Shipping Stopped Being Evidence of Anything

14,700 new subscription apps launched in January 2026 alone. The number of new apps globally rose about 60% in the first quarter against the same period a year earlier, with the iOS store growing by as much as 80%.

Set against that, one figure does more work than the rest: apps launched during the 2025 vibe coding era account for roughly 3% of revenue.

Why the 3% matters more than the 14,700

The launch volume tells you building got cheap, which everyone already knew. The revenue share tells you what happened next, which is that almost none of it worked.

That gap is not a comment on quality. Plenty of those products function correctly. It is a comment on what launching now proves, which is nothing. For most of software's history, shipping was expensive enough to be a costly signal: getting a product into the market meant somebody had committed real money and real time, which meant somebody had at least been forced to think about whether it was worth doing.

Remove the cost and the signal goes with it. A launched product is now evidence that a person had an idea on a Tuesday.

The retention side

“Launching used to be a costly signal. Remove the cost and the signal goes with it. A launched product is now evidence that a person had an idea on a Tuesday.”

Many AI apps lose 79% of annual subscribers before month twelve, generally because the product met a short-term curiosity rather than a recurring need.

This is the same finding from the other end. Cheap building produces a lot of products that are pleasant to try once, because a thing that is pleasant to try once is the easiest kind of thing to build. Products people return to weekly are hard, and they were hard before the tooling changed.

Launching used to be a costly signal.

Where the saturation actually is

Five categories are conspicuously oversubscribed: AI writing assistants, which have more than a hundred funded competitors, customer support chatbots, meeting summarisers, logo generators, and resume builders.

What those have in common is that they are the products you think of when you have a general capability and go looking for something to apply it to. They are the output of starting from the tool. Everyone starting from the tool arrives at the same short list, which is why the list is crowded.

The categories with the least competition run the other way: compliance tooling, vertical software for unglamorous industries like HVAC, pest control and roofing, agent infrastructure, senior care technology, and fintech infrastructure. You do not reach any of those by starting from a capability. You reach them by starting from a problem you have personally watched go unsolved.

What replaces shipping as evidence

Someone paying before the thing exists. That is the only signal that got more valuable as building got cheaper, precisely because it is the one thing the tooling cannot manufacture.

It does not have to be much. A deposit, a paid pilot, a signed letter of intent, a pre-order from three people who have the problem. The amount matters far less than the fact that money moved before you built anything, because money moving is the only evidence that survives contact with the fact that everyone can now build.

The corollary is uncomfortable for anyone enjoying the building part. If you cannot get a single person to commit anything before you start, the correct conclusion is usually not that they need to see it first.

See zero dollar validation experiment templates for the cheap versions, kill criteria before you start for deciding in advance what would make you stop, and where capital is crowding for the funded end of the same map. The waitlist is where the structured version of this runs.