OpenClaw — Stop Buying Mac Minis

I just got back from an OpenClaw event. Around 1,000 people in the room. Everyone was interested in AI and OpenClaw but I’d say less than 5% actually knew what it was.

They ended up auctioning off a Mac Mini because that’s what everyone’s setting up OpenClaw on supposedly at the moment and someone paid $20,000 for it.

TWENTY THOUSAND. For a Mac Mini.

In that moment it propelled me to come out and write this blog because I realised no one actually understands what it is — I’d bet over 95% of people there — couldn’t even tell you what OpenClaw actually does, let alone set it up themselves.

So let me break it down.

---

What Is OpenClaw, Actually?

OpenClaw is open source software that tries to turn any computer into a personal AI agent hub. You install it on a computer environment, connect it to any AI model, and you can set it up to browse the web, pull information from platforms, run tasks, and interact with your tools on your behalf; but this has been possible with terminal AI coding agents which I’ll get into in a bit. The main moat is its gateway however, it’s ability to interact with the AI you setup over Telegram, WhatsApp, Messenger etc…

It was created by Peter Steinberg — a guy who was retired and came back because he heard of these AI coding agents and its ability to code and push software much more efficiently now. He published and developed over 43 projects on his github in the last year that he has come back to develop software before OpenClaw finally caught fire. OpenClaw has now racked up 190,000 GitHub stars in under two weeks, making it the fastest-growing open source project we’ve ever seen.

The software itself is genuinely impressive. But you don’t need to spend $ on a Mac Mini to run it. Any old laptop will do — any computer with an OS can run it.

---

My Setup

I spent a weekend getting OpenClaw running on a spare laptop I don’t use anymore. As you can see above, I have it sitting on my desk, docked and running 24/7 in clamshell.

Here’s my setup:

• Hardware: Old laptop, always on, charged and docked at home
• Model: Open source AI model that’s running through OpenRouter

Nothing fancy. No need to purchase a Mac Mini — in fact you can run it on a Raspberry Pi, a small hand-sized computer.

---

What I Actually Use It For — Honestly

I’ll be straight with you — I’m not using OpenClaw for as much as you might think. A lot of people online are making it sound like it’s running their entire life. For me, it’s more of an environment to run things on.

I originally planned to use it for code reviews — having it hooked up to my repos so it could review code I push autonomously. But then Claude Code released their remote access feature, and I just remotely access my Claude Code session from the same environment I already had OpenClaw running on. And in general all the software that I setup for OpenClaw I build with Claude Code anyways rather than building it out and trying to brute force build it through the AI model I have setup through the OpenClaw gateway.

And that’s the thing like you build out the infrastructure at the end of the day when setting OpenClaw up. You want to connect to Xero, G Drive, Meta ads whatever? You will probably need to build that out OpenClaw out of the box has some integrations but my experience is unreliable and most likely won’t work with the use case you want. You ideally for example want to build out a skill or write a file that says use this command line tool that uses Xero API to talk to xero if that makes sense? But yeah at the end of the day you need technical understanding of what you’re setting up to actually make it work but you can try brute force setup and tell Claude Code what you want that’s the route I’d go if I were not technical. Go back and forth with Claude Code on that environment, explain, give docs on what OpenClaw is and say this is what I want; What needs to happen?

---

The Risks Nobody’s Talking About

Here’s what bothers me about the OpenClaw hype — nobody is talking about the risks.

As mentioned you can connect OpenClaw to a lot of different things. APIs, platforms, services, your own tools. That’s the whole point. But on top of this it also comes with inherent risk. When you give an AI agent the ability to act on your behalf — browsing, clicking, submitting forms, making requests — you’re trusting it to do the right thing. And right now, there’s no real way to measure what’s happening in the backend. You can’t easily audit every request the LLM is making, every action it’s taking, every piece of data it’s accessing.

That’s a real problem. Especially if you’re connecting it to anything sensitive.

And again on top of all that — you need genuine technical understanding to set this up properly. Connecting APIs, configuring models, routing data between platforms, debugging when things inevitably break, consistently managing the environment etc… If you don’t have technical understanding, OpenClaw is a waste of your time in my opinion you’re better off learning how to use Claude Code, Codex, Claude CoWork, Claude Excel etc… I wouldn’t recommend going down OpenClaw path. You’ll end up frustrated and realise you wasted a whole lotta time.

---

The Bigger Picture — Software Has Changed

Here’s where the OpenClaw story gets interesting, and it’s not about the software itself.

You may know the pain of creating software it used to take months — sometimes years — to produce software. Full development teams, project managers, QA cycles, deployment pipelines. Building anything meaningful required serious time, money, and expertise.

That’s changed. Dramatically.

Peter Steinberg was retired. He came back and built 43 projects in a year because, in his words, creating software now gives you a dopamine hit. I was a third-year computer science undergrad and I’ve solo-built a full AI-powered SaaS with real paying users. That would have been unthinkable 2 years ago — even for a senior developer with a couple decades of experience.

If you have technical expertise, the tools available today mean you can build software way faster what they are calling inference speed.

And that changes everything for businesses.

---

Why Every Business Needs a Custom Tech Stack

If software is this much easier to build, why is your business still paying thousands a month for off-the-shelf SaaS tools that doesn’t even really quite fit?

Think about it. You’re paying endless subscription fees for software that was built for everyone — which means it was built for no one in particular. You’re locked into their pricing, their feature roadmap, their limitations. Every month, the bill goes up. Every year, you’re more dependent on platforms you don’t own or control.

Now imagine the alternative. A tech stack that’s fully custom, built specifically for your business. Tailored to your workflows. Designed around how your team actually operates. Something you own, can manipulate however you want, and never pay a recurring licence fee for and that your business entirely owns.

That’s not a pipe dream anymore. That’s the reality we’re living in. These tools are fundamentally changing the software game you can now build from scratch internal tools, automations, and systems that replace the bloated software stack your business is currently bleeding money on.

---

This Isn’t Optional Anymore

Let me be blunt. If your business isn’t thinking about this right now, you’re going to get burnt.

The businesses that move now — the ones that start building custom tools, integrating AI into their workflows, and replacing expensive SaaS subscriptions with systems they actually own — are the ones that will have an unfair advantage 12 months from now.

The ones that sit back and wait? They’ll be paying more for less while their competitors operate leaner, faster, and smarter.

Sam Altman (CEO of OpenAI): “AI won’t replace you. But people who don’t use AI will be replaced by people who do.”

This is not a threat. It’s the reality of where we’re heading.

---

What To Do About It

This is exactly what we do at Automate Wise. We help businesses cut through the noise, implement the right AI tools, and build custom systems that replace expensive software stacks — in weeks, not months. No fluff. No generic chatbots. Real modern tech that moves the needle.

If you’re watching all of this from the sidelines wondering where to start, look at the kinds of systems we’re already building and then book a conversation. That’s how you can move from “AI is interesting” to “AI is saving us money and making us faster.”