I used to have a folder. 📁 Mental, mostly. Sometimes a notes file. Ideas I'd build "when I got a weekend." Then a quieter weekend. Then maybe a holiday.
Most of them sat there for years. A few sat there long enough that the problem they solved had already been solved by someone else, twice.
Something has shifted in the last year or so.
The math on testing an idea has changed
The whole vibe-coding wave — Cursor, Claude Code, the rest of it — has changed the math on what an idea costs to test. Not write a thinkpiece about. Actually build, in a way you can click on.
With the right tools, and what I've come to think of as strict prompting — mostly being annoyingly specific about stack and architectural choices upfront — I can have a working MVP of most ideas in a day or two.
Not a polished product. Not something I'd put my name on yet. But enough to actually feel the shape of the thing instead of imagining it.
The strange part isn't the speed
It's what the speed reveals.
I'm a hands-on type. I don't really learn why an idea is bad by thinking about it harder. I learn by building far enough into it to hit the wall. 🧱 And now I hit the wall faster.
Sometimes the wall is that the data I assumed existed publicly, doesn't. Sometimes it's a regulatory requirement I had no idea applied to whatever niche I'd wandered into. Sometimes it's just that the problem dissolves the moment you try to describe it to a real user.
Roughly 99.7% of my ideas turn out to be bad
Or unrealistic. Or solving a problem that already has six perfectly fine solutions I just hadn't bothered to look up.
I knew this in theory. I just never had to confront it before, because the ideas were safely parked in the drawer where they could remain potentially brilliant forever.
Now they meet reality on a Tuesday evening, and reality is usually unimpressed.
The remaining 0.3%
The ones that survive are worth more than I thought, though. Because they survive contact with an actual prototype — not contact with my optimism.
That's a different filter than "this would be cool to build." It's "this is still interesting after I've seen the messy data, the unknown regulation, the user who shrugs when I explain it."
The drawer is mostly empty now. I'm not sure if that's progress or just a faster way to be disappointed in myself.
Probably both.