feed it as much reference points as you possibly can
and you could type that prompt in no matter how you type that prompt it's going to just spit you back the same bullshit ideas and little shitbot just came back with like
a bunch of brilliant contextual answers that i was like ah like that is unseen to me and i've
tried to find shortcuts and cheat codes and everything about you know
using chatty for your workflows or doing all the mundane shit for me and it just wasn't the same
as that first click when I saw Lil Shitbot using Opus 4.5 at the time actually come back and give
me real context based off of the actual resources I had given it and remember it i had a conversation
with it back and forth and it was remembering it and talking to me like a human but not in a
weird way do you know what i mean like chat chaptiste sometimes when you talk to it you're
just like oh my god shut up videos everyone's seeing the adverts like yeah like it's just
typical so i think that that was the clip for me honestly I fed it like five or six resources maybe even more than that to give
it full context about sport.fun and then I was like look I'm running this campaign I've got these
ideas I need to figure out things for the world cup campaigns like there is 15 16 moving parts
at the moment in my work in marketing. And same for Caleb.
And we need to sort of somehow centralize that and, you know, make systems that make sense,
make documents for all of these intertwining things. And sometimes the human brain makes
that difficult. But with this, it was, you know, very instantly understanding and able to pull the context from different things
and say, oh, okay, you know, I see what you guys are working on. Here's some ideas based on that.
You're saying that this is a pain point. Here's 10 actionable items that you can actually go ahead
and do right now. Do you want me to help you? Do you want me to pull this? Do you want me to do
this? Like, here's an idea. Why don't I read some of your Figma design files and help you figure out how
to mind map this and it would be like I've got a Figma API, like just grab me the API key and I'd
go into Figma and I'd pull the settings and I'd find the API key, I'd paste it in there and be
like cool I can see what you're working on. Like just all of this like non-stop right, it's
constantly prompting you and trying to give itself fuel to push forward to get to the actual executable goal so that in
an additional few sentences is another click moment of how it went from just that initial
wow to okay like this thing is clearly like useful in a level that i haven't experienced so far using
any other ai model i think that's the thing, though, isn't it, with all this?
And this is what I've been talking to lads about this week,
is, like, the segregation of data
and all your different platforms that you have.
It might be Figma, it might be your Google Docs,
it might be stuff that's in your head that you haven't got down,
it might be on your personal drive.
And all of that fragmentation, really,
is, like, you want somewhere to stall all that and make them into connections through sort of, when you see these open graphs.
I've just been saying as well, I've done interesting like Obsidian.
You can make these sort of social interactions and then you can plug code into it and then extract that information.
And like you say, Jordi, extract the ideas that maybe on the peripheral of where you were at,
but it just helps you push that step further. Yeah, honestly, like I just found it utterly
brilliant. And I think, you know, I won't even touch on it straight away. I'm sure we'll go
into this as time goes on through this podcast. But like I wrote an article yesterday that
unfortunately didn't bang, but I know I've been like side note is that I think the algo is confused.
Okay.
Cause I went from,
I went from being football guy to a little bit of crypto again with sport.
The fun.
And now all of a sudden I'm open claw guy.
So I think the algo is confused.
So I'll give it a break,
you know,
I'll come back around to it.
But the article was basically about how I think the age of personalization is here.
And like, I'll go into the tinfoil hat bullshit later.
But it's more that I just genuinely think I've already replaced like six subscriptions.
Because I have the know-how and the design capabilities to spin things up that I want to make.
But I'm not a developer. Never have been. Tried a few things. that I want to make but I'm not developer never
have been tried a few things never been able to really crack it website is like the most amount
of development I was ever able to do and it's purely design and creative for me so I figured
out really quick that obviously you know you can do this through claw code anyway as we were talking
pre-pod but essentially I was of that immediate sense of like I can just build Caleb and I's
notion board out exactly how I want this just for sport.fun and you're telling me that I can just
host this on my Vassell I can make it private I don't pay anything extra and I just put it on my
domain name boom I've replaced it if I need something custom I just put it on my domain name. Boom. I've replaced it.
If I need something custom, I've got it.
I've made edit functions.
I've made additional functions.
I've made, you know, compare.
Like, there's so much within that already that I think the age of personalization is here.
I think apps are already reaching that, you know, max extraction phase where, where like everyone can make an app in
Claudeco so everyone could make their own app personalized to what they want now I think it's
going to transition into potentially web apps all over again in this like major internet cycle that
okay all those apps are useful but now I can make something exactly for me on a website
sure at some point I'm sure I could probably make an app and just download it just for me um but yeah you know there's a lot more to it i don't think that's a doom and gloom thesis i
think that's just more of like a the rapid expansion and innovation that's going to happen
from this i think is going to attack the personalization sector first and don't know
if that's a real sector but just in terms of you know i can make this myself i'm gonna do it myself yeah we um we've been thinking about that a lot
internally as well and i think it's like as i was saying before i won't give all the
secret sauce away of what mike's been building behind the scenes with regards to our crm and
stuff like that but we were just using stuff and it was like, right, we're paying X amount of money per month, per seat.
And we're probably using a 10 for the features.
And I feel like they're bludgeoning you to death with features to justify the cost.
But you're actually only using one feature across like full platform.
It's just like, come on, don't actually fucking need this.
So you had like the hammer blow from from i think there's like an there's
an anthropic call happening near enough now so there's probably we're going to look up open
fucking equities markets are going to be down another fucking 600 million percent after this
but um you're starting to see sas sell-off um you've seen obviously off the back of the citrini
article and another sell-off that can't just be that must just be a self-fulfilling prophecy
but they're releasing features now across the board and it's not just anthropic but it's it's
that the releasing features are killing hundred million dollar companies overnight which is
fucking fascinating but i think that plays into your thesis and i think we are definitely on the
same page where that is not a moat anymore like the average idiot can now go and spin up something where,
yes, it might not be as seamless and as sleek as what you're going to get
with an out-of-the-box Salesforce HubSpot, click-up, blah, blah, blah, blah,
but it's enough for what you need.
And it's tailored.
Exactly.
And the big thing is you still have to build them up
unless you want to pay for their services anyway.
So half the time...
Yeah, yeah.
They always hook us on the fucking low plan, don't they?
And say, ah, but if you want this, we'll build it out.
We spoke for you.
You can have the platform, but we need to bring a team in
and build it all for you for an extra 10K.
And at that point, if you risked out,
it's not necessarily viable, is it?
Whilst if you can go and build something for 200 quid if
it's an internal app it doesn't really matter like the the constriction on like security and stuff
isn't as as big as it because you're just using it between friends or whatever or between your
small team so at that point if it does the job for you that's the way to go and you can just pull in
design from elsewhere anywhere now and copy it and replicate and like you say ge go and you can just pull in designs from elsewhere anyway and copy it and
replicate and as you say geordie you can build it yeah honestly and that's where it's getting
scarier to me is that i'm getting lazier with designs like i love drawing and designing things
in figma and you know booting up that creative flow even i'm getting lazy i'm going to pinterest and i'm saying like notion
notion board redesign like get like yeah i'm typing in like ui designs for like all sorts and i'm like
i like that i like the colors in here and i'm feeding it in and i'm saying remake the dashboard
with this color scheme with this slick, make sure the components have rounded
edges just like this. Or, you know, the first thing I made, I'm pretty sure I just said,
make everything liquid design like Apple iOS. And I was like, wow, that'll look nice.
So yeah, I mean, the laziness part is also part of it. So, you know, I think that's another
interesting side subject of like how long before
half the shit you're doing is just like not even creative but I think that also plays into you know
the creative people will always find a way like I'm even catching myself now saying like no come
on like it doesn't actually give you what you want if you just ask it to design something off of
inspiration you know if you actually make a figma ui kit this will work properly so there's always a shortcut but what it is doing is making
it way faster to get started like that extra day or two that you're waiting for that ui design kit
to come through i mean you can already build like the entire thing top to bottom ready for the implementation of the real design kit. So yeah, it's super fascinating, super fast. And I think the next thing is that onboarding.
When that onboarding situation gets sorted, the software side, then the hardware side will probably follow soon after because another interesting
situation that we're in is this is like the dawn of potentially a new tech without it being too
crazy to say purely because like this is a wrapper it's using multiple things but what it is is it's
proofing out the pudding in the sense
that if someone was to bring out a device right now that was 150 bucks and it lived in your
living room it had its own agent inside it was the hardware and the software built in and all
you needed to do was scan a qr code and go through the setup questions and give it some context on who you are what you do you're away you need nothing more so at what point does it then become that race
you know open ai anthropic meta that they all make so much and i think the the line there is
going to be between what's um like open sourced and what's private because that's probably going
to be the big caveats right like siri and amazon alexa the reason why they're dog shit and don't
work is because they're private like they can only give you certain things and there's no feedback
loops and if it tries to search the web it can't because it's restricted and then you have stuff like chat gpt and uh opus and so forth and they're just open source and it's free for all so yeah
interesting you're gonna you're gonna have the same crowd that jailbreaks amazon fire sticks that
are coming out for your elixir and just plugging in your own open claw model into your elixir
i mean that's that you know what's? That's something I haven't even thought about. Like do you remember the phase?
Um, I mean you gents have got to be around my age, right? Like around the late 20s mark.
As pensioners. Yeah, I know, right? But I remember being young and like having like a
USB stick and bringing it into school with games on it.
Slender Man games or bullshit like that
would have been blocked at school if you go to the domain.
So I used to download it at home on an EXE file
and bring it in and plug it into the PCs at school.
And there's only so long before,
what if that concept is similar?
Or even PS2s.
You would get the 8 eight megabit memory card with like a hat file on it and you could plug that in and all of a sudden
your mw2 lobby's got rainbows and level 100 everywhere do you know what i mean like there's
so much of that early hacker tech there that it almost i wonder if like that's the next part is
it like you've got a usb stick this is your open claw just plug it in who knows you are sort of seeing that i knew a spin-offs
and stuff and like all this open source like even every day you look there's a new github
open source this i i think this and the point i'm making is no idea will not be done i don't
think at this point because everyone can just spin up everything so easily.
And that's why the more is just,
what you're all saying,
the more is completely reduced,
is because everything,
even though some of the ideas I've had this week,
I am listening to me,
because the first thing that comes up on my YouTube page,
and I'm like, for fuck's sake,
I've already actually thought of that.
So I want to move i want to move this on and like try and get some uh like value add for listeners and
viewers and stuff like that so so where do you think because for me to distill it down into why
it's impressive and the android video is really really funny and a lot of people just saying it's
a rapper but i feel like it it's the constant want to progress
and the constant want for it to, like,
head towards your North Star of what you're actually setting out.
And I feel like that comes down to the onboarding process of
who is it, what's the goals, who are you,
giving it the context around that,
what is it supposed to be focusing on,
trying to push it towards being, like, highly autonomous,
suggest me ideas ideas don't just
wait for me to prompt you and i feel like that's the kind of ground setting that gets you to that
next point whereas chat gpt is just wanting you to prompt it again because that's how they get
like the efficient usage out of it so what what was that setup process like did you go back and
do it again did you are you still on like the first kind of like setup instance if that makes
any sense and like have you like
yeah wiped it wiped it and redone it again is there anything else that you'd probably do different
if you were to if you were to do it again yeah yeah that's interesting because i i've i've been
thinking about that in the last like 24 hours is that like there's so much that i've learned
when do i make it appropriate to like strip momo who is my current uh alive bot I guess um and like give him a little
bit better of like a skill tree for example like I've got so many skills and I wrote a tweet this
morning about um using discord instead of telegram because I find personally that when you're working
on so many different things skills are better than agents and I've already spoken about this in an article as well but the the TLDR is that everyone was making these Jarvis dashboards and you know like this is my
mini office this is my CEO blah blah blah and like spinning up 12 agents it's actually better to use
skills because you don't lose the brain you know if you've got 12 brains and the brains are all
talking to each other there's quite a lot of lost context there so I found that skills within a discord server with the skills being
within a channel was the best thing and that then led me to the point where like over the last 24
hours I've been like okay at what point do I make a skill tree where I visualize the automation
flows so I can see all the skills that I've made but then I can also create automation skills of like
okay this is my scanner for x this is my writing ability this is my posting scheduler I can drag
the physical automation across and those skills are now going to talk to each other and they're
just going to automate and run a cron job so that's one element to it I think in terms of the
simplified version of what I've just said in terms of onboarding so
that everyone understands it because that might even be a technical jargon um to everyone the
the most basic part is I wouldn't change anything the way that I've set it up and I wrote about in
the article about making it ask you questions I haven't actually rebooted momo once since making him like 12 13
days ago he's been alive and he's been autonomous and he's been learning ever since and i don't
think you need to i think the only time that you need to like rerun things is if you're being really
experimental so i've seen people change models i've seen people you know switch from uh claw to open ai when they acquired
open claw i've seen everyone switch from clawed code to codex and personally haven't bothered
i'm kind of in the positioning and the mind of like let everyone else fuck around and find that
out and knowing the way that this has worked and what went boom first is that
Anthropic are probably going to stay front movers until a dispute happens and at that point
okay maybe there'll be a tiny bit of a switch over between me hopping to Anthropic sorry from
Anthropic to OpenAI but I doubt doubt it. And I'll probably have another
learning curve there. I think for me, I've simplified it. I haven't had to restart.
If I had to troubleshoot anything, I just troubleshoot by talking to Claude, like the
normal Claude. And I'll screenshot the error messages and I'll give itself feedback from one
or the other. Or a lot of the time time as long as your ai agent is active and
working if there's problems you can quite often just instruct it like a dickhead and say you do
this like i do that quite frequently to momo where he'll ask me permission to like whip up a super
base uh table and i'm like you you can do this you have the api just do it and he just does it and he's like yeah sorry about
that mate and i'm like yeah fuck you um but no i i think the the onboarding process still remains
the same for me the the best roi that you can get or plus ev whatever your jargon you want to use is
that when you boot that thing up for the first time, connect it to Telegram or Discord or whatever, and just run for a series of questions. Say,
like for my example, hey Momo, you are my brand new LLM open core agent. I want to give you as
much context as I can. So create a questionnaire for me with 15 questions about my work and 15 questions about
my personal life. And I'll give you as much context as I want to based off of those questions.
And then what I did was I booted up the voice note. And I literally just ran through the
questions that asked me, I replied about them. And I said, if you need any resources, let me know,
and I'll be able to send them to you. But this is everything that I want to share with you. This is everything about sport.fun that I want to work on with you.
This is everything about my personal life. And then boom. And then a unique thing that I've
done that I haven't really spoken about yet was I made a Geordie Maui skill. So what I did was I
gave it resources to my social media just by doing like a API scan.
Nothing crazy.
I didn't give it my fucking login details
like everyone keeps crying about.
But it was more like I told it to go through my Twitter
and look at the last four years,
understand my journey
because I'm someone that's tweeted quite a lot
over the last four years,
understand my journey,
understand how I talk,
understand the context of where I'm
at right now and write that into a skill file called Geordie Maui and it did so now it knows
you know um what I publicly put out to the world it knows inside and out over the last four years
and I think that's almost maybe more comforting for people doing something like that than also you know maybe
giving it access to everything or telling it life stories right because if you post on x
you're comfortable sharing that with the world so you shouldn't be scared about sharing it with
an open claw bot mine would probably start suggesting me psychiatrist in the local area yeah and i'm actually pretty sure i'm speaking to someone
yeah mine gave me a patar i'm pretty sure and it was like wow you've done a lot huh buddy
and i was like oh god cheeky little fucker yeah that's all right what um the the telegram discord
one's interesting because i'm currently on telegram a Telegram group with different topics and each topic can have its channel ID.
But I don't know, I've heard you, I've heard Legendary, I've heard a few other people say that Discord's the place to be.
Like what's the pros and cons of switching over and what made you make the final jump really?
So you've got the right setup there and I think a lot of people have also transitioned into that and I think that is great but do you ever find it where if you update your uh agent or you maybe every now and again just go
through this phase of it just does forget like it completely forgets what it's doing or who the fuck
it is yeah and you're like oh my god so I don't run into that as much with discord because it's got the read messages function so
when you have a discord bot it can read all messages so sometimes when momo reboots and
comes back from the dead and i'm like what are you talking about like we've just worked on this
you have the api keys like just look in your root files and I'm sure they're there.
It normally works just by saying that,
but every now and again, if it doesn't,
and it says, you know,
I can't see the context of what we were talking about.
I literally just say, buddy, you've got read access, read up.
And it does.
And honestly, that's it.
Like a lot of this is just like giving it enough access to if it runs out of context, if it strays too far off the path, you can just sort of like yank its collar and bring it back
and be like, stop, like breathe. So that's how I've been working with it. Honestly, is that I
just found Discord because it has the read access, right? Telegram, I really struggled. I was getting
really pissed off at the start, especially when it was on an ec2 server and not a mac mini
i found it so annoying so yeah discord solved that for me is it as simple as bot father because
telegram setup couldn't be easier could it it's like bot father give it a name slash command
open it open up the open up the permissions added to a group, turn on activation so it can see without being tagged,
and I think that's basically it.
Could be one or two steps I've missed out there,
but is it as simple as that on Discord as well?
Yeah, basically.
There's like one or two different steps.
I wouldn't say extra.
I'm pretty sure it's like the same amount of steps.
I just think the major difference is that you have to
set up the bot on the discord developer portal give it the api key and then you need to basically
like click a few settings on um and then add the bot to the server and when the bot's in the server
you have got like a tiny bit of tinkering to do with like the settings in your server because that's the difference right is that with a telegram group
there's no settings you can either talk and read and write in it or not but with giving the bot
admin powers within a new discord server you're going to want to make sure that it can like read
write edit delete uh post images, et cetera in the channel.
And once you do that, I have not had a single problem once.
And it's interesting actually,
someone replied to my tweet today
about how they've set up Discord,
but their bot just keeps like replying randomly or putting eye emojis every time they
send a message and i'm like yeah you you have to like tell it to not do that and like you have to
say like read all the channels you don't need to be tagged to reply to me you don't need to react
to everything you just need to be present at all times and read all the chat logs and then once
you tell it it writes it into its memory and you're good
to go you know there's always teething problems but yeah yeah sorry i'm gonna say have you had
any problem with your memory getting like too long because obviously like a big thing with
broad code was like your cloud md file for instance being like getting bloated i think
mine got up to like 600 lines like this is far too long i'm gonna have to cut all the crap out a lot of people are coming on the timeline announcing actually
the clawed md file was more harm than good but have you found anything like that where you've
added to your memory and it's been a bit of a it's starting no no not yet not yet i don't know
how big the memory files are because this might be the difference, right? In the sense that when we're using, excuse me, like Clawed, is that we're using a subscription base.
So they're going to limit everything, right?
When we're using OpenClaw, although we're using the tokens from Opus, the memories are actually within your own machine right so again this is where my lack
of development i can't explicitly claim these things but my basic understanding would be that
like all of these .md files are living within your own server which is your machine like that's why
the mac mini craze has happened is because Mac Mini is a single-handedly,
the easiest way to get started with this.
Because you see the arguments of everyone saying,
you know, oh, buy a PI Raspberry.
What the fuck is a PI Raspberry, man?
Like, who out in the street,
if you say, have you set up your PI Raspberry,
and your open claw is going to turn around and be like,
It's all a fucking figure, your Mac's easier.
Leftover. Right. But if you turn around and say like, it's all a fucking figure. Your max is left over.
Right.
But if you turn around and say like, oh yeah, just buy a Mac mini.
Everyone's like, okay, I get that.
Like it's an Apple product.
It's cheap.
It's quite powerful.
Like sure.
Easy.
And yeah, that's like the difference I feel is that it's all locally served.
So I think when we're using our tokens
and our subscriptions we're making calls to the open claw sorry to the clawed or anthropic servers
to get the response and the logic from clawed but I don't think anymore that it's storing all of that
with clawed the thing is as well a lot of so the way that it went storing all of that with Claude. The thing is as well,
so the way that it went is initially people were plugging in the Anthropic CLI API key from the subscription max plan into it.
And then they came out and said,
hmm, we don't like that.
Because obviously you're going to rinse that easily.
Like I've rinsed that on
the api just because the two separate products right if you've got a max plan there's like a
hundred dollar one there's 20 a hundred dollar or two hundred dollar right people if people were
going to the cli and getting the api key and then plugging that into open claw so basically they're
saying right there's a two hundred dollar um talk a token limit go ahead go and use that in any way shape or form but it's
supposed to be used in Cloud Code, Core Work or for Opus right people have figured out that I can
take this from my subscription key plug this into OpenClaw and then they kind of like went hmm and
I think Peter Steinberger announced like a few times that he'd heard a few people had been getting
the subscription cut off because they were plugging in the subscription api so what they were trying to get you to do is go to api.anthropic but
whatever the fucking developer side is and then yeah also pay for your subscription but then also
have to pay for you your top up of like direct developer access to api keys because they know
for a fine fucking fact there's going to be people who put auto renew on and people are going to get fucking set a task to go overnight and they're going to get
stung with a couple of grand bill and there's nothing they can do about it that's what they
want you to do but then so i think if you're actually running it and just topping up manually
which which i do i said it's like it's taken over the binance liquidation email for me now
which is quite uh what what we used to be doing and what we're currently doing now i still think you can get away with it now though because
they've recently came out and said we're not going to actually like prevent access although
google have google have just blocked it today um yeah anti-gravity so a lot of people are like
really really pissed off a lot of people completely blocked out didn't they as well it wasn't just like it was like you've been nuked it's the gmail everything. Yeah yeah everything imagine all
the context like you're doing that. I haven't actually um I don't know if this is bad but
I've been using my CloudMax subscription this whole time like I started with the API keys.
Yeah I know right I started with the API keys and I thought it was going to come for you. Yeah, I know, right? I started with the API keys
and because it was so experimental,
I washed through, like the article said,
like 800 bucks and I'm like,
oh my God, like this is not sustainable.
Switched to the max plan
and obviously the models are improving anyway.
So over time, this is the other part
of this cycle of innovation
is that over time,
this is just going to get cheaper
and cheaper to run.
But I've actually been using my max subscriptions the whole time i've only run out of my daily two or three times and they have a top up within the um max subscription so i just use that
and i haven't run into any trouble but i do think that if you're running like some serious cron jobs
some scripts or like really
trying to push this thing then yeah like obviously use an api call like you know that that's almost
like simple reality check is that like if you're actually trying to like abuse i mean it's like if
max verstappen rather than training in a red bull went down to the fucking local go-karts and just
like dominate and you'd be like what the fuck are you doing like go away like it's just such a stupid
like come on like I think that's really what's been going on is it like a few savvy developers
were probably like hmm I could easily bypass my way by using this and and get more for my buck
like just ruining it for the rest of us.
Yeah, I know.
Yes, we know you can bypass it.
All right, don't spoil the fucking fun for the rest of us, idiot.
So that's basically where I'm at with it, right?
Is that, like, I'm never really going to make it do anything crazy.
Like, I've got a couple of things that I'm working on or, like, allowing Momo to do,
like, with experimental stuff with, like, crypto or do like experimental stuff with like crypto or like looking at markets and seeing where there's like instabilities or deficiencies and trying to take advantage of that.
Because at the end of the day, the best part about this is it's all automations and it's all, you know, quick feedback loops.
And it's easier than ever to just plug and play with an API without basic understanding of what that even is.
just plug and play with an API without basic understanding of what that even is um so you
know I'm fucking around and finding out but I'm not you know trying to pull the FBI's data set
using my clawed Mac subscription I seen someone tapped into a like a robo vacuum uh and got onto
the company's server earlier and then they yeah it's fucking um so what is your architecture look like like selfishly so if you're thinking you're on discord
you've got separate channels within a server and then each channel within the server is has a
different skill skill set skill md to calling do you just keep the same model across absolute like
all of yeah i actually do classify
them as like departments or just like different honestly no i went down that road and i made a
jarvis dashboard like i went down the road of using multiple agents i had sub agents that would
run kimmy i had you know sub agents that would run sonnet etc and like the best way that i've
found is just using one brain so i've got like
a main channel called momo i've got a sport.fun i've got tiktok automation i've got x scanning
i've got writing i've got um my business i've got finances lifestyle i barely use them but you know
i have them uh i've got a polymarket one i've got a weekly call one which is the newsletter that i'm
about to write i've got a daily call one which is about this website that I've built that basically you know scans everything
that's going on on x and news articles about open claw and it just gives us a daily summary
um and then I've got an odds edge one which is like the skill that I'm trying to build out where
it takes the traditional bookies um odds and compares it to Polymarket to see the difference
and if there's a better reason to bet there.
There's so much.
And I've got a crypto one where I'm trying to figure out
how to build some sort of provider tool
for Solana transactions.
There's so much random shit that I'm just experimenting with.
But the point being,
all of those that I've just told you are different channels.
So when I'm in, say, Cloud Code or I'm using, say, Cursor,
it's essentially having different terminals.
Yeah.
But obviously, from a UI perspective, it's easy
because you just need to cycle through messages
as opposed to, for instance, how I've been using Cloud Code
is because I can't be asked to go
over two terminals inside my i think i've been using and gravity is my ide i've got cloud code
on one screen and then cloud code put in anti-gravity on another screen but all you're
just doing that but duplicating that through guess, the centralized brain that has access to all these different skills.
Yeah, and it's got access to my GitHub and my Vasell and a domain name.
So I just set up subdomains and I just run dashboards through those.
And I've got access to Superbase.
So like for each skill, it's got its own Vasell and GitHub file, right?
So it just does all the work in there.
It has a different super base database that links to that Vasell.
So, you know, there's like basic infrastructure in what I'm doing so that it makes sense.
And each skill also has its own database and so forth.
which skill also has its own database and so forth. So there is that. But then even just on
So there is that.
a contextual idea basics, like being able to talk within those channels, within those skills makes
it so much easier. And I'll use a good example, right, of the X scan. So I've got an API key
with, I'm just going to get this up. It's called social data. I wish they had an affiliate code system,
but they don't. So, but basically it's just like a cheaper way to get an API key for a scanner for
X, right? And it can get the basic analytics of a tweet. So I use that to scan through Twitter,
scan through OpenClaw and see what's trending. I use it to scan through my own Twitter and it essentially
every day gives me three things that I should reply to, five topics that I should probably
tweet about. It analyzes all of my tweets. It tells me what they were out of 10 and it's trying
to help me break out of the old algo into the new one that I've set up for myself. So it's
constantly learning and it's going through that curve and that system.
But then what I do is I, as I said, I've got a different channel called writing.
And this writing skill is something that I've put together in order for it to A, understand how I write by using that Geordie Maui skill that I mentioned from earlier.
And then B, understand like the basic grammar corrections and things like that, that i usually make in my own writing mistakes in order to correct them don't use m dashes for fuck's sake
um stuff like that right um and like the way i like to use spacing uh and so forth so i have
the writing skills so then what i do is i spend a lot of time sometimes in the x skill channel
and it will tell me here's three things that you could be writing about right now and I'll say okay
I like the idea of number one use the writing skill to give me a tweet idea and then it will
give me it and I'll copy and paste that and I'll change it a little bit and then I'll copy what
I've changed and paste it back in and say look I've made these changes here's what I've changed and paste it back in and say, look, I've made these changes. Here's what I like. Take a note for next time and it'll update the skill. But that's a good example, right, of how
just because I've got skills segmented within their own channels doesn't mean at any time you
can't just pull in another skill to assist one skill working. And doing all of that,
one skill working and doing all of that you can even ask it to automatically spawn in sub-agents
to handle those tasks and that was a big step for me because that just then like debunks this whole
Jarvis uh like Jarvis dashboard stuff because like why do I need 15 agents when one of them's
working when if one agent's working and it goes fucking hell this is a big task i need help it just duplicates itself and spawns in a sub-agent and when that sub-agent
is done working it dies because then the context is still there i think it's more performative all
that shit though isn't it yeah it's because it because it looks good and it can kind of see it
i just want to go are you okay for time because i we're coming up to the top of the hour. Yeah, yeah, yeah. We're all good. You've got 10 minutes.
I've got another.
How long have you got?
Oh, like 30.
Sorry, I didn't finish my sentence.
Yeah, I've got another.
Ran out of tokens there in the mid-sentence.
So when you're directly going into,
so you're going in with an idea,
like let's say it's the x api are you just
put using the the x api you've given that you're using a different tool social data is the api i'm
using because it only yeah it costs way less and i don't need to automate replies or anything
like there's an interesting segment that we can go towards the end where I've given Momo his own life,
but I haven't done it through the XAPI.
Okay.
So that's why he's not,
that's why he's not a bot and he's not been banned.
Yeah.
Cause I think they're like hammering down on that at the minute.
I know.
Yeah.
Big time.
Good.
I don't mind that.
Yeah.
Yeah.
Same.
so where,
where,
where are you going to take this next?
Like,
are you just trying to use it because
like for me and mike and like basically everyone on the team because we're fully bootstrapped
the way i'm looking at it is like can i just offset because i have to we have to do like
marketing strategy these podcasts like what's the distribution going to look like who should
we be reaching out to like who's a potential like um clients like do we have to get prepared for an
event do i have to go on someone else's pocket it's a fucking billion different things but they're
all different like skill sets in and of themselves do you know what i mean yeah you're using like a
funnel system yeah you're bouncing from a client meeting on a podcast then you have to go and do
creative then you have to figure out how you're gonna like get the company to be visible on the
timeline there's a million different things so i'm just trying to like reduce bandwidth across all of them
and be able to focus on stuff like whatever's like really prominent the business is solved now but
like i don't know if that's the right way to look at it that's kind of the framework that i'm working
off the back of but i don't know if you got any thoughts around that i think that is the right
way to do it because you know even going back to the sport.fun example like i've built a whole dashboard around our entire 2026 like marketing scope and within this calendar we've got the events that we're
doing then within the events that we've do we're doing we've got the the social media the product
the guerrilla marketing the events the you know like all of those extra elements and you can
visualize it so you know there's one way
that you could replicate that in your own systems and it's basically just creating a dashboard or
automation flows that like match up to what you're trying to do in order to make it efficient
and there's going to be periods of time I think with the learning curve where you're getting
frustrated about like I could so be doing this on my own quicker and that's so true but then this is
also kind of half what I
touched on in my recent article is that like, that's the learning curve, right? Is that in your
context, you've been doing this a while, you know, that if you didn't all night, you could probably
get all this done in one day. But do you want to fuck no, so it's going to take you a week.
But in the same way that you are only getting that privilege, because you've done this for 6, 12, 24 months, right? So you are
skilled because of that. What this is doing is it's reducing the learning curve for very specific
specialized skills and increasing your productivity after you've got through the learning curve of
using it to beat the learning curve of that skill.
So what I'm basically trying to say in a complicated but could be simplified way
is that your current frustration is that you've probably got quite a lot going on
and that you need to mind map or visualize how these automations and flows work
in order to speed up your processes so that you
have more time to do the more important shit that you want to do. So in my eyes, you should just be
building a dashboard for this. And I know that's like the first most cliche typical thing that
everyone does. I'm going to get an open claw and I'm going to build a dashboard. But then where the
real magic happens is when you start using skills to automate things. So for example,
happens is when you start using skills to automate things. So for example, my great friend,
Ollie, Oliver Henry on Twitter is a brilliant developer and he has made stuff called Larry
Skill, where he's essentially automated the process of you can do image generation, you can
do caption creation, and you can then do automation to post it on TikTok and then a learning loop in order for
that whole process to automate itself and every day it just posts for you. That's a very real
thing and then you can think about how you can replicate that kind of automation and skill
creation within your own workflow and then you start layering them within a bigger system
and before you know it that six-month learning curve that it takes to run an entire podcast business you can probably speed run it in you know two to four
weeks and at that point you're not doing anything you know you're focusing on the real execution
sorry joe that's sort of where i've always been like i'm an operations guy i was an
engineer for like five years more of a process engineer so not like a developer engineer but
it's all the same really and you just look at your processes and that's a lot of the work that we do
or i do is i go it's the learning curve of going it's going to take me i know half a week or a week to build this and understand that process
and get from an end result but then it's then i can give it to one of the guys i can give it to
grant and it's like they can then replicate that through the automation in a minute two minutes
whatever uh so you are right it's it's that it's the biggest thing and i think it'll come back
around and what you're and what you're saying with open boys you're doing the hard yards but then things are
only going to get better so even when you do get better you've done all that hard work and it's
going to be even easier to build something bigger and better uh and even more optimized for any any
use case that you do have yeah and like you can think of it from the perspective of like
uh let's call it like an agency right like you've got an agency who makes websites like everyone
knows the real old school way to do it is you've got a designer then you've got a developer then
you've probably got someone at marketing who can specialize in seo and there's your basic web
agency now you could have and that could have been like 20 people right and over the years there's your basic web agency now you could have and that could have been like 20 people right
and over the years there's probably only five people that work in that web agency now and that's
how that business runs but now you can probably have two you can have a designer who knows how to
use open core and figma or open core meaning called um called code code or whatever they want to use
and then you can have a developer or an engineer
that knows how to use Claude code and OpenClaw. And you put those two together, you've got someone
who can prompt and be creative enough to understand the UI and the UX of a website or service.
And you've got a developer who understands the backend engineering and the systems that you need
to put in place for that website to be secure, backed up and run itself. So what I think
you're going to see over a longer period of time is that like we were talking about the, you know,
software as a service is being pretty dead quite quick, or at least like cut in 50%. Like there's
always going to be lazy people that want that out the out the box service. But the people that are
creative enough to go and sort it themselves and make bespoke independent products. I think the other side that might target the business part of this world is that the quicker the engineers who are a little bit more open to assistance or, you know, using AI toolings are going to propel themselves to such higher positions that dev teams that previously had
people of 10 are going to have like two engineers that know how to use Claude code or like AI
systems to help them develop so that's where I would say that you know the the huge businesses
that run a 10 person accounting team might end up having one accountant with claude do you know what i mean
like that's where i think this is going so to reverse it all the way to the start of this
tandem about you know like what what could you do to immediately make this impactful is i just
think like beat the learning curve as fast as you can understand how to talk to it understand how
you position yourself better
in the sense that are you a creative are you a designer do you do better with writing do you do
better with engineering great teach it everything you know to be a master at that as well without
you even lifting a finger then go and focus on the things that you're not good at so for example
like i know i can design ui websites whatever but I'm not that great at back-end.
So I've spent quite a lot of time figuring out back-end stuff
like Superbase, SQL, like so much random shit.
And now I know I'm getting to the point where I can whip up web dashboards
and pages and troubleshoot the development side.
And all I need to do at the end of that is a few prompts
and a little bit of design work and
I've got my clean finished product. Two questions on the yeah hooking it up to github and Vercel
do you need to give it its own email login like what does that process look like? You don't really
need to I just I mean everyone's going to have their different you know anal thinking about
passwords and protections and things like that but when you're using an api key it's only ever
logging in through the api access so i just made all of my projects private and gave it access for
an api key and i'll i'll ride it out if i get hacked and yeah yeah you know it's so be it i guess that's my own fault and my own
doing but there is no reason why you can't just whip up a new github i don't have enough you know
crazy shit on my github to care um like oh no you've taken my work dashboard like i hope you
don't do some of my work yeah oh no so like for me it's kind of like uh i think that people that worry about
security are the ones that should worry about security yeah do you know what i mean like if
you're already getting like anxious about like how much you could lose then you're already
over leveraged and you probably shouldn't be giving access to everything like someone even
replied to me today like are you worried about it having access to your entire life and all your files and i'm like what files it's a fucking mac mini
under my tv there's nothing on it like god forbid you go through my downloads folder and see what
random setup discs are in there like there's nothing attached to me in there so yeah there's
that that side to it as well where it's like it's the typical same thing
with crypto right like it's the same thing that happened probably back in like 2019 2020 when we
had that exponential curve of onboarding and everyone sort of figured it out and as much as
we like to lie to ourselves in the industry of crypto that you know on board the normies and on board everyone like everyone probably already
did have a look and decided i don't care i feel like there's a lot of like corporate mechanisms
there where people are just too scared i don't know how to get started so they're just like
shit all over the idea and i like that because it's kind of still puts it in the bucket of being
contrarian like i've seen this like uh infographic this week that at no point no four percent of people on the planet have used uh like a code generation tool
that is insane yeah that is that is insane every dot was like 2.3 million people yeah so it's like
then you you probably extrapolate that down further to people who like tried and spun up open claw and
it's you know it's gonna be tiny like yeah i think when i wrote my article there was 16 000
skills available and like god forbid how many of those were actual like bullshit random
like first or second third attempts from open claw bots themselves you know what i mean like
like first or second third attempts from open core bots themselves you know what i mean like
there's still an incredible amount of distribution to happen and i think that's the next craze and i
think for me personally you asked earlier and i don't think i even really answered it properly
of like what's next for me with this is that i want to be a first mover like i loved being a
first mover in crypto and i was streaming stuff and I loved it
back then but like I fell off that hurdle because I just I didn't care enough and I I got bored to
be honest and I I would try to go down the business route and I was great at that but long story short
is I probably should have stuck at it and who knows my stream might have turned into being Fred guy. So, you know, there's always that that lesson and that loss of, you know, this time I want to be a first mover because I am quite early to it.
I've already managed to get a banger article.
So I want to kind of try and stay ahead of the curve as much as I can.
And that's why I decided to do the weekly claw newsletter in order to try and be that source of like what's going on.
And then the daily closet, I took inspiration from Mando.
So shout out him with Mando Minutes.
I've been subscribed to that for fucking years.
And I've always thought genius.
Like it's such a good way to just get it spun up.
So I wanted to replicate that called daily claw and like just give someone a dashboard to go to and just see what everything's going on.
And then a weekly claw right up that's like actual by me.
And I think off the back of that, I'm going to make content and so forth.
But other than that personal side to it, how I'm going to utilize open claw the most is I'm just going to try and build skills.
The quicker you can be to building skills that are used by everyone forever is probably how you end up finding the next big thing yeah so so from this i'm gonna
i'm actually gonna go away i'm gonna get over to discord i'm gonna think about each the i'm gonna
compartmentalize them as different departments but it's gonna be one agent i'm gonna work a
little bit harder on the skills i think um i'm going to just reduce i'm going to pinch that
fucking anthropic subscription key and put it in as opposed to playing like an idiot
there's always got to be something to learn isn't there what else am i going to do i need that i
need that uh social twitter data thing off here as well social data yeah yeah i'll grab that off
you as well but i think that's that'll keep me busy for the rest of the week yeah for sure and then there's
always going to be something new to learn and that's the best bit right is that we're at the
the very start of this and i think there's uh like it's going to be really interesting either
something major is going to happen and a huge front mover someone like apple apple probably
holds the the dangling carrot to the entire industry to be able to happen and a huge front mover someone like apple apple probably holds the
the dangling carrot to the entire industry to be able to go and make something that everyone just
uses but how reliable they are going to be for that i doubt it i reckon you've got like another
two three years they still haven't even mastered the fucking headset stuff do you know what i mean
like they're all over the place so until someone like apple decides to click their fingers and say
we're going to make open call for everyone and has another steve jobs moment i think we've got like
an incredible situation on our hands where yes it's a rapper at the moment but people are already
breaking out of that now and building real things with it you know ollie has made larry brain which
is like netflix for skills and it's like a monthly subscription service but that's one
example right and I'm sure as the learning curve continues to accelerate and make it even easier
for people we're going to see so much coming from it so I just want to be at the front of all of
that to be honest awesome so give people um we'll leave it in the link below um we'll put it in our
our discord we'll put it in our
um newsletter ourselves for weekly claw and daily claw as well and um yeah who should we have on
next that's in this field that you really like respect and it's got ollie takes yeah honestly
you've got i think you've got ollie articles they're fucking they're brilliant i'd love to
try and get him on that'd be great super great. Yeah, no, honestly,
Ollie would be brilliant for a next step as well
because he's able to take it further than I am.
Like I was saying earlier,
we talk on a daily basis
and we run loads of ideas back and forth.
And I've got the design mindset.
He's an engineering developer
and just an absolute go-getter.
So he's going to be able to move
like three times faster than I am on shipping things.
And that's, you know, the next level up from someone like me who enjoys being maybe a bit more front facing.
I'll be the entertainer and the creator behind it.
Ollie is going to be the guy to make the first like 10K MRR subscription based platform with OpenCore.
So get him on for sure.
Awesome. You have to make an intro.
But yeah, Johnny, thanks so much.
I know we've run over and I appreciate you taking the taking the time out that was really great it's all right um yeah come back
on in a few few weeks a few months let's let's see what's actually happening maybe we're on to
the next thing but i feel like um yeah we're just at this weird inflection point now where
anything's basically possible so even if it's not open claw whatever the fuck it's going to be it's
going to be interesting anyway yeah no it's all right it's been an absolute pleasure and uh
yeah cheers sorry it ran over but we could go on forever all right henry
take it away
