Six tools, zero lines of code written by hand, and a Swiss army knife I didn't know I needed.

For most of my professional life, I've had ideas that needed a developer. Simple things. A tool that does one specific task, solves one specific problem, saves thirty minutes a week. The kind of thing that's too small to commission and too technical to build yourself. So you either find something that almost does what you need, with twenty features you'll never use and a subscription you'll resent, or you just keep doing the thing manually.
That changed recently. And I didn't expect it to happen the way it did.
I started experimenting with AI agents to solve small, annoying problems. Not big ambitious projects. Just things that had been sitting on a mental list for a while. A tool that pulls PDF order forms from a folder and calculates revenue in one click. Another that tells you which product was sold to which customer on which date. Tiny things. But they worked. And when I showed them to a couple of friends who run small businesses, they immediately started using them.
That's when I realized: the bottleneck was never the idea. It was always the execution. And vibe-coding, if you want to call it that, removes the execution bottleneck almost entirely.
Here's the thing nobody tells you about building tools with AI: the hard part isn't the building. It's knowing what you want.
You can't just say "build me a digital literacy quiz" and get something useful. You need to know what questions matter, what the right answer categories are, how the scoring should work, what a useful result looks like. The AI handles the code. You have to handle the thinking.
This is actually where domain knowledge becomes more valuable, not less. I could build the digital literacy quizzes because I've spent years researching how algorithmic systems affect human behavior. I could build the running tools because I run ultramarathons and know the problems from the inside. The vibe-coding part is fast. The expertise that makes the output actually useful takes years.
And yes, there's a technical minimum. Getting something deployed somewhere other people can actually access it requires enough understanding of how GitHub Pages works and what an API call looks like. Not a lot of technical knowledge. But some.
My first published tools came from my research work. Long reports are useful, but most people don't read them. An interactive tool that takes two minutes and gives you a personal score is a different thing entirely.
The Digital Wellbeing Score Calculator measures five dimensions of how you relate to digital environments: Screen Quality, Algorithmic Dependency, Identity Coherence, Social Comparison Load, and Digital Boundaries. It's based on the Digital Wellbeing Index research framework published on Zenodo.
The Digital Literacy Quiz helps you understand where you stand in the digital world. Fifteen questions, results place you as a Passenger, Navigator, or Pilot.
The Digital Habits Quiz for Kids Ages 9-12 covers screen time awareness, gaming habits, online safety, and real life balance. Results are encouraging rather than judgmental.
All three are free, no signup required.
The second set came from a different place entirely. I run ultramarathons. And runners have problems that are annoyingly specific and surprisingly unsolved.
Running Outfit Weather started from a very wet morning. I put on my best waterproof jacket for a rainy trail run and came back soaked from the inside. The tool pulls real-time weather data, applies wind chill and heat index formulas, adds warmth for longer efforts, and recommends what to wear.
Pace to Playlist came from a simpler frustration. Have you ever been mid-tempo run and a slow song comes on? You feel your legs respond before your brain does. This tool takes your target pace and returns ten songs in the right BPM range.
Name Your Run is smaller but useful for anyone who shares their runs. You finish, open Strava, and stare at the activity name field. "Morning run." Again. This tool generates a name and motivational line specific to that run.
I'm a industrial engineer by training. I research digital behavior professionally. I run ultramarathons. None of those things made me a developer. But all of them gave me problems worth solving and enough domain knowledge to know when a solution actually works.
Vibe-coding, for me right now, is a Swiss army knife. Fast, practical, good enough for problems that aren't too complex. The repetitive tasks, the small annoyances, the "someone should really build this" ideas that never get built because they're too small to commission.
More complex things? Maybe later. For now, I'm genuinely enjoying solving small problems quickly and seeing people actually use what gets built.
Life is more automatable than we realize. And most of us don't know it yet because the execution barrier always got in the way first.
All tools at: https://github.com/recepzerk
Research at Zenodo. Running at recepzerk.com/running