The Gold Rush That Sells Shovels to Itself

Ihtesham Ali, writing on X:

I tracked 89 indie hackers building businesses around OpenClaw... 67% are generating revenue. 34% hit four figures in their first month.

This isn't about another AI tool. This is about the wealth transfer happening right now while most people are still asking "what's an AI agent?"

It is, in fact, about another AI tool.

The piece lists ten OpenClaw-based businesses allegedly "printing money." The revenue figures are precise - "$800-$1,500/month per client," "profit margins above 90%," "$8,400 across 7 clients in January" - but none are attributed to named sources, linked to evidence, or accompanied by any methodology for how "89 indie hackers" were tracked. We're asked to take the numbers on faith, which is a lot of faith for an article whose central thesis is "act now before you miss out."

What's instructive is the contradiction the article can't resolve. Points one through eight describe a thriving ecosystem of profitable businesses. Points nine and ten describe a platform where 7.1% of community-built skills leak API keys, 15% contain malicious instructions, and 18,000+ instances are exposed to the internet. The article treats this not as a reason for caution, but as a further business opportunity. The building is on fire, but have you considered selling fire extinguishers?

The "picks and shovels" insight - that you can make money servicing a platform rather than building on it - is legitimate. It was true during the actual gold rush. It was also true that most of the miners went broke. Articles like this never mention the failure rate, the client acquisition grind, the months of unpaid setup work, or the operator who built the agent and couldn't find a single paying customer. Survivorship bias isn't a bug in this genre. It's the entire format.

The closing line: "Send this to anyone still talking about ChatGPT instead of building." That's not advice. It's a forwarding mechanism. The article is engineered to spread, and the best way to engineer spread is to make sharing it feel like an act of generosity rather than what it actually is - free distribution for the author's brand.

None of this means AI agents are worthless. Automating client communication, repurposing content, and extracting structured data from the web are real, valuable applications. But an honest assessment of those applications wouldn't need to dress itself up as a breathless countdown of "money printers." It would name its sources, acknowledge its uncertainties, and trust the reader to make a decision without being told that the alternative is getting left behind.

The professionals I keep meeting don't need to be told the gold rush is on. They need someone to help them tell the gold from the pyrite. That's the harder article to write, which is probably why it doesn't go viral.