I Published 283 Apify Actors. People Mostly Run the Boring Ones.

Nobody sets out to build 283 of anything

My Apify profile says I joined in August 2025. It now lists 283 public actors. That is not a plan. It happens one client question at a time, each one ending in some version of "can we pull that automatically," and the answer keeps being yes. A year later you have a fleet, and a showcase site with a battle-arsenal theme nobody asked for.

So I pulled the numbers for every one of them from Apify's public store API to see what people actually do with a fleet like that. The answer is humbling, and if you are about to publish your first actor, it will save you a year.

183,095 runs. A third of them belong to ten actors.

All 283 actors together: 8,584 total users and 183,095 runs. My profile says 93.2 percent of runs succeed. The median actor has 9 users. Nineteen actors have one user or none, and an actor with one user is not a product. It is a diary.

The top ten actors by runs account for 60,444 of them, 33 percent of everything. The other 273 split the rest. If you have ever wondered what a power law feels like from the inside, it feels like maintaining 273 things for the love of it.

The flashy actors get tried. The boring ones get scheduled.

The actor with the most users is Influencer Discovery: 412 people, 1,389 runs, a little over three runs each. The AI image generator has 304 users and 1,413 runs, under five each. People show up, try it, get a picture, and leave. That is a demo, and a demo is a fine thing to be. It is just not a business.

Now sort by runs. First place is Tech Stack Detector, a BuiltWith alternative: 12,527 runs from 182 users, about 69 each. Second is the BBB Scraper: 10,824 runs from 169 users, about 64 each. Third is Company Contact Enricher: 7,312 runs from 168 users, about 44 each. Nobody clicks Run 69 times by hand. Those are pipelines, calling the actor on a schedule, feeding a column somebody needs filled every day.

The BBB one made me laugh, because in my entity association test Perplexity answered "is this company legit" by citing the BBB five times. Somewhere, 169 people are running the same witness check the AI engines run, at scale, on purpose. They were right before the robots were.

What I would tell anyone about to publish an actor

One. Build the thing that goes inside somebody else's pipeline. One input, like a URL or a company name. One output, clean JSON, the same shape every run. The actors that win are the ones a developer can wire up once and forget.

Two. Charge for results, not effort. All 283 of mine use pay-per-event pricing, so a user pays for the records or actions they got, not for how long my code thought about it. It puts my incentive in the right place. The 6.8 percent of runs that fail is where the real work lives: blocking, layout changes, and inputs I would never have imagined anyone typing.

Three. Assume the caller is a machine. Every actor I publish can post results to a webhook, which is how they end up inside Zapier and Make pipelines. Eighteen of the 283 have MCP in the name, and Apify runs an MCP server that lets an AI agent find and call Store actors directly. The next heavy user of your actor is probably not a person.

Nobody wants a scraper. They want a column filled in.

The actors that get used answer one question, in a format a machine can read tomorrow and every day after. The rest are demos, and the fleet needs a few of those too, because demos are how people find the boring ones.

If you want to build your own, or just run one of mine, start at Apify. My whole fleet is on my Store profile, and yes, the Tech Stack Detector is still the most exciting boring thing I have ever made.