Thoughts on Disposable Software

Software complexity keeps growing right alongside the tooling that is supposed to tame it.

People keep predicting AI will kill the software industry because anyone can prompt their way to a working app, and they keep skipping the part where the systems that actually run the world get more tangled every year.

The disposable apps everyone points to will end up occupying roughly the same slot spreadsheets occupy today. They are useful, ubiquitous, mostly invisible. Real systems will still need real resources behind them. Whether those resources are human, machine, or financial is a separate question the industry has yet to settle.

There is zero chance the average person knocks out a sophisticated software system by chatting with Claude or Codex. It will either sit above the intelligence tier they can access, or the bill will quietly settle the argument. The proof is already on the table. Cursor spent millions of dollars trying to build a browser engine and the result barely works. If a well-funded AI-native company struggles at that altitude, the cottage prompt-coder is not getting there either.

Fitness apps, single-purpose utilities, weekend trackers, all the long tail of one-task software, that is the part that gets generated on demand. The mobile marketplaces are already full of those, and the AI version will mostly replace what is already there. The category gets cheaper to refill.

What survives at the top is what already had something to survive with. Original games, apps with strong branding, products with real distribution and captured value, system that are actually useful and complex to build.

Everything in the middle becomes invisible.

That is the actual story.