Six years ago we got a new wall charger for our Nissan Leaf. You could set the speed, and delay when it started.
I was trying to keep the battery between 80 and 90 percent. A lithium battery that sits near full every night ages faster.
Every night, my wife would ask me what delay to set. I’d stand in the garage working out how much charge we needed by morning and when the charger should start so we didn’t overshoot. Usually it was fine. Sometimes I’d forget to answer, and we’d wake up to a car that hadn’t charged.
I kept thinking about building it. The math was easy enough to write down, but a tool you actually reach for in the garage, on your phone, that’s polished enough to use instead of calculating from scratch — that’s a lot of work. Especially if you’re imagining other people using it too. The ROI never quite tipped. I’d wish I had it, then balk at the weekends it would take.
We had a backup car, so forgetting to charge wasn’t catastrophic. But it still nagged me. Some mornings I’d have range anxiety when the car was actually fine. Other nights we’d be lazy, just plug in and let it sit at 100 percent, knowing it wasn’t great for the battery.
I’ve been using Claude for work and for side projects for a while now, mostly larger things. Businesses, tools, substantial builds. This problem was always in a different category. A small itch. Never worth the time.
This isn’t a story about discovering AI and how it’s great for building things. It’s about noticing that the calculation for “not worth the time” has quietly changed. Building a small tool now fits inside a couple of hours. The threshold for what’s worth doing has dropped.
One evening, after doing the same mental math I’d done hundreds of times before, I just sat down and decided to build it.
The logic had been with me for years. I just had to describe it, piece by piece, and watch it take shape in front of me. Claude wrote the code. I talked about how it should behave. That loop is fast. Two hours, maybe. Most of it was the UI, iterating on the flow until the thing felt right in my hand. The math part was surprisingly quick because it was already figured out. The code part was almost a byproduct.
Honestly, the part that took longest was the logo. AI has gotten remarkably good at writing code and almost embarrassingly bad at clean, crisp visual iteration. A lot of back-and-forth with shape coordinates and color choices until something clicked. That, and only that, felt slow.
I have a computer science degree. I used to pride myself on how well I wrote code. Elegant algorithms. Simple solutions to tangled problems. Seeing patterns where there only seemed to be chaos. That was what I loved.
But when I look back, I don’t think I ever loved writing lines and lines of code. The typing. The boilerplate. The tenth implementation of the same pattern. The debugging that turns out to be a version mismatch. I tolerated it for the rest.
With LeafCharge, I let Claude pick the framework. I had heard of it but never used it. I could have looked at the code. I didn’t feel the need. Funny coming from someone whose career is code. What I was doing was the part I’ve always done: describing how the thing should behave, working out the logic, watching the pattern take shape. The rest was a vehicle.
I picked up some Svelte along the way, and how automated deployments on Cloudflare actually work. But the bigger learning wasn’t the technologies. It was getting to see the whole process at once. Usually, writing the code takes up so much energy that the other pieces either get glossed over or handed to someone else. With this build, I orchestrated the whole thing. Design, logic, copy, icons, deployment, SEO. All of it stayed mine to decide, without getting mired in any one step.
Working through every step of getting something out of my head and into the world — a different kind of satisfaction than solving a tough problem. Both take time and effort, but this one felt more tangible.
I use it now. Drag a couple of sliders, glance at the answer, set the charger, done. A few seconds. The math I used to do in my head is just there.
I’m already thinking about what to tweak next, which is probably the point. Things you’ve built tend to keep asking you questions. But the core thing, the thing I built it for, that’s just quietly working. A little daily friction, gone.
LeafCharge lives at leafcharge.dasasian.com if you want to try it out.
I came back to this observation later, from the other side. Working Faster Was Wearing Me Out is about what AI’s lowered threshold was costing me in the rhythm of the work itself.
What happened after months of using it: Pull Handle, Push Door.