Module 1 · The basics of using AI
You speak many times faster than you type. As long as no technology exists that beats it, speech is the fastest way to capture what's in your head and turn it into results. Whoever clears this hurdle early and practices it has a clear advantage — today, and even more so in the future.
The live proof: this page
This page is itself the example. It was created from a single voice recording — spoken freely, with repetitions, circling thoughts and everything that comes with it. The unedited transcript was handed directly to an AI agent, no intermediate step. Compare for yourself:
📝 Raw transcript
The original, word for word, uncorrected — exactly as it came out of the recording (in German, the language it was spoken in). This is what the AI got as input.
📖 Formatted version
The same content: structured, repetitions merged, language smoothed — deliberately kept rich, nothing invented.
There is obviously no problem with recording a lot and at length — you still end up with a good result. That is exactly what you use the strength of LLMs for: getting to the point, summarizing, structuring.
Why speaking makes the difference
Speed
Even very fast typists can speak several times more content in the same amount of time. On complex topics you simply won't get as far writing as you will speaking — the AI receives more substance and delivers more of what you mean.
Freedom
There is no person listening who has to stay patient. You may circle, repeat yourself, ramble — any time, for as long as you want. The sheer amount of content and the effort of explaining are what make the quality of the result, not the perfect wording.
Practice that compounds
Every recording trains exactly what counts in the AI era: describing, formulating, gaining clarity about what you actually want. The more often you do it, the more ease and efficiency you develop — and a freedom you didn't see before.
And quite soberly: someone who gives input, plans and orchestrates by voice, versus someone at the keyboard — the advantage clearly sits on one side. There is no need for a long debate about that.
The objections — and what’s true about them
"Spoken language sounds different from written language."
True — this is the most understandable hurdle. But it doesn't have to stay that way: speak everything out first, just as it comes, have it structured, and then go over it yourself. Everyone will agree that it's harder to start with a blank page than when something is already there. And there are ways to have the text phrased very close to your own writing style — more on that in a module of its own.
"The first attempt never produces the right thing."
Expecting something to be perfect on the first attempt is utopian — no matter how much effort you put in. But: the first attempt is usually already good enough, especially for things you give to friends and family — and better than self-typed texts with mistakes, where most people never get to the key point. With a second iteration — naming what you notice, following up once or twice — you're done. That is remarkably efficient, and I dispute that, across many different people, there is a generally better and faster way.
"AI only produces average output anyway."
The more input and content you provide yourself, the more the result corresponds to you — and to what you want to achieve and convey. If you let the AI fill in everything, the result is nothing special; if you give too little, neither. The AI can smooth, structure and fill gaps — the substance comes from you. If you repeat things in a transcript that you don't want in the end, though, don't be surprised when they show up — that costs you a correction.
"Do I have to speak everything from now on?"
No. For small aspects, corrections or very specific inputs, short typed prompts are efficient — if no long recording is needed, don't make one. And if you don't really care how something is implemented, you don't need to speak much into it either; then grant the AI the freedom you're leaving it. The rule applies to everything where your vision, depth and intent should shape the result: better to say a bit more than too little.
"And what if the transcript makes mistakes?"
Transcription today works with well over 97 % accuracy in German and English — even on local models. A few extra sentences help further: if you swallow a word, context corrects it. And checking is easy — ask for a summary and see whether your important points are in there, instead of re-reading the whole recording. Or have it implemented right away and judge the result.
How to start
- Record instead of typing. Take your next more complex matter — an idea, an email, a project — and speak it out as it comes. Your phone’s dictation feature is enough to begin with.
- Have it structured. Give the transcript to an AI: summarize, outline, smooth — or have it implemented directly. Say what the goal is and why you’re doing it; that’s enough, even without a detailed vision.
- Go over it and follow up. Read the result, name what doesn’t fit, iterate once or twice. Done — and next time it will already come easier.
Positive experiences come from trying — trial and error beats every prejudice and every inhibition. Whoever doesn’t apply this misses out on opportunities to realize things with the help of AI. And that would be a shame.
What comes next
That was addition and subtraction — the basics everything else builds on. Over time, more modules will follow here: the most important tools and the most sensible approaches for using AI efficiently — today, and for whatever comes next.
A personal word
“ I carry a great deal in my head on any topic — and because of the sheer volume and the speed at which I see connections, it is hard for me to always set the right priorities. For me, AI is a great gift: I use it to prepare my content for others.
I've been told this is "impersonal". The truth is the opposite:
I invest more time in the person and the content this way, not less. In a recording I can name things in depth and say how they are meant — and then get help being understood, without overwhelming others with the sheer volume. That way, time, effort and nerves are spared — theirs, and mine too. Without this help there were often follow-up questions, long processes, frustration — up to people avoiding starting topics with me at all, because a short answer is rarely possible from my side. I don't find that fair when I give my best and it isn't enough. So I use today's technologies to turn this weakness into, perhaps, even a strength.
I am open about it: yes, the AI helps me phrase and structure — and I read what I send, and I stand behind it myself. It would be something else to send messages one has never read; that would no longer be personal interaction. This way, however, it is a sign of respect for the other person's time — and should suggest importance, not impersonality. Not that everyone has to do it; everyone is free. But I hope understanding grows here — and that this example encourages you to take the first step yourself, in this "future is now" moment. ”
Created from a single recording · 2026-08-11 · Module 1 of X