As of today, you cannot buy a base Mac Mini from Apple. Here's 95 Mac. The base Mac model is sold out. I've been building local AIs on this channel on
Mac Minis and other machines as well for a while. So, when the sponsor of this video, Abacus, pitched me a cloud service whose whole premise is you
don't need to buy a Mac Mini, I had to laugh. Good timing, right? Let's talk about whether that pitch actually holds up. But first, I got to check
this. Here's a Mac Mini on the Apple Store. M4 chip. Wow. currently unavailable. Are you serious? Okay, so what are people actually doing? Well, I'm
in Washington DC and this guy near me is actually selling a Mac Mini, a base model with OpenClaw pre-installed for, yeah, 1,500 bucks. And this one is
charging almost 900 bucks just to come to your house and install Open Claw on hardware you already own. So that's where the market is this week. Turns
out there's another option. Okay, so how does this work? Abocus paid for this video. They did not pay for the take. And there's obviously pluses and
minuses to running OpenC claw hosted or locally. Abacus claw is open claw hosted for you. That's the whole pitch. But it's more intricate than that.
Instead of setting up a machine, installing a framework, wiring up API keys, exposing it to the internet safely, and keeping it patched, they run it
for you. You just log in, run your own AI system with memory, personality, and real world task automation. Start Abacus claw. Now, you have to have an
account and sign in, of course, first. Here you're going to describe to your agent what you want in plain English and it's live. The thing that
they're charging for is not the software. OpenClaw is open source. It's free. You can run it yourself today. The catch is you have to have the
hardware to run it and you have to set it up and all the things I mentioned earlier. What you're paying for with Abacus is the part where you don't
have to do any of that. And it's not going to cost you almost 900 bucks like you'd pay to this guy. Actually, it's going to cost you a lot less. And
look, on this channel, the answer is usually run it yourself, right? That's what we're doing here. But the number of people that actually follow
through on self-hosted AI is pretty small. If you ever started a HomeLab project and watched it collect dust, you know what I'm talking about. So,
does this actually work and is it worth what they're charging? Let's find out. Welcome to Abacus Claw. Here you can set up WhatsApp, Telegram, Slack,
or Discord. You can always manage that later if you want to. If you're ever on this screen, you can just do that up here. Configure WhatsApp,
configure Telegram, other connectors, and you can get to it through here, too. If you want to start a brand new session, new session button. It's not
that complicated. Okay, but it's already going. I didn't need to do anything. I just clicked a couple buttons and it's off. Hi, I'm Alex. Get me
today's Mac Mini availability. Got to keep it busy. Oh, real quick to anyone new to the channel or new to the whole space. What is OpenClaw? So,
OpenClaw is an open source AI agent. It is self-hosted. It runs on your own hardware. You plug it into a model, whether it's a local model, clawed,
GPT, whatever you're using, and it acts as the connective tissue between that model and the rest of your digital life, whether it's email, your files,
your browser, and your messaging apps. The difference between this and a chatbot is a chatbot just talks back and forth. An agent actually does that.
Plus, it does things for you. You can tell it, "Go check my inbox," and it summarizes everything important. Or, "Ping me on Telegram," and it goes and
does that. and it can work all night for you at 3:00 a.m. while you're asleep without you babysitting it. That's basically the pitch for Open Claw.
You run it on your own and it's yours. So, Abacus Claw is basically the same software without the rerunning it part. I gave it a pretty difficult
problem actually because if you go to the Apple website, they don't make it easy. It's not like each link has a separate page, separate URL. It's a
very JavaScript heavy type of website. But let's see. This is cool because I did not expect it to actually get this done because wow. But it did it.
Bad news, Alex. The Mac Mini is completely sold out everywhere as of today. We already knew this, but I'm impressed that the agent got it done. I have
a lot more faith in it now than I did before. Even I'm going to start a new session. Act as my content repurposing assistant. When I drop a blog post
draft or a link into Telegram, Agentbot automatically turn it into a Twitter post. Punchy conversational rewrite as a LinkedIn post. professional tone
and save both versions to a notion database called content pipeline. Boom. I don't have any connections set up yet. Let's see how far I can get. So,
it's telling me that we need a Telegram bot. So, we need a bot token and we need to configure notion integrations. Once we do that, it'll build the
skill. So, here on my phone, I've got Telegram. I searched for Botfather 10 million monthly users. Wow. Get some instructions. Start. And then we go
to new bot. Or I can just tap on that. Choose a name for your bot. A clawbot. Boom. Now choose a username for your bot. How about a clawbot? Boom.
Sorry, this username is taken already. Ah, aclawbot bot. There we go. It gave me a token. Keep this token secure and store it safely. Um, here's my
Telegram token. Wow. Telegram channel added already. Now needs a pairing code. Now, I know that this is still a bit of a setup, but it's guiding me
all the way through it. So, it's pretty easy. For Notion, I need three things. Notion API key. So, I went to my notion internal connections, create a
new connection, and I also get a token. Everything is called a token these days. You notice that? Here you go. Database ID from your content pipeline
database. Let's try that ID right there. Ah, well, that's handy. It tells me exactly how to connect Notion content database to my integration. Now, so
I go to my content pipeline connections and there's my claw right there. Add to page. Done. Boom. Now, while it's doing stuff, you can always go to
the dashboard here, and it opens up a new page on the right here, you have a little bit more insight into what's happening. So, for example, right now
it's using Miniax M2.7. But if you want to tweak a few things, and this is what's cool about Abacus AI also is you can use any one of these models
here that are available. These are top-of-the-line open- source models like Kimmy K2 and not even open source like Frontier models too like Claude
Opus 4.6 Gemini set the uh reasoning effort here. So you can definitely tweak and configure things as well. This is like the simple view and then you
have the background view if you want to dig into it. You can dig into it even more by pressing this button and then you get to view the computer
screen, the actual virtual machine that's running all this. cuz between you and me, it's it's a virtual machine running in the background. There it
is. It's a Linux box. Sh. Don't tell anybody that I got in there. I'm I feel like I'm hacking this thing or something. So, it's Ubuntu. We have
soul.md. So, let's go with cats soul.md. And this is basically the chatbot right here. This is like the soul of your clawbot. I'm not going to go into
details about the architecture of how open claw works, but you can read up on it. And this is it. This is that file. Basically, the entire setup.
Agents.mmd is here. Memory skills. You can update that if you want to. You can add additional things into it. So, you have a lot more control if you
want. One could say it's kind of like having one of these around, but you don't have to manage it. Good news. Telegram is already running and notion
database is fully connected. And look at this. It created this whole template for me without me having to manually configure all this. Ooh, I got to
try this with Google Docs and Google Sheets because that's notoriously difficult to work with through the API, but I bet our friend here can do it.
That's for another time. Let me know if you want to see that video. It's ready to go. How to use it. Drop a blog post into your Telegram bot either by
pasting the text directly or sending a link to the article. Really? Okay, here we go. Paste. Go. I basically pasted that 9 to5 Mac article about the
Mac minis being sold out cuz that was on my clipboard. My bot is answering me. Yeah, supplying chain issues continue to hammer Apple. Really? The base
M4 is completely gone from Apple store. Not just backorded, but currently unavailable. Period. And back in Notion, here we are. There's a test entry
that it created while I was setting things up. Date LinkedIn post. Let's see how professional this sounds. Applebase Mini is completely out of stock.
Blah blah. Yes, it's got some hashtags at the bottom there. source URL is there. And there's my Twitter. An entire thread. I would never post this, by
the way. Don't do this. This is just I hate it when people do this kind of slop. So, I'm not going to do that myself, but it's a good uh proof of
concept here. Okay. So, for most people, and I mean most people, not most people watching this channel, Abacus Claw is a good thing. It's a good
solution. Hosted Open Claw. The reason self-hosted agents mostly don't happen is that people don't want them. It's the gap between I'd like an agent
that checks my email and I have a running secure monitored 24/7 agent that checks my email. Now the security argument alone is worth sitting with. I
haven't tried I'm kind of a late comer to this I think considering my line of work and I have not set up open claw locally yet because I'm concerned
about security. an agent with access to your inbox, all your tools running on your local network, exposing everything to the internet on a box that's
under your desk, then that's a bigger target than most people are prepared to defend. And I'm not being dramatic about that. That's just true. It's
kind of early days and bad actors are going to take advantage of that. But if you're watching this channel, you're probably not most people. You're
probably the person who'd rather run the Mac Mini cluster than rent the cloud. You want your data on your hardware. You want to pick your own model.
You don't want a subscription because knowing is part of the point for you. Local still wins. It's more work, sure, and it fails in annoying and
spectacular ways. Sure, you'll spend a Saturday afternoon you don't plan to spend trying to tweak and fix things, but hey, at least you'll own it, and
nothing about it gets worse because a company you don't control changes its servers. What I'd actually recommend if you've never gotten a self-hosted
agent working end to end, Abacus Claw is a cheap way to find out whether agents are useful to you at all before you sink a weekend into setting up
Open Claw locally and trying to hunt down one of these on Craigslist. A few bucks a month for finding out is a much better deal than 40 hours of
finding out. And if you decide agents are useful, then go and build a local version and then you end up building it knowing what you actually want to
do with it. So that's the honest answer. Link is in the description. Go check out Abacus Claw and thanks to them for continuing to be a sponsor on the
channel and I'll see you in the next video.