Why I Keep Building Things That Don't Need to Exist

I spent part of this year wiring a vintage dice roller to Twitter and part of it building life-size robotic dogs for the White House holiday décor. Neither project had a business case. Both taught me more about how I build than most client work did.

DiceBot started as a question: what if you didn’t have to be in the room to push the button? A Raspberry Pi, a motor, a camera, OpenCV counting pips, a Ruby script listening for #RollTheDice. About $130 and a weekend, plus the lighting fights that never fully ended. Gizmag covered it as an IoT curiosity. People tweeted at it from places I will never visit. That still delights me more than it should.

Bo-Bot and Sunny-Bot were the other end of the same impulse. Same motors, same iteration, same willingness to burn a part and try again, except the finished work ended up in the White House holiday décor. We posted the plans. If a kid builds her own dog-bot in a makerspace, the project did its job.

I know how this reads. Hardware toys. Side projects. The thing you do instead of shipping the real work.

Here’s what I actually get from them. Client software has constraints that hide your taste. Scope, stakeholders, deadlines, the politics of whose idea survives the meeting. A weekend build has almost none of that. You find out what you reach for when nobody is watching: whether you care more about the clever architecture or the thing that works when a stranger pokes it, whether you stop at “it runs” or keep going until the interaction feels inevitable. DiceBot didn’t need a microsite and a watermarked photo. We built those because the roll felt unfinished without them.

The objection I take seriously is that this is nostalgia for a smaller scale, a way to feel productive without facing the harder problems. Sometimes it is. I’ve used a soldering iron to avoid a hard conversation about a product. I’m not proud of those weekends.

But the other half is real. Every serious tool I trust, I trusted first in a toy. Computer vision on dice pips before computer vision on archival photographs. Servos and sensors in a dog head before I trusted myself to talk about robotics with Fellows and OSTP staff who lived in a different craft. The play is rehearsal. Not always. Often enough that I protect the habit.

I don’t have a clean rule for which side projects are worth the nights. The ones that pull me back to the bench after dinner tend to be the ones. The ones I have to schedule usually aren’t. That isn’t a methodology. It’s just the only signal I’ve learned to trust.