OpenClaw is the first AI agent I’ve seen that makes me think “okay, this is actually something.” Persistent background agent, open source, runs locally, talks to your apps, executes real tasks. Peter Steinberger built something that crossed the line from impressive demo to thing-you-actually-want-running. That’s rare and it deserves credit.

Now here’s what’s happening around it.

People are buying Mac minis they don’t need. Delivery times on high-memory configs have stretched to six weeks. Raspberry Pi stock is getting pulled into the hype. There’s a cottage industry of setup guides, YouTube walkthroughs, and paid courses - all focused on getting OpenClaw running. API providers are watching credits burn 24/7 from agents that are… monitoring things. Doing stuff. Supposedly.

And malware authors are already targeting OpenClaw config files and gateway tokens, because of course they are. Where there’s a stampede, there’s a pickpocket.

The most telling signal: the most popular OpenClaw content right now is about how to set it up. Not what to build with it. Not problems people solved. How to install it.

FOMO. Not utility.

The Spending Doesn’t Make Sense Yet⌗

Let’s do some quick math. A Mac mini with enough memory to run OpenClaw comfortably costs $500-1,500. On top of that, you’re paying API costs for whatever frontier model your agent is talking to. Persistent agents aren’t cheap - they’re making calls constantly, and at frontier model pricing, that adds up to real money fast. Some people are reporting costs north of a dollar per interaction for basic tasks.

So you’ve got hardware costs plus ongoing API costs plus the time you spent configuring everything. For what? What’s the workflow that justifies this?

For some people, the answer is real. If you had a genuine automation problem before OpenClaw existed - something you were already trying to solve - then this is a powerful new option. Those people aren’t the ones driving Mac mini shortages.

The shortages are being driven by people who saw a demo, felt the anxiety, and bought hardware the same week. No use case in mind. Just the fear of being left behind.

Who’s Actually Making Money⌗

Follow the money and the picture gets clear fast.

Apple is selling Macs to people who already had perfectly good computers. API providers are collecting credits from always-on agents. Content creators are selling setup guides and courses. The whole ecosystem around OpenClaw is thriving.

Notice who’s not on that list: the person who bought the Mac mini. They’re the customer in every direction. They bought the hardware, they’re paying the API costs, and they bought the course that told them to do both.

OpenClaw the product is good. The gold rush economy forming around it is extractive. Don’t confuse the two.

The Discovery Argument⌗

Here’s the counterpoint, and it’s a real one: tinkering without a clear goal is how a lot of breakthroughs happen.

The early web was full of people building things nobody asked for. Most of those things died. Some became actual businesses. You can’t always know in advance which experiments will produce value. Sometimes you just have to play with the tools and see what happens.

I believe this. I’ve written about it before. The collapse of execution barriers means more people experimenting, which means more surface area for genuine discovery.

But there’s a difference between discovery and anxiety.

Discovery looks like: “I have this recurring annoyance in my workflow. I wonder if a persistent agent could handle it.” You’re starting from a problem and exploring whether the tool fits.

Anxiety looks like: “Everyone is setting this up and I don’t even know what it does but I need to have it running by this weekend.” You’re starting from fear and working backwards toward a justification.

Most of what I’m seeing right now is the second one. And the problem with anxiety-driven tinkering is that people bail before they find anything useful. The motivation was never curiosity. It was the fear of missing out. Once the fear fades - or the next shiny thing shows up - the Mac mini goes back to collecting dust.

The Taste Problem Again⌗

This connects to something I keep coming back to: AI tools are only as good as the person’s ability to point them at something worth doing.

OpenClaw can monitor your email, browse the web, run scripts, manage your calendar, and coordinate across apps. That’s a lot of capability. But capability without direction is just electricity. The agent will happily run 24/7 doing whatever you tell it to do. The question is whether what you told it to do actually matters.

The people who’ll get lasting value from OpenClaw are the ones who already had a clear picture of what they needed automated. They had the taste - the judgment about what’s worth building - before the tool showed up.

Everyone else is configuring an agent and then waiting for it to become useful. Cart before the horse.

The Pattern⌗

We’ve seen this exact cycle at the enterprise level. Impressive capability gets announced. Spending surge follows. Then a slow, quiet correction as people figure out what actually works versus what just looked good in a demo.

OpenClaw is this cycle playing out at the consumer level, in real time, with people’s personal money instead of corporate budgets. It’s faster and messier and more visible because it’s happening on YouTube and Twitter instead of in boardrooms.

The technology is real. The capability is real. A subset of people will find genuinely transformative uses for persistent local agents. I’d bet on it.

But the ratio of spending to value creation right now is wildly inverted. Most of the money being spent on OpenClaw this month isn’t buying productivity. It’s buying anxiety relief.

If you have a real problem and OpenClaw solves it - go. Don’t wait. The tool is good and it’s only getting better.

If you’re buying a Mac mini because a YouTube thumbnail made you feel behind, sit with that feeling for a week. See if it passes. It probably will. And the Mac mini will still be there if it doesn’t.