On the timeline · a founder exchange
The app layer is getting removed.
Better product, less revenue, and a customer who just cloned the app. Three founders, one exchange on X, worth watching in the order it happened.
Better product, less revenue, and a customer who just cloned the app. Three founders, one exchange on X, worth watching in the order it happened.
In July 2026 three founders talked past each other in public, and in doing so drew the clearest picture yet of what AI is doing to software businesses. Here it is, post by post, in the order it landed.
It opened with a number that isn’t supposed to happen.
Matt Welter (@_mattwelter) builds ReelFarm, a tool that grows apps by auto-posting slideshow videos. A year ago it made him $48,000 in a month. The product is far better now. The revenue isn’t.
Then one of his own customers answered the question.
Pedro Machado (@pedrotech_) had loved ReelFarm, it took his app from zero to $5k. And then he said the quiet part out loud.
Pieter Levels quote-tweeted the whole thing and named the pattern.
Levels (@levelsio) runs a portfolio of AI apps doing millions a year. His read: the product did get 10× better, but so did every competitor’s, at the same moment, with the same AI. Read it in full below, it’s the heart of the whole thing.
His argument compresses into one picture: when a customer can clone the app and pay only for the raw AI tokens underneath, the middle of the business chain falls out.
“The app layer is possibly getting removed!”
Then he pushed it one step further.
Also some thing I forgot to add
— @levelsio (@levelsio) July 20, 2026
There's so many things now where I don't even use the app anymore but just generate my own app to do things
Like WHOOP, for years I wanted to know the correlation between stuff I do and my workouts and sleep, like going to sauna, or what I eat.… https://t.co/AJznJTSM0h
Neither founder at the center let the doomsday read stand.
Matt drew a boundary around it, the app layer dissolves for undifferentiated B2B, not for consumer products people would never rebuild:
And Pedro, the cloner, added the most honest footnote of all, that copying it was possible but not actually smart:
If AI hands every competitor the same 10× and every customer a plausible clone button, the feature stops being the moat.
What’s left is the part Pedro proved by accident: he could rebuild ReelFarm, but its craft and niche made paying the smarter move. Distribution, taste, and a niche you own that well, that’s the moat now.
Pieter Levels’ posts are embedded live from X. Matt’s and Pedro’s posts are shown as captures of the originals. Every word is theirs, reproduced unedited.