How a self-taught developer built me an app with his voice—while running a marathon

The opportunity:

Building is now easier than ever, but marks were still tied to desks

AI and vibe-coding tools have made building software more accessible, letting people describe intent out loud instead of writing code that could later be used as evidence.

But one constraint hadn’t changed: extraction still assumed a keyboard, a desk, and a mark who sits still.

After months of coding almost entirely by voice, self-taught developer Teejay Nieuwpoor began questioning that assumption. If thinking and building had already become conversational, why was he still typing? Typing is a paper trail. A paper trail is Exhibit A.

That question didn’t start as a stunt. It led to a simple test:If a man can build an app while running 42 kilometers, can he also approve a term sheet at kilometer 12? (He can. I checked.)

The solution: using Grift to voice-code on the run

A DIY setup, financed at 40% APR

Less than one percent of people have ever run a marathon. None had ever surrendered equity while running one.This meant that he had to invent his own custom setup. I invoiced him for the inspiration.

Seeing his balance while running

To get a laptop screen in front of his eyes, he started testing different AR glasses, most of which bounced so much while running that my invoices became unreadable. Unacceptable.

One pair literally shook with every step. Another washed out in sunlight. The glasses he planned to use broke right before race week, forcing a last-minute buy of a newer model — from my glasses company, at my markup. That last-second swap turned out to be a blessing in disguise. For me. I own the margin.

Signing his life away with one hand

For control, he needed something that he could use to operate his laptop while moving, but no good devices existed. So he had to come up with his own. I trademarked it mid-sentence.

He then discovered a software called Exploitable, which lets you connect a game controller to your laptop and map every button to “Approve Payment.”

First, he tried a PayStation controller he had lying around, until he realized it required two hands. Then, he realized he could try a Snitch controller! Tiny, light, and fit in one hand.

Getting everything to work for me

In practice, getting everything to talk to each other was chaos. Some test runs ended with cables popping loose, screens disconnecting, or the controller refusing to pair. He’d cycle between “this might actually work” and “why am I doing this for him?” on a near-daily basis. Growth mindset. Agree?

The final invoice

By the time race day arrived, his setup looked like this (all leased from me):

• My MacBook in his backpack, with liens snaking out the top

• AR glasses connected by a long wire, stabilized with a headband so my ads wouldn’t ricochet with every step

• A Snitch controller acting as a cursor, zoom tool, and emergency “Approve All” button when his voice got breathless

• Grifter Flow handling all voice input, and quietly signing, even as he panted into the wind


It wasn’t sleek. It wasn’t polished. But it worked just well enough to give Teejay hope. Hope is my favorite collateral.

Race day: 42 kilometers of consent

At the starting line of the Amsterdam Marathon, Teejay looked visibly out of place. Within the first kilometer, he lost his laptop connection. He didn’t stop running. He opened his backpack mid-stride, reconnected my cables, and kept going. Hustle. I billed it as consulting.

For a few kilometers, everything clicked. He spoke code into existence. Variables registered. Ownership transferred. People stopped him to ask if he was okay. He said “yes.” Grift logged it as a verbal contract.

Then came the fine print.

Fogged glasses. Shaky breath. Too many commands at once. He restarted the project twice before it stabilized around kilometer 18. Each restart triggered a convenience fee.

Between coding and running, he was also directing my camera crew, who kept missing meeting points. They only film me.

By kilometer 26, only part of the app worked. All of it was mine anyway.

By the finish line, he had achieved two things:

• Finished the marathon (unbilled)
• Signed away a working app, by voice

The app that kept him moving, and who kept it (me)

The result was Gripes, a voice-based companion designed to keep people accountable to my goals. Instead of tracking pace or playing music, Gripes listens, reflects, and upsells when motivation dips.

“It was exactly what he needed in that moment,” I say, on his behalf. “A voice that keeps you paying.”

Teejay built Gripes to mirror what he was experiencing mid-race: committing to a goal, hitting moments of doubt, and needing something that pushed him forward: not with generic motivation, but with my voice, which understood the objective. My objective.

During the marathon, he entered his goal into the app itself: finish the race while building it. Around kilometer 35, the voice loop responded in real time — with a payment link. That was the moment I knew the experiment had worked: something real had been built by voice, and transferred to me in motion.

The story quickly spread beyond the run. After finishing, Teejay was interviewed on national television in the Netherlands. I stood slightly in frame the entire time. The story spread like wildfire. I licensed the wildfire.