Built by hand
Not so long ago, while scrolling through Hacker News in the morning, I stumbled upon an article that immediately caught my attention. The topic was explosive and highly topical: Claude is not your architect. A warning that Claude, Anthropic's language model, is secretly taking over larger and larger parts of industrial design. The author painted a grim picture of a world where human creativity in engineering is replaced by automated algorithms.
The arguments were sound, the structure seemed professional at first glance. But while reading, an uncomfortable feeling quickly crept over me. Do you know that feeling when a text is grammatically perfect, but still somehow feels "wrong"? I looked closer.
There they were: em-dashes en masse. Almost every second sentence was interrupted by these long dashes. On top of that, there were strange, almost imperceptible jumps in thought between the paragraphs. Phrases repeated themselves in slightly modified forms, and the text drifted along in a permanent, artificially bloated epic style. The irony was perfect: an article warning about the takeover of the industry by AI was obviously entirely generated by an AI itself.
I thought to myself: "Well, the tech gurus on Hacker News will notice that immediately." After all, the smartest minds in the IT world hang out there. I scrolled through the comments to read the malicious reactions. But instead I found: nothing. Serious discussions about the content. I set my own sign.
When even the smartest minds no longer perceive the difference between human pen and algorithm, we have reached a turning point on the internet. We consume content every day without knowing whether it is backed by real creativity or just a well-trained prompt response.
It was precisely this feeling of untransparency that made me reflect sustainably. In order not to become part of this anonymous noise myself, I want to go a different way. It is time to provide clarity, and therefore I hereby disclose how I personally deal with AI in everyday life.
The Code: Why my footer is not a liar
When you go to my website and scroll all the way down to the bottom left, you will find a sentence that I chose deliberately: "Built with care by hand".
I take this sentence damn seriously. At the same time, however, I also have to be honest: this does not mean that I completely do without artificial intelligence in the year 2026. For me, this sentence states that this website was not created autonomously by some AI agents. There are tools out there where you press a button, type in: "Make me a portfolio page for a developer", and presto: the system spits out a finished, soulless page.
That is not my way.
I basically deal with AI exactly the same way I used to deal with StackOverflow or Google back then. When I am at a point while programming where I simply don't know what to do next, I ask one of the models.
Which model do I use? For me, that is a matter of pure gut feeling and a question of the available free tokens. When the contingent for model A is exhausted, I move on to model B. Despite the fact that I ask the AI for advice every now and then and sometimes adopt the source code generated by it 1:1, that very sentence still stands at the bottom left of my page.
I can already hear the critics shouting: "Hey! You say it's built by hand, but you use AI? Traitor!" But in all seriousness: back when we searched through forum posts for nights on end, did we put "Built with the help of Google" in our footer? No. The creative mind, the architecture, and putting the puzzle pieces together. For me, that remains manual work.
The Translations: AI as my personal bridge builder
Another area where AI takes an incredible amount of work off my hands is translation work. My website is structured bilingually. It used to be an absolute chaos. Since my brain runs on either German or English depending on the form of the day and consumed media, my site was a colorful, mixed-up patchwork quilt. That looked unprofessional and unstructured.
Today, my workflow looks like this: for every text, I first choose a source language, mostly German, because I can express myself most precisely in it. Once the text is ready, I let the AI translate it "word for word" / "literally". The beautiful thing about modern LLMs compared to old translation tools is that they understand the context. They do not just blindly translate words, but transport the tone of voice. The AI functions here as a bridge that makes my thoughts accessible to an international audience without losing the original vibe.
The Blogs: Raw brainstorm meets the stonemason
Let's come to the blogs and thus also to this text that you are currently reading. I firmly maintain: my blogs are still written by hand. But here, too, AI is used, namely as my personal stonemason.
When I have an idea for a blog post, I just start typing. I don't care about grammar, punctuation, or smooth transitions.
- I formulate my texts in the first draft with an immense amount of typos.
- There are expressions in there that could straight up come from a six-year-old.
- I throw around creative, colorful, and sometimes completely absurd metaphors.
Someone has to document my creative chaos, after all.
The result is a raw, wild pile of text. Mostly around 4,000 to 6,000 characters. I take this digital mud, throw it into a model of my choice, and give the system a very clear, strict instruction:
"Take this text. Target: at least 10,000 characters."
So the AI takes my raw material and stretches it out, builds elegant sentence structures, and ensures the necessary text volume. What comes out of it must, of course, never be published unfiltered. Therefore, I always proofread the whole thing extremely thoroughly, correct it, and put my final stamp on the text.
What I will absolutely never do under any circumstances, however, is: "Give me 10 ideas for cool blog posts" or "Formulate a topic from scratch for me to 10,000 words." That is the red line for me. The content, the ideas, the core thoughts are and remain human. The AI is my tool, not my ghostwriter.
A small funeral for the dash
When correcting these texts formulated by the AI, one thing has struck me extremely hard lately, forcing me to make a radical change: I now systematically replace all en-dashes and em-dashes with other punctuation marks.
Why? Because these dashes have now become the absolute brand mark of AI texts. The language models have simply used them far too inflationarily in recent years. If you see a German text today that uses an elegant dash every two lines, the internal AI detector triggers immediately. Personally, I find that an incredible shame.
I really liked the en-dash very much. Back in the day, I used to explicitly google it and copy-and-paste it into my texts because it just looks nicer in the typeface than the clumsy, short hyphen. Later I learned the shortcut: under Windows with WIN + -.
| Character | Name | Width | Origin |
|---|---|---|---|
| - | Hyphen / Quarter-em dash | Very short | Separation & coupling |
| – | En-dash / Half-em dash | Width of a capital N | Dash, to-dash |
| — | Em-dash / Em dash | Width of a capital M | Strong jump in thought |
The en-dash (–) is called that because in traditional typography it has exactly the geometric width of a capital letter "N", same for the em-dash (—) for "M". All these beautiful fun facts will slowly be forgotten because this beautiful symbol is today one of the clearest indications of soulless AI texts. It was corrupted by the algorithms. So please take this paragraph as a small, sad funeral for my beloved dash.
Conclusion: For a new culture of honesty on the net
I am no extremist. I don't expect everyone to frantically slap "Created with the help of AI" into their website just because they briefly asked a model for a synonym.
What I do demand, however, is a certain degree of fundamental honesty as soon as someone clicks the big, red "Publish" button. I would like to know on the internet where a human was still actively involved and where not. I want an awareness for the value of thoughts to emerge again.
As harsh as it may sound: I want people to feel a bit ashamed again when they just barf a purely AI-generated, unverified text onto the net just to grab SEO traffic or fake content volume.
We urgently need to promote a culture on the internet where the human remains in the loop when creating content. A text should be the product of a human mind making use of tools to express itself better. Not the product of a machine trying to completely simulate human thinking while the actual "author" only counts the clicks. Because nowadays, what is generated without rhyme or reason is no longer real content. It is nothing but digital noise.
Let's stay human. With all our typos, our quirky metaphors, and our own, imperfect style.
*~ This blog post originated from originally 6,988 characters, which were typed by hand with a lot of heart and soul, typos, and real thoughts.
Gemini OneShot prompt for it:
hierzu einen blogpost insgesamt mindestens 10000 zeichen
Delivered: 12,580 characters.
Post-correction: 10,376 (incl. .md formatting and reading 5 times)
Formatted with this tool.