Real experiments only happen where failure is allowed. Drop the weight, the tries multiply, and unexpected paths open. Today we unpack this principle through an AI pet gacha.
A common mistake when picking up a new skill — you try to make "something proper." Portfolio-worthy. Boss-worthy. Business-valuable. And strangely, the moment you aim for that, a whole day passes without a single line written. Why?
This essay explains where that freeze comes from and how to break it. Even if you've never built anything, you can follow along — we'll go slowly. Today's example is a silly AI pet gacha, but the principle applies to any skill, any era. Three years from now, whatever Claude is renamed to, the spine of this essay still works.
Start with the principle. The biggest block when making something isn't skill. It's the weight of failure. "This has to succeed" — once that weight lands, you can't write line one. You're afraid it won't be perfect. So a week goes to research, three days to architecture, and the actual building never begins.
The opposite — a "failure is fine" project with zero weight — starts immediately. You build what comes to mind, delete if it's bad, build again. Five tries in an hour. Try-count determines learning. That's a law you can't break.
But somehow we forget this as adults. As kids, when the sandcastle fell, we rebuilt it. That's how we learned construction — in our bodies. As adults we dismiss "silly sandcastles" and study real architecture books until exhausted. Weight blocks learning. That's the paradox.
My story. One bored weekend, I asked Claude to build me an AI pet gacha. "No rerolls. One pull ever. Random from 20 pets. AI generates a cute name." Completely useless. Can't put it in a portfolio. Won't make money. Can't even brag to friends.
But in the two hours of building, I learned all this:
Doing this as a serious project would have taken days. I'd have read a book, followed tutorials, copied examples. But the "make something useless" frame meant failure didn't matter, and because failure didn't matter I tried anything, and the tries piled up into learning. I deleted the gacha from my server two days after launch. What stayed was five skills stuck to my fingers.
Picture a five-year-old building a sandcastle. The first one collapses. No tears — they rebuild. It collapses again. This time they pack the base harder. Collapses less. In that process the child learns the importance of foundations. No architecture book involved.
Tell that same child "this sandcastle has to be perfect — we're showing Dad," and watch what happens. The child can't start. They stop halfway and say "maybe later." Weight blocked the experiment.
A creator's projects work the same way.
Flip this principle and it's a bit shocking. The more useless, the better the learning. By the numbers:
| Project type | Avg tries/month | New skills picked up | Completion |
|---|---|---|---|
| Serious (portfolio-grade) | 1-2 | 1-2 | 30% |
| Middle (side project) | 5 | 3-4 | 60% |
| Light (goofing around) | 20 | 10+ | 90% |
Strange, right? Light projects finish more often. Because no weight means you ride it out. The portfolio one gets scrapped the second you mutter "this is mediocre." The goofing-around one is goofing-around anyway, so you finish it. Finishing forces mistakes, and mistakes force solutions.
Here's the first aha moment.
Light projects are what build real skill muscle.
Serious projects are clothes in a shop window. Pretty — can't actually wear them. Light projects are work clothes. They get dirty, but that's where the body actually moves. Skill comes from work clothes.
How to apply this? Ask yourself one question each week.
"What useless thing will I make this week?"
Truly useless. Three criteria.
No business value. Can't go on a résumé. No calculators, to-do lists, accounting apps. Pet gachas, mood guessers, weird-name generators, rate-my-day apps. That stuff.
Small scope, no weight. Define it as a "month-long project" and it gets heavy. "One evening's worth" is the right size.
Fun is required. A useless project without laughter is just wasted time. You should laugh when the pet gets a ridiculous name. Laughter holds attention.
One line to remember: Light. Often. Toss. Three words.
Today's principle in summary.
Heavy projects block starts. Light projects multiply tries. Tries make learning. We used an AI pet gacha as the example, but any skill, any tool works the same way. Python, drawing, instruments — all of it.
One useless thing a week. Un-showable to the boss. Done in two hours. Makes you laugh. In projects like that, the skill actually becomes yours.
Three years from now when Claude is renamed and Artifacts vanish, the principle keeps working. Technology changes. The principle doesn't.