Lately I’ve been running into a problem I didn’t anticipate: AI doesn’t just help you build faster—it multiplies how many things you try to build.
One side project turns into two. Each of those spawns a “small reusable library.” Those turn into tools. Those tools deserve their own apps.
And suddenly you’re not building a side project—you’re inside a fractal.
This coming Tuesday marks six months since I started seeing Claude Code every day. Please, whatever you do, don’t tell my wife. These next few posts are reflections and cautionary tales.
Lately my main takeaway has been about pacing yourself and saying, “No Claude, let’s not build a standalone library for this utility app.”
Perhaps it’s “Claude creep” or maybe just ADHD “shiny thing” syndrome, but I’ve found the ideas keep coming and for each app I start to build, I find a need for a cool library or tool that no one’s built (or built “right”) yet.
I started out just trying to finish a pre-Claude CRUD app with a pay tier—my longtime evening/weekends white whale.
Then I realized I could build the lightweight notetaking app I didn’t know I needed.
Then I might as well move those components into a free component library. Which should probably ship with a CLI.
And then—did you know no one’s built a free scrollable color palette + font picker with live reload that embeds into your dev site?
And also no one’s built an isometric voxel floating cloud visualizer for Claude subagents.
And actually—there’s no good lightweight isometric game engine right now?
And I really need to redo my portfolio site.
But I’ll also need a clean studio site…
Not because you can’t build it—
but because you can build too much.
When every project branches out to three more…
you’re in an infinite, recursive fractal.
Sometimes you have to say—
“Uh uh, Claude. Not today.”