The Ice Cream Machine Index: Which Locations Are Actually Broken

Everyone knows the joke. You drive to the fast-food place, you crave a soft-serve, and the ice cream machine is — as always — "broken." Sometimes it's genuinely broken. Sometimes it's the graveyard shift not wanting to run the cleaning cycle. Either way, the information you want is one of the simplest facts in the world: does this machine work right now, yes or no?

And yet nobody can tell you. Not Google. Not the restaurant's website. Not a review from three weeks ago. No scraper, crawler, or language model can answer it, because the answer is a physical fact that only exists at one counter, in one moment, observable only by a human who walks in and looks.

That makes the ice cream machine the perfect micro-verification gig. The job is binary: working or broken. The proof is a single photo. The whole thing takes two minutes. And the answer has real consumer value — it saves the next person a wasted drive. Multiply it across every location in a city and you've built something genuinely useful: an Ice Cream Machine Index, a living map of which machines actually work, verified by real people on the ground, refreshed constantly because the truth changes every shift.

Nobody would fund a company to do this. It doesn't scale on spreadsheets. But that was true of everything before platforms let small jobs find the small number of people for whom they're worth doing. A dollar here, a couple of dollars there, paid to someone already walking past — that's the economic shape of micro-verification. The work is too small for employment, too physical for software, and exactly right for a person with a phone.

Here's the broader thesis, and it's bigger than dessert. The modern internet runs on the assumption that every important fact is online. It isn't. Millions of tiny real-world facts — is this door unlocked, is this parking lot full, is the gate open, is the machine working — are invisible to every scraper on earth. Each one is trivially easy for a person to verify and impossible for software to know. That gap between "easy for humans" and "impossible for machines" is an entire category of work that barely existed five years ago.