OpenClaw Will Never Actually Run Your Business… Try This Instead.
By Liam Ottley
Summary
Topics Covered
- OpenClaw Traps You at 5-10% Potential
- Claude Code: Robust Harness Beats Indie Wrapper
- Ferrari Engine in Honda Chassis Fails Scaling
- Custom Claude Code Scales Complexity Exponentially
Full Transcript
So, if you are a business owner who's been using OpenClaw and you think you're at the cutting edge of this technology, I can promise you you're only getting maybe 5 to 10% of the potential out of this kind of technology at the moment.
And on the other hand, if you're someone who's maybe been less impressed or struggling with using something like this, there is a reason for that. And
I'm going to be revealing in this video why that is. The system that I've created and I've been using that ticks all the boxes of OpenClaw and far beyond. And the reason I'm making this
beyond. And the reason I'm making this video is because I've been running into so many successful and smart entrepreneurs who are excited about this technology naturally and the potential it has for being able to actually get
them some time back and actually be able to do things for them. So them as a founder aren't just change their business like a job. But the thing is Open Claw is open flawed. There are
major issues with it that are going to stop you turning it into the true AI assistant that runs your business and does things for you. And in this video, I'm going to be breaking those down so that hopefully I can convince some of you to make the transition to the system
that I'm going to talk about so that you can truly realize the benefits of this incredible technology that is now in our hands. And if you guys are new to the
hands. And if you guys are new to the channel and don't know who I am, my name is Liam Mley and I've been building AI businesses for the past 3 years. I've
been in business for the past seven. And
this stuff I'm going to talk about in this video is by far the most exciting thing I've ever seen in my career as an entrepreneur. I've barely left the house
entrepreneur. I've barely left the house this week. I've barely slept at all
this week. I've barely slept at all because for the first time for me and the friends that I'm here with who are also nerding out on getting these systems set up. We can actually see us moving towards a future where we can use a system I'm going to explain here to
actually automate our entire business.
And over the next 30 days, I'm actually documenting my process of automating as many of my tasks as possible. And so
long story short, the main difference I'm talking about here is not using open claw or open floor. We are using clawed code.
And if that sounds scary, stick with me because yes, this may sound like a dev tool, but it is actually the most powerful AI assistant and productivity machine that I've ever seen. And in this video, I'm going to be breaking down why
you must make the switch. Now, I'm going to do a side-by-side comparison of all the things you love about this and how Claude Code beats it in every realm. And
at the end of the video, I'm going to be explaining what it actually takes to get started on building the system that will go like this rather than open claw, which is going to go like this and plateau for you. Okay, so the first
thing you need to understand is what open claw actually is and what clawed code is. So it starts off with a model
code is. So it starts off with a model of some sort. We have the model as a sphere.
We call this our model. So this might be uh anthropics uh opus 4.6. It might be open 5.2. Whatever it is, this is some
open 5.2. Whatever it is, this is some kind of LLM or model. That is our starting point. Now what Claude code is
starting point. Now what Claude code is here is a harness. You may have heard of ChachiBT rappers. That was kind of a
ChachiBT rappers. That was kind of a slur term used by people to throw shade at basic applications that took the underlying model and put a thin layer around it and called it a business or
called it a SAS. So that's like a wrapper.
And this will all make sense in a second. You have to understand the
second. You have to understand the difference to this to really fundamentally know why you must make the switch now because once you realize this, you cannot go back. So we have a wrapper here. Recently we've had the
wrapper here. Recently we've had the emergence of what's called a harness.
Now, I've still been trying to figure out where we made the jump and and why one took over the other. You don't hear much about rappers anymore. You hear all about a harness. So, I don't really know how to draw a harness, but recently I've
been drawing them like this. Think of it as this big uh we have this powerful orb. And you guys can drop a like for my
orb. And you guys can drop a like for my drawing skills cuz I think that's pretty damn good. We have this powerful orb
damn good. We have this powerful orb here, which is the model. so good at understanding text, understanding images, being able to generate things, generate code, uh reason through things, analyze large amounts of text and code.
And what we're doing is we're putting it into a harness where I mean, this is maybe a nice way of putting it. We have
this harness holding it in place and giving this model a whole bunch of new features and abilities in a very robust way. So, I haven't really figured out
way. So, I haven't really figured out how to draw these yet. We can maybe put them as like this.
So it gives it these very powerful arms that stick out and allow it to do things in the digital world.
And the best example of a really great harness is Claude code where Anthropic has poured billions of dollars into the engineers and people behind building out and training their harness to be able to
do things like uh these actions being like reading files, creating files, deleting files, searching the web, spinning up sub agents to do deep research or spinning up sub agents to dig into file systems. The harness that
Anthropic has created for claw code is so well created and so adapted to the digital environment it lives in. It is
it seems like an absolute genius. It can
plan everything out. It can open files, edit them, search the web, pull things in, write code, write text, read images, and read PDFs and basically anything you throw at it. And that's what has made this such a powerful and explosive tool.
Now the thing is when it comes to openclaw it is essentially a wrapper or a harness that has been built by an indie developer to do the same thing as this but also
connect it to telegram or to connect it to something that's handy for you. So
we're going to go into the features in a second but I want to just point out that we've taken essentially the same model.
One of the one of I guess perks of open core is that you can swap the model out but at the end of the day if you're trying to make a really powerful system would you really want to use anything but the most powerful models opus 4.6 six on it. The new one just came out as
well. So, we have instead of this really
well. So, we have instead of this really robust and powerful harness that has been had hundreds of millions of dollars poured into it, we have instead something made by an indie dev. How do I
make like a janky looking harness here?
Barely even hanging in there. And no
disrespect to the dude at all. I I'm not saying I could do any better. It's just
the reality of one dude making something that just trying to replicate what has been done by a one of the most valuable AI startups or technology startups in the world right now with the best talent
in the world. And so it's essentially clawed code and you've ripped off all of the the arms. Say you've ripped these off. You've taken away all of its
off. You've taken away all of its ability to read files, view files, uh search the web, etc. And you've stripped it back to just this model and then you've dumped it in a crappier harness.
Again, no disrespect. and then tried to kind of Frankenstein and duct tape on the same arms. For example, there's a whole bunch of great stuff that comes out of the box with claude code, like web searching, which is super native.
They spent tons of money training it and and doing post training to get it to really use these web search tools very well to be able to spin up deep research agents to go out and search the web uh separately, come back and feed you
information. and all of this stuff you
information. and all of this stuff you now have to like kind of plug in using uh third party tools and try to recreate this with a inferior harness that has not been trained with nearly as much uh
effort and resources. So that is the clear difference and that is why out of the box this thing is so much more powerful has been refined over billions of dollars and this thing is made by an indie developer where you have to
basically plug in all of the things just to get it to baseline and by nature that means it's not going to perform as well.
Now I hear what you're saying you may be like oh but I can connect this to Telegram. Cool. We can do that, too. Oh,
Telegram. Cool. We can do that, too. Oh,
but I can set up chron jobs and it'll do stuff for me. Cool. You can set that up, too. Everything that you could do with
too. Everything that you could do with with OpenClaw and that you you love about it, you can do with this in a completely custom setup that ties into all your own workflows. And ultimately,
it is far more controllable and scalable longterm than something like Clawbot where the performance is going to flatten out as the complexity scales.
I've been building my system now for coming up 2 weeks and I can assure you the levels of complexity that I'm reaching now with the workflows, the layers of data, the documentation.
There's no freaking way you could achieve what I'm getting out of it right now. And I've got a video coming down.
now. And I've got a video coming down.
It might have already been out, but you can see all of the stuff that I built and also the things that I'm looking to build in the future. So, understanding
what a harness is that this is a great harness proven with lots of money and this one is fundamentally not as good. I
think the best way I put it is you've essentially taken a Ferrari's engine.
Well, you've got a whole Ferrari here and you've taken the Ferrari's engine out, ripped the model out, and then dumped the engine, the really powerful model into a crappy chassis. So, you
chuck it in a Honda. That's the best way to explain what has happened here.
You've got the incredible power of the core model, but it's just in a really bad structure to allow it to do what the the model or the engine was intended to do. Okay, so I just quickly want to riff
do. Okay, so I just quickly want to riff through some of the things that when I try to convince entrepreneurs that I meet to make the shift, they're like, "Oh, no, but like I've got it working and you know, I can put it on my Telegram and I've got this and this."
So, let's just run through a few of the the the best the biggest sales points of something like OpenClaw and why you can do all of it and better with Claude
code. So, firstly, Telegram,
code. So, firstly, Telegram, WhatsApp, etc. deployment. You can access your
etc. deployment. You can access your open claw through a platform of your choice. Cool. You can do that too. In
choice. Cool. You can do that too. In
fact, because it's custom and you can control everything. I have created mine
control everything. I have created mine on my phone to be far more powerful. I
can't go into it into this video, but where you can have layers and layers of complexity where you have orchestrator agents, you have separate subtopics.
I've fully customized it to have daily briefs coming into the Telegram and different channels. Full customization
different channels. Full customization of my Telegram setup and that is ultimately where we're trying to get it where we can run our business from the phone. So secondly, chron jobs.
phone. So secondly, chron jobs.
This is a core feature of open claw that allows you to set up recurring jobs. So
it'll ping, it'll say, "Hey, okay, can you please do this thing?" One, the way that it's done open core is very confusing and in my opinion, not that great. And two, you can set it up in
great. And two, you can set it up in claw code yourself fully custom. You can
have layers and layers and layers of them. And just like everything else,
them. And just like everything else, it's much more controllable and you can fully understand what's going on much more easily. So that's tech. Third is
more easily. So that's tech. Third is
memory.
Uh, another great thing is like, oh, it learns about me and it updates. And the
thing is, it's a memory system that is kind of opaque and harder for you to maintain yourself and and make modifications to. Not only do I have
modifications to. Not only do I have memory in my cloud code setup, I have a very structured context OS that allows me to continually keep that updated also via documentation of new features. Very
importantly, I have a full database with all of my business data condensed into one place. So, I pull from all my data
one place. So, I pull from all my data sources into one place and my Telegram assistant has access to be able to query that data, get any information I want.
So overall, I have a more comprehensive memory system that I have full control over and I have it connected to a database that has all of my business data in it as well. Fourth is web searching.
This is a pretty easy one. Claw code has native web searching in also deep research agents that can be kicked off.
Uh so the web searching and claw code is native. It's been trained on it is
native. It's been trained on it is categorically better than using some plugged in third party tool uh to provide web searching capabilities to your open claw. So we win there as well.
Fifth is model agnostic.
Uh, sure. Open claw allows you to plug in a whole bunch of crappier models than Opus 4.6 or whatever the latest one is.
Personally, I'm using this thing for some really complex business processes, analysis of data, researching trends, uh, strategy. I want the Ferrari engine.
uh, strategy. I want the Ferrari engine.
I don't want the Civic engine put in this thing. And Anthropic does have
this thing. And Anthropic does have three different model tiers, Haiku, Sonnet, and Opus. I use Haiku for my Telegram orchestrator to keep the cost low there. And then I only bust out Opus
low there. And then I only bust out Opus for the big task when I need it. So, I
mean, sure, you could maybe say you beat us there, but at the end of the day, this is going to be used for very complex things. I want the best engine
complex things. I want the best engine or brain driving that. And if I want to, I can always downgrade to a cheaper model and bring my cost down. You can
plug in your anthropic plan and pay what is it 50 or 90. I think 50 bucks is one of the lower plans. I'm on the 200 plan and I'm absolutely smashing this thing every day. I've only hit limits once or
every day. I've only hit limits once or twice. In which case, it is well worth
twice. In which case, it is well worth the $200 I'm spending. Okay, so I hope you can agree here that this is clearly the answer. But what does it actually
the answer. But what does it actually take to get this set up? What does it cost to make the transition? You might
feel like you've sunk a whole lot of time and effort into this already. And
just on that note, you really need to understand this graph here. You will get quick results with your open claw. But
as complexity scales and you want to build more and more and more into it, it is going to suffer. May it might even start performing worse over time.
Compared with something like clawed code where when you use the proper template to get things set up a good base and build on top of that in the right way, you can see compounding benefits. There
is no limit that I have found so far to the scale and complexity that I've built into my system. Unfortunately, to get started, this is not as easy as just running one command and getting things off the ground. There is a bit more setup involved and a little bit more of a technical barrier, but that should not
stop you from pursuing this. I've
literally been sitting here. We're doing
this generational cla code lock in here.
If you want to see the behind the scenes of me and my friends absolutely geeking out on this stuff up till 3 4 5 a.m.
Absolutely nerding out, going all in on this stuff to build these systems for our businesses while I've been here with them. Some of them are completely
them. Some of them are completely non-technical, never touched line of code in their life. And by using my template, which is the right folder structure set up to get you started in the right direction and also the right commands to use where Claude code is forced to become your assistant in
getting this set up where you just say this is what I want. It's going to walk you through a bunch of questions to figure out exactly what you want and then technically spec that, search the web, figure out how to do it and then from there it can actually build and implement it into your workspace. The
other thing you're going to need aside from a good starting template is to be surrounded by people who are also doing this. That's why I'm supposed to have
this. That's why I'm supposed to have left Cape Town already, but I decided to stay because my friends are here and they're also all over this and we are just sharing learnings around. When I
build something here, I just tell my agent to pull it all into a markdown file, air drop it over to them, they tell their agent to install this, and within a few minutes, they can have the same functionality that I just created, and it took me a few hours to do. So,
with all that in mind, I'm actually running a webinar that's going to be kicking off next week, depending on when you're watching this, but there'll be a link in the description. There's far too much. I just did literally did a
much. I just did literally did a workshop with my accelerator community today, walking them through it. There's
just so much to get through. So, I'm
going to be doing a full free webinar for any business owners who want to understand what this is, see what my system is and how I'm running it, understand what steps I need to take to get their set up, and also to be able to grab my free template. That is exactly
what you need to get started and moving in the right direction. Also, as part of that webinar, you'll get added into my Telegram channel, the AIOS or AI operating system, which is what I'm calling this. You'll get access to the
calling this. You'll get access to the Ground Zero Telegram channel where I'm constantly dropping in my learnings and findings that come from my friend group here, that come from the workshops I do at my accelerator and the community there, too. and you'll also get put on
there, too. and you'll also get put on my newsletter list where you'll get an email every single day breaking down what happened in that Telegram channel direct to your inbox so that you can apply it to your business. So, I'd love to see you on that webinar. It'll be the first link in the description. I really
want to help like tens of thousands of business owners to see and understand this stuff and apply it because the future is already here. This is it. This
is how businesses are going to be run in the next 2 3 4 years. You can have it right now. And I'd love to be able to
right now. And I'd love to be able to share with you what I'm learning. As you
can tell, I'm pretty fired up by it. And
if you are impatient and you want to get access right now rather than wait for the webinar, you can see if you're a good fit for my accelerator program where every single day, Monday to Friday, I'm doing an hourong workshop with my community, showing you what I'm learning here, doing live builds, doing
buildongs, workshops with the community.
It's community where you can go to be around people who are at the tip of the spear and pioneering this new future of business. And that's all for the video,
business. And that's all for the video, guys. I hope I've been able to convince
guys. I hope I've been able to convince you to at least be open-minded enough to consider this because I can tell you, you definitely want this outcome, not this for your business. If you want to watch something else from me to get more context on how I plan to automate my
business and get 60 to 70% of my work automated in the next 30 days, you can check out this video up here where I do a full breakdown. But aside from that, guys, hope to see you in the webinar.
Hope to see you applying this stuff and subscribe to the channel for more on this because I'm going to be absolutely ripping it over the next few months for you guys. I'll see you in the next one.
you guys. I'll see you in the next one.
Loading video analysis...