Over-engineering a solution for a non-existent problem is a rite

I've realized that the danger starts when we mistake personal enjoyment for product-market fit. It's easy to spend a month building a complex AI workflow that automates a task taking five minutes a year, then convince ourselves it's a "game-changer."

To stop this cycle of over-engineering, I've started running every new idea through a technical audit:
1. Who specifically has this problem?
2. What is the current manual workaround?
3. Is the friction of switching to my tool lower than the pain of the problem?
4. Am I building this for a user, or just to play with a new library?

Both paths are valid—hobby projects are where we actually master new tech—but being honest about the goal prevents burnout.

I'm looking for examples of "technically impressive but useless" builds. Specifically, things that worked perfectly from a deployment standpoint but solved a problem that didn't actually exist. What was the stack, what did it actually do, and what was the specific technical lesson you took away from the failure?