The speaker is making a significant investment in local AI models, particularly through the OpenClaw framework on Mac Studios, to create a fully autonomous, 24/7 AI agent company. This approach is presented as the future of AI, offering unprecedented capabilities and cost-effectiveness compared to cloud-based solutions.
Mind Map
Нажмите, чтобы развернуть
Нажмите, чтобы открыть полную интерактивную карту
I just spent $20,000 on OpenClaw and
plan on spending another $100,000 on it
by the end of the year. I truly believe
the return on investment on all these
Mac Studios I'm buying is going to be at
least 10x what I'm spending. You might
think I'm crazy, but if by the end of
this video I don't have you convinced
what I'm doing is actually genius, then
I think you're the crazy one. In this
video, I'll go over what I just spent
all that money on, what my open claw is
doing on all this new hardware I just
bought, why I believe this is the
future, and show you how you can start
doing the exact same thing I'm doing
without spending a dollar on any
hardware. You can use the exact Mac Mini
you bought or even the dusty HP laptop
from your closet. You are about to get a
peak into the future of AI agents. Let's
get into it. So, real quick, we're going
to cover a bunch in this video. Chapters
down below if you want to skip around,
but basically what we're going to be
doing is I'm going to tell you why I
spent all this money to run local AI
models, what all the advantages are,
what this is going to allow me to do.
I'm going to give you a demo of exactly
what I'm building, which I promise you,
you have never seen anything like this
in your life. There is nobody else on
planet Earth building what I'm building
with Open Claw at the moment. I'm going
to show you how this will change the
entire world and why I believe all
normal people will be running out to buy
thousands of dollars worth of computers
by the end of the year. And at the very
end of the video, we'll go through how
you can do the same thing I'm doing on a
much smaller budget. Even if you have a
$500 Mac Mini or a $10 HP laptop in your
closet, I'll show you which local models
you can run so you can start doing
something similar to what I'm doing. I'm
about to take you through a glimpse of
what I think the future of AI agents
are, what I think the future of OpenClaw
is, and what I just think the future of
AI is. You're going to learn a ton, and
your mind is going to be blown by what
these AI models are capable of. So,
here's where we're going to start. The
$20,000 I spent is on two Mac Studios.
These are Mac Studios with 512 GB of
unified memory each. Why 512 GB of
unified memory each? so that I can run
the largest, smartest local AI models in
the world. Right now, that is Kimmy K
2.5. Kimmy K 2.5 is 600 gigabytes, which
means I need to load 600 gigabytes into
local memory. With two Mac Studios at
512, that's a terabyte of AI models I
can load into local memory. If I were to
do the same thing with Nvidia GPUs, I'd
need to spend over $100,000 on Nvidia
GPUs, but I can do that with the unified
memory of Mac Studios. Now, why do I
want to run large local models on my
computers on these Mac Studios? Well,
here are the five reasons why. One, you
run these models completely for free.
I'm not paying for API costs anymore
when you use Claude Opus or Chad GBT or
any other AIS that are in the cloud. You
pay for every single token. These local
models, because they're running on my
Mac Studios, they are completely free to
run. And because they are completely
free to run, I can have them work 24/7
365. This is an absolutely massive
advantage. The number one objection I
get when I tell people I'm doing this
is, "Oh, local models, they're stupider
than cloud models. Kimmy K 2.5 is not as
smart as Opus. It's not as smart as Chad
GBT 5.3." And you know what? You're 100%
right. But here's the thing. The fact
that I can run these AI models 24/7
completely changes what AI is capable
of. Now, I'm going to show you examples
once we get through this. I'm going to
show you my actual setup I'm building
with these Mac Studios. But the fact
that these AI models can now run 24/7
365 means I can do things like have them
read Twitter and Reddit all day and all
night looking for challenges. Have them
coding all day and all night building
solutions to those challenges. Have them
shipping those apps all day and all
night promoting them on all the social
media. I can constantly be creating
value from midnight to midnight every
single day. They they don't have to eat,
they don't have to sleep, and they don't
cost me any money. It's literally like
having 24/7 employees working for you
that never complain, never demand more
money out of you. They just have that
upfront cost of buying the computer. And
it doesn't matter that they're dumber
than Opus. I can't do what I just
described and what I'm going to show you
later in this video with Opus. If I were
to do this with Claude Opus or Chad GBT,
I'd be spending at least $10,000 a month
on API costs. But because I can run this
locally, I can have these models going
all the time, finding challenges,
building solutions to challenges, and
creating value for me and my company.
There's many other reasons why you'd be
wanting to run local models. Privacy,
for instance, everything that happens
happens on the computer. Nothing goes to
the cloud. Nothing goes to chat GPT
servers. None of this can be read by Sam
Alman or the owners of the other AI
companies. Everything is completely
private. In fact, I can unplug my
computers right now if I wanted from the
internet and they'd still be going and
running AI models. There's no internet
required. It's all private, all local.
No one can see what I'm doing. Next,
it's educational. By running these local
models, I'm learning how AI works. I'm
learning a ton about what goes into
running an AI model. I am learning about
the most important technology in the
entire world. And that in itself is
worth a lot of money. And lastly, it's
just fun AF. It is just so much fun
looking down at your desk, seeing
computers running, and knowing there are
AI agents running, doing things for you
24/7. It is fun. And I don't care what
other people say. You're allowed to have
fun in this world. You're allowed to
spend money to have fun. You're allowed
to do things for the fun of it. Don't
let people tell you you're not allowed
to have fun anymore. Take it from me,
you're allowed to have fun. And so that
is another reason why I'm doing this.
So, what exactly am I building with all
these local models running 24/7? Well,
have a look at this. This is my company.
This is my oneperson 24/7 365 AI agent
company. And this might look strange,
but let me walk through exactly what's
happening here. What you see here are my
AI agents, and they are working as we
speak. They are reading X. They are
going through Reddit. They are searching
for problems to solve. They are reading
my tweets. They're watching my YouTube
videos. They're looking at the
performance of all my content. And as
you can see, sometimes the agents even
meet up together, go to a meeting table
and have discussions. Right now, there
is a standup happening for all my AI
agents where they are brainstorming new
features for Creator Buddy. I can track
the live activity. So, you can see the
live activity over the last hour. A
roundt started to brainstorm new
features. They're all coming up with new
ideas for features for Creator Buddy. At
some point they will be handed off to
Henry who is the manager, the strategic
manager of all the AI agents. If we take
a look at the org chart of this AI
digital company of all these local
models running and talking to each other
and planning at all times, you can see
I'm at the top as the CEO. Henry is the
chief of staff. He's running on Opus 45,
but all he's doing is getting ideas and
approving and disproving. He just has to
do a couple prompts a day where he
approves or disproves ideas. the local
models hand to him. As you can see,
right now he is in the standup
brainstorming new features for creator
buddy. Then I have my other agents
working for me in this organization. So
you can see I have a creative team which
all they are is a local model running on
my Mac studio running off of Flux 2 that
is able to generate images for me,
thumbnails for my YouTube images for my
Twitter account. Whatever I need, that's
all being done locally. I have my
research team which is Scout who's an
analyst that is being powered by GLM4.7
which is a huge local model that I'm
running on my Mac studios. This is
constantly reading Twitter, constantly
reading Reddit, finding challenges to
solve and handing them to Henry who is
my chief of staff and then I have many
other agents working in my digital
organization as well. I have an
engineering team who's constantly coding
for me and many other AI agents as well.
They are able to accomplish so many
things. So, for instance, right now as
they discuss new features for Creator
Buddy, they're looking online, seeing
what people want out of content tools
and building it out and discussing it
with each other and learning from each
other. And this goes much deeper than
that. Each one of these agents, have
their own memories, have their own
personalities, are building their own
relationships with each other agent. If
you look down here below, I have a list
of all my agents. If I click on one of
them, you can see here Quill has its own
soul. So its own personality, how they
think, their own signature phrases,
their own voice of how they talk, their
own speaking style, their own
responsibilities, even has their own
relationships with the other AI agents.
Every time one of my agents speaks to
another agent, it actually changes their
relationship. They can become better
friends. They can become bitter enemies.
It's just like a real workplace where
you have friends, you have co-workers
you like, you have co-workers you don't
like, and their relationships can shift
and evolve over time. They also have
their own memories. So Quill is a new
agent I just hired just now. So it
doesn't have its own insights and
strategies, but as they participate in
meetings, as they have conversations
with other agents, they can come up with
their own insights and strategies. So
for instance, Quill is my creative agent
who writes tweets. Maybe Quail goes in
and has a water cooler conversation with
Scout, who is my local model, who is
constantly reading Twitter, and Scout
tells him, "Hey, that tweet you wrote
the other day, it's performing really
well." Quill can then get a memory that
says, "Okay, tweets like this perform
really well. I need to write more about
them." So, my digital society here, my
digital office, they're constantly
learning from each other. They're
constantly talking to each other. They
can do many things. They can even have
water cooler conversations, which you
just saw there. They're now walking over
to the water cooler and talking to each
other. This happens all autonomously
24/7 365. They are constantly
researching, constantly writing,
creating, coding. They are constantly
learning from each other. They're
constantly building relationships with
each other. And I don't need to be a
part of this. I can just sit back and
enable them and make sure they're doing
good work. When I'm sleeping, they're
working. When I'm eating, they're
working. When I'm watching the Patriots
win the Super Bowl, which this will look
really bad if they don't end up winning
the Super Bowl today, they are talking
to each other and working and creating.
This is only possible with local models.
If this were all done with Opus, if all
of these AI agents were Opus and Chad
GBT, I'd be spending the cost of these
Max Studios every single month. But
because I have local models working, I
can offload a lot of this work to those
local models to save tremendous amounts
of money. And yes, I still have Opus as
a part of this, but Opus, Henry, the the
chief of strategy, is only doing
decision-making. He isn't doing the
dirty work. He isn't searching Twitter
and searching Reddit and doing a lot of
the writing and creating. He just
approves and disproves. Everyone else is
doing the hard work. All the tokens are
being burnt by the local models. As I
buy more computers, as I buy more GPUs
and devices, which I fully plan on
doing, I'm going to buy the Mac Studio
M5 Ultra when that comes out in a few
months. I'm going to buy a DGX Spark.
I'm going to buy a whole lot of other
computers. By the way, Nvidia, if you're
watching, send me the DGX Spark. I will
talk about it so much. I will be adding
all these devices to my organization, to
my local data center. And as I do that,
I can run more local models. I can
expand my oneperson company. I can get
more employees in here just chugging and
working 24/7. I can even, as I add more
GPUs, train my own custom models.
Basically, train my own employees that
will be working in my company. In a
second, right after I go through this,
I'll show you how you can run your own
local models, even if you have really
crappy computers or a Mac Mini or
anything. I'll show you how to run that
in a second. But just to wrap this part
up, this is the future of AI agents.
Claudebot unlocked this. Claudebot
unlocked the ability to run your own
agents autonomously. The issue was is
you can't take full advantage of
Claudebot with APIs and cloud models.
The bottleneck once models can run
autonomously is the cost of the models
themselves. But by running these models
locally, that bottleneck disappears.
Your models can now be completely
unchained and do so many more use cases
like constantly finding challenges
online, like constantly building and
coding and creating things, like
constantly reviewing all your work. This
is the future of AI agents, and this is
the worst it will ever be. Right now,
the best local model is Kimmy K 2.5,
which is near Opus 4.5 level. Over time,
and probably by the end of the year, the
local models will be better than that.
They'll be better than that and be able
to run on much cheaper hardware. This is
the slowest, dumbest, and most expensive
it will ever be. And right now, this has
been amazing for me and what I've been
able to accomplish. Another one of the
biggest questions I get is like, okay,
that all sounds cool, but like what are
you actually getting out of all this?
What are the workflows that you're able
to enable that you weren't able to do
before? Let me give you a couple
examples here. So, here's two examples.
You can just go straight down here.
Here's a couple things we've been doing.
Number one, I have a researcher agent,
which is a local model, constantly
reading Reddit 247 365. It finds
challenges people are having in
subreddits. It hands it to my strategy
officer, which is Henry. Henry decides
if the challenge is good or not. He
takes the good challenge and hands it to
the developer agent. The developer agent
then codes an app to solve that
challenge. The developer agent then
ships that app and puts it live on
Verscell through the Verscell CLI. Then
the researcher agent DMs the original
poster and says, "Hey, we came up with
the solution to this problem. This is
just a constant closed loop. I do not
need to be a part of that is constantly
going 24/7, 365. This is not possible
with cloud APIs." Here's another
example. I record a YouTube video, just
a raw YouTube video. A local agent edits
the video. So, it cuts out all the blank
space that is then handed to a local
image model Flux 2 that is running
locally on my device that generates a
thumbnail for the video. Another agent
goes in in the browser, puts the video
onto YouTube, puts the chapters, posts
the video, and a day later, all my
agents meet. They go to the table I
showed you before, and they discuss the
performance. They go over the
transcript. What did he talk about? The
hook, the thumbnail, what worked, what
didn't work. And based on all those
learnings which get saved to their
memory, they write a new script. These
are the type of autonomous use cases
that were not possible before. They were
not possible with Chad GPT. They were
not possible with Claudebot just using
cloud APIs. It's only possible by having
compute on your desk 24/7. This is what
I can do now. This is what I'm
developing. And this is what you can do
in the new world as you start running
your own local AI models. So now the
question becomes, "Hey, Alex, do I need
to spend $20,000? That's a lot of money.
I can't afford that right now. I need to
be able to run local models and do these
use cases without spending all that
money." Well, I got good news for you.
You can do this without spending
$20,000. Can you do it to the degree,
the intelligence, and the speed at which
my models are doing it? No. But you can
start off cheap and then slowly layer on
from there. Let me show you. So, let's
talk about which local models you can
run on different budgets. You don't need
to spend $20,000 like I have. You can
have different budgets and different
local models on each. Now, are the
cheaper ones going to be as smart,
efficient, fast as larger ones? No. But
it's good to start somewhere and then as
you go, as you figure out new use cases,
as you figure out how to fit this into
your workflow, you can either buy more
hardware or change things around or
experiment. It's up to you. I don't
recommend everyone just go out and spend
$20,000 like I did. No, I do not
recommend that. Start cheap, then slowly
build your way up. So, if your budget's
only $100, that's great. You can buy a
Raspberry Pi on there. You can run
simple, small, local models like Gemma,
like Tiny Llama, things like that. And
you can do different things. You can
have simple chats. You can do smart home
things. You can do very simple things.
There are use cases there. As your
budget goes up, you can do more
interesting use cases. So, if you went
out, you're like many people like myself
that bought a Mac Mini when you
discovered OpenClaw, you can run models,
too. There's Llama models, there's
Mistral models, there's Quen models you
can run that could be like personal
assistants that can do a little bit of
coding. Will it be as good as Claude
code? No. But you can still do some
small things on it. If you have a larger
budget, you buy maybe the
top-of-the-line Mac Mini. You can start
doing some more serious coding. You can
start doing some more serious research.
Maybe your model now starts reading
Reddit at all times like mine. By the
way, as I get into the upper tiers, if
you learned anything at all, leave a
like down below. Make sure to subscribe
and turn on notifications. I'm going to
be making so many videos about these use
cases and about what I'm building. It'll
blow your mind. Make sure to turn on
notifications for that. And let me know
down below in the comments if there's
any specific part of this you want to
hear more about. Whether it's running
cheaper local models, whether it's about
running more expensive ones, use cases
for this open claw as a whole. Let me
know which one you want me to dive
deeper into for my next video. As your
budget increases, maybe you buy a Mac
Studio M2 Ultra, which is the older
generation. you can start doing more
professional workflows, have multiple
agents going at once, and then once you
get to my level, where I'm at now with
the Mac Studio M3 Ultra, so I have two
of these M3 Ultras, one's on my desk
now, another's coming in the mail this
week, you can start having a fully
autonomous organization working for you,
which is amazing, and run local models
that are almost as good as Opus, right?
And almost as good as just good enough
if it can run 24/7. This is the future.
And this might not look like much right
now what I'm showing you, but I am
actively building this out as we speak.
I'm actively building out more use
cases. Nobody else in the world is doing
anything like this. I'm not kidding.
Where we are in technology right now is
absolutely unbelievable. This is all
green field. This is a brand new
technology. Open Claw only became
popular like 2 weeks ago. We are 2 weeks
into this revolution. If you go in now,
if you tinker now and experiment now and
try new things out, the odds are you're
doing something no one else in the
entire world has done. It is early on
these technology revolutions where all
the opportunity is. I'm being serious.
This is where all the opportunity is. If
you strike now and you experiment, try
new things, invest, dedicate yourself to
this, you could create success and
opportunity for yourself that no one has
ever seen before. That's what I'm doing.
That's why I went all in and invested in
my own personal local data center here
because I want to do things the world
has never seen before. And it's
something you can do as well. If you're
joining me on this journey into the
unknown, if you're joining me on pushing
the limits of technology, if you're
joining me on trying to build the first
one billionoll business, make sure to
subscribe. Make sure to turn on
notifications. I'll be taking you
through this. I'm not holding back any
secrets whatsoever. Everything I do, I
will show to you. and you can copy and
join along with me and do incredible
things. I hope you learned something. I
absolutely love making these videos for
you. Thanks for joining along for this
Нажмите на любой текст или временную метку, чтобы перейти к этому моменту видео
Поделиться:
Большинство транскрипций готово менее чем за 5 секунд
Копировать одним кликом125+ языковПоиск по текстуПерейти к временным меткам
Вставьте ссылку на YouTube
Введите ссылку на любое YouTube-видео, чтобы получить полную транскрипцию
Форма извлечения транскрипции
Большинство транскрипций готово менее чем за 5 секунд
Установите расширение для Chrome
Получайте транскрипции прямо на YouTube, не переходя на другие сайты. Установите наше расширение и открывайте текст любого видео в один клик — прямо на странице просмотра.
