Why I'm Building Apps with AI
Instead of Learning to Code

My father tried to teach me to code when I was a kid. Not in a serious way — more of a "this is the future, you should know this" kind of conversation. I don't remember the language. I just remember the feeling: Hello World, a blinking cursor, and a complete blankness about why any of it would matter to me. It felt like being handed the wrong instrument for music I hadn't heard yet.

I went into marketing. Words, ideas, audiences, what moves people and why. That felt right. Software was someone else's problem.

The first crack in the wall

Years later, I started doing CSS work for clients' eshops. Small things — a header redesign here, a product card layout there. I'd open Chrome DevTools, change a value, watch the page respond instantly, change it back, try something else. There was no compilation, no abstraction. Just: adjust this, see that.

It felt like playing. More specifically, it felt thrillingly curious — the same way I imagine people feel when they first pick up a musical instrument and accidentally make something that sounds good. I wasn't learning to code. I was just exploring what was already there, which is a completely different thing.

That was the first hint that the gap between me and software wasn't as wide as I'd assumed. It was just that the entry points I'd been shown before were the wrong ones for how I think.

Scripts, ChatGPT, and a pattern I didn't notice at the time

ChatGPT launched in late 2022 and I was immediately curious. I started using it in my marketing work, and at some point something shifted: I stopped asking "what problem do I have that I need help with?" and started asking "what could I actually build with this?" The tool came first, and it opened up a whole new layer of thinking about what was possible.

Google Ads scripts were the first real test. I wanted to automate things the interface just couldn't do — custom bid logic, smarter reporting, bulk operations that would take hours manually. Before ChatGPT, scripts meant finding a developer or giving up. Now I could just describe what I wanted: what the script should do, what data it should touch, what the output should look like. ChatGPT wrote it. I tested it, found the edge cases, described them back. It fixed them. I ran it.

Each new AI feature that arrived expanded that mental layer further. Not "what problem does this solve?" but "what new thing does this make possible?" That curious accumulation — of tools, of ideas, of half-finished concepts — is what set everything in motion.

Every tool that almost got there

By 2023 I had real app ideas collecting in my head and in old chat histories. I tried everything that promised to close the gap: ChatGPT's code generation, Perplexity, Google AI Studio with Gemini. Each new one came with bigger promises than the last.

For small, self-contained things they delivered. I built jeuzbitcoinza100k.cz — a fun app for tracking whether Bitcoin would hit $100K — and a QR Clock that generates a scannable code from the current time. Simple, satisfying, done.

But anything more ambitious hit a wall. SpekPlatz was supposed to be a community map for sharing beautiful spots — curated places, a sense of collective geography. I had a clear vision. I could describe every detail. But every tool I tried either produced something that missed the point, or something so fragile that adding one feature broke three others. It sat unfinished, month after month. That gap between "I can describe this clearly" and "this actually works" was still wide for anything beyond the trivially simple.

"I had ideas. I had the ability to describe them clearly. What I was missing was a tool capable of acting on that description at the level of complexity I needed."

One evening in February

In February 2026 I finally opened Claude Code and described SpekPlatz — properly, the full thing, the idea I'd been carrying around for years. Interactive map, place collections, community sharing, a custom visual design. It was live that same evening.

Something I'd been stuck on for so long, finished in one focused session. That felt significant enough that I went back through my old GPT and Gemini conversation history and started pulling out everything I'd given up on. Old ideas, abandoned prototypes, half-baked concepts I'd shelved because the tools weren't ready. I started trying them again.

They worked. Most of them worked on the first or second attempt. Music Mouse — a generative instrument where cursor movement becomes music. ConsentKit — a proper cookie consent library I'd previously been paying for as a SaaS. DayLog — a habit tracker I built for a friend in one evening. VagoshIt — a full group activity planner with real authentication and a live database. TimeInvoice — a freelancer invoicing tool I'd been paying a monthly subscription for.

After watching this work, I decided to build this website — a place to share everything I'd made, freely, so others could use it and maybe feel the same nudge I did: that building something yourself is closer than you think.

Why everything is free, and what I'm actually learning

All my life I've been curious about things that interested me — not because I wanted to make a career from them, but because exploring them was exciting. Sometimes those things turned into work anyway, and people started paying me for what I'd originally done just because I liked it. Marketing was like that. CSS was like that.

Building apps feels the same. I don't want to put these behind a paywall. I want to give them to people, see what they do with them, build more. But I'm aware that the skills I'm accumulating — understanding what makes software work, how to think through a product, how to collaborate with AI at this level — are going to be genuinely valuable in ways I can't fully predict yet. I'm learning something real here, even when it feels like I'm just having fun.

So everything on Resonant Labs is free and open source. Use it, fork it, build on it. And if it makes you want to try building something of your own — that's more than enough.