Fair point on where the trajectory might seem to be heading.
If we reach a stage where anyone can effortlessly prompt any bespoke piece of software into existence, the perceived utility of sharing simple hobby projects definitely changes.
However, I think there’s a crucial distinction between something being "made by AI" and being "accelerated by AI":
* When we see projects like the recent viral clean-room Adobe clones or browser ports, they aren't the result of an AI just conjuring complex software out of thin air. There is still a human sitting there making architectural decisions, defining boundaries, separating good code from bad, and fixing edge cases. The LLM acts as an incredible accelerator for boilerplate, but the intent and system design still come from a developer. A simple prompt like "build me a shooter" or "clone this complex app" still falls apart 99.9% of the time.
* And when it comes to games specifically, the code is only half the story. The rest is taste, pacing, subtle game-feel, and constraints. When you prompt something for yourself, you only ever get a reflection of what you already expected. Playing someone else's game - even a tiny, flawed hobby project - is about experiencing someone else's vision and surprises.
Maybe the web will be saturated with instantly generated disposable software, but I'd argue that makes deliberate, opinionated human craft more refreshing to encounter, not less.