The story of this chart is the people who made the apps, not the users who didn't show up
If a tree falls in the middle of the forest and nobody's around to hear it, does it even make a sound? If somebody who wouldn't have made an iOS app in 2023 makes one in 2026 but then nobody even uses it, is it an app?
Think about all those people and what they're experiencing right now. They weren't making iOS apps in 2023, so they're normal people. They were probably pretty nervous to try AI in the first place, but they did try it, and they got something out of it. They liked the thing they made enough to fill out some forms and pay $100.
Now their app is out in the world and nobody's using it. They're probably feeling some disappointment. They also probably are reflecting on, and integrating what they learned. They'll probably keep using AI. Next time they'll either invest more in distribution, or embrace forever-prototyping and scale back their ambitions (like I have).
They're probably not going to pay a fee next time, or fill out a ton of forms. Maybe what they build won't even really be an app.
Application is a weird word
What even is an 'app'?
I remember the first time I wondered this the question filled me with a immediate terror that I'd uncovered yet another basic, obvious Computer-Thing I didn't at all understand.
It was refreshing, then, to slowly realize I wasn't missing something. An app is just a weird distinction with inconsistent, strange boundaries.
It was particularly fun to read about the early days of Microsoft and Oracle, and how strange the very idea of software was back then. You're telling me I have a computer, that can do anything and was very expensive...and you're going to sell me lines of code that turn my anything machine into a one thing machine? Ridiculous.
Hardware -> Firmware -> Software -> Slop
Slop isn't software. It's softer. If we hadn't called software software, maybe software would be a good word for it but that one's taken.
Slop even flows, kinda. I'd say it's a consistency similar to ketchup. If you can define a container you almost just...turn on the spigot and watch it slowly fill up.
It looks like software, It compiles like software. But brother? It ain't software.
Last analogy: what food do you bring on an ocean voyage vs what do you want for lunch?
Stewart Brand's Maintenance is (predictably) quite good! I just finished the first bit, which frames maintenance through solo round-the-world sailing.
Think about what you'd bring to eat if you're packing for a 10-month sailing trip, and how the constraints would change what you could bring.
It can't go bad.
It will get wet (with saltwater).
It has to be healthy enough.
The more compact the better.
Given all that, it's going to be hard to also get the nice-to-haves of variety and yummy.
Now think about the constraints of making software without LLMs, to be used by people who also don't have LLMs and don't even really know what an array is.
Tough to make anything that's any good! You have to painstakingly filling your short term memory with a bunch of very local knowledge of a complicated system. And it better not break!
If there's a bug? If your software starts misbehaving for that array-ignorant user? They're gonna be so mad there's gonna need to be a whole job called Customer Support, a whole profession of people who's whole thing is just being a layer between that visceral anger and the people who make the software.
In 2026, from first principles, wouldn't it make a lot more sense to try to build software that could do this:
What you'd build for the user and their own LLM to riff on and improve is very different from what you'd build to be bullet-proof and never break unexpectedly. Different enough that to even try to compare it to software would be unfair to both software and slop. That's the story of the chart.
There's an exponential flood of ...something!... coming into the world and the people making it are naively trying to shove it into software-sized holes.
What should we do with all this slop? What should we be trying to make if not software, but faster?
Well...what are you hungry for? What would hit the spot right now? What sounds good not for the next ten months, but just for lunch today? Make that.