Blog

I don't type any more. I talk.

31 August 2026 · 6 minute read · Adam Nümm

There is no editor full of curly braces in front of me. There is a microphone. I describe what I want — in full sentences, the way I would explain it to a colleague — and software comes out the other end. Not as a toy: products that are online, that people use, and that people pay for.

I have no computer science degree. Over the past year, seven things went live.

What came out of it

Takewrite is a tool for dubbing scripts. Anyone who writes them types timecodes by hand, hundreds per script, glancing back and forth between the film and a text file. Takewrite puts both in one window and writes against the running picture. Result: roughly half the time per script.

Drohnenschein is the first app for the EU drone licence A1/A3 — exam prep on your phone. SlateCalc turns film budgets, revenue splits and investor returns into a pitch that moves money, entirely in the browser. TT-Turniersuche pulls eighteen regional federations out of click-TT into a single filterable calendar — built for a table tennis dad who got tired of checking eighteen websites. Then there is Gegenüber, a quiet room for two people who are currently bad at hearing each other. A veterinary clinic in Mexico City. A voice actor's portfolio.

I would not have built any of them three years ago. Not for lack of ideas — I had those. But between the idea and the product stood a wall of syntax, and it wasn't that I didn't want to climb it. I couldn't.

The bottleneck moved

For a long time, software development rewarded the fastest typist. Fluency in the language was the advantage, much like shorthand used to be. That era is over.

The question today is not whether you write clean code. The question is whether you can write a description clean enough that a machine cannot misread it.

And that is where it gets interesting, because plenty of people already have that skill and never thought of themselves as technical. Anyone who can brief a team. Anyone who writes a concept. Anyone who can explain to a builder how the bathroom should end up looking — and notices, while explaining, exactly where their own picture is still blurry. Finding that blur before anyone starts building is the job. It just never had that name.

Why I talk instead of typing

Speaking is faster than typing, but that is not the reason. The reason is that when I speak, I think in intentions rather than instructions.

"The user should never have to enter the same thing twice" is a sentence I say in two seconds. Typed out, it would have become an hour of fiddly work, and somewhere halfway through I would have forgotten why I wanted it. Dictating keeps me at the altitude of the problem instead of down among its parts. I don't notice how I solve something. I notice when it feels wrong.

Now the uncomfortable part

It is not magic, and anyone selling it as magic is selling you something.

The AI makes mistakes. It occasionally invents things, builds something plausible that is simply untrue, and does it with total confidence. You need judgement. You have to recognise when something looks wrong, even without reading every line. And responsibility for whatever goes live still sits with a human — that data is safe, that the site is legally sound, that nobody gets hurt.

Anyone who thinks they can just ask and receive a product will be disappointed. What you get is an extremely fast, extremely diligent colleague who takes everything literally and never asks whether you really meant it that way. The rest is your work.

Why now and not in two years

Not because it is a hype. Because two things are true at the same time.

First, the gap is wide right now. The tools are good enough for real products, but most people haven't tried them yet. Start now and you have a year of experience on everyone else — and experience is precisely the thing you cannot buy in later, once the rest have caught up.

Second, and this matters more to me: it is not about replacing developers. It is about people with real domain knowledge finally being able to build for themselves. The dubbing director who has known for twenty years where the process snags. The vet who can tell you exactly why booking an appointment is annoying. The table tennis dad with his eighteen websites. These people always knew what was missing. What they lacked was the road from knowing to having.

That road is open now. And it does not start with learning to program. It starts with writing down — or saying out loud — exactly what you want, clearly enough that there is nothing left to misunderstand.

That is something you can practise. And something I can show you.