The Comfortable Lie I Told Myself
Whenever someone asks why I build my own tools instead of just using what's out there, I used to have a tidy answer ready: to save time. It's the answer people expect, and it sounds responsible — a busy person automating away the busywork.
It's also not really true. If I'm honest, half the tools I've built took longer to make than they'll ever save me. Nobody optimizing for time would spend a weekend building a thing to replace a task that takes ten minutes a month.
So if it isn't about time, what is it about? It took me an embarrassingly long while to name it. I don't build tools to save time. I build them to force myself to figure out what I actually think.
You Can't Build a Tool for an Idea You Only Half-Understand
Here's the thing about an opinion in your head: it's allowed to be vague. I can believe "people worry too much about a single retirement number" and never once be forced to say what I'd put in its place. The belief floats there, comfortable and unexamined, sounding wiser than it is.
Then I try to build a tool around it, and the vagueness collapses instantly. A form field can't hold a mood. A rule can't run on "it depends." The moment I try to turn the intuition into something that actually works, every soft spot in my thinking lights up like a warning panel.
Building is the most honest editor I've ever had. Writing lets you hide behind a good sentence. A working tool doesn't — either it does the thing or it doesn't, and if it doesn't, it's usually because I never really understood the thing in the first place.
The Question Every Tool Drags Out of Me
The clearest example is the Retirement Runway Snapshot. I wanted a tool that told someone, in plain English, roughly where they stood — no scary number, no advice. Simple enough as a wish.
Except the second I started building, it asked me a question I'd been happily dodging for years: what does "on track" actually mean? On track for what? Measured how? Against whose assumptions? I couldn't write a single line of that tool's logic until I'd answered it for myself. The tool refused to let me stay vague.
That happens every single time. The tool becomes a series of pointed questions I can't wave away, and answering them is the education. By the time it works, I understand the problem in a way I simply didn't before I started — not because I read more, but because I had to decide.
Building Is Thinking You Can't Fake
When people picture "using AI to build something," they imagine the output — the app, the automation, the finished thing. What they miss is everything the building demands of you along the way. Every small tool quietly forces you to commit to answers you'd otherwise leave blurry:
The decisions a tool won't let you skip:
- What exactly is this for — in one sentence, not a paragraph of hedging?
- What's the smallest set of inputs that actually matters, and what am I brave enough to leave out?
- Where's the line between helpful and harmful — what must this refuse to do?
- What does "good enough to be useful" look like, concretely, today?
- What do I genuinely believe here, once "it depends" is no longer an allowed answer?
Answer those five honestly and you understand your own problem better than most people ever bother to. That's not a side effect of building — for me it's the entire return. The app is just the receipt that proves I did the thinking.
Why This Only Became Possible Lately
I couldn't have worked this way five years ago, and that's the part I think people are still underrating. Building used to be far too expensive to use as a thinking tool. If standing up a working app cost weeks of setup and code, you'd only ever build things you were already sure about. The cost forced you to resolve your thinking before you built — exactly backwards from how understanding actually works.
AI flipped that. When I can take a half-formed idea from "I wonder…" to a working prototype in an afternoon — a static page, one webhook, a careful prompt — building becomes cheap enough to use as a way of reasoning, not just shipping. I can now build to find out what I think, instead of only building things I've already concluded.
That's a genuinely new privilege, and it isn't reserved for engineers. Anyone willing to wrestle an idea into a working shape — a small-business owner, an advisor, a curious generalist — can now use building itself as a tool for thinking clearly. That, quietly, might be the most important thing AI changed for people like me.
🛠️ See what the thinking produced
The Retirement Runway Snapshot is what came out the other side of "what does on track even mean?" — a plain-English read, no scary number, no advice. It's the receipt for a lot of forced clarity.
Try the Snapshot →Build the Thing to Learn the Thing
If you take one idea from this, let it be this: the next time you want to understand something properly, don't just read about it — try to build the smallest possible tool around it. You'll hit the exact spot where your understanding runs out, usually within the first hour, and that spot is where the real learning starts.
You don't have to ship it. You don't even have to finish it. The value already happened the moment the vague idea in your head was forced to become a specific decision on the screen. That's the trade I keep making, over and over: a weekend of building for a level of clarity I couldn't get any other way. It's the best deal in my entire toolkit.