The Day I Walked Away From 299,000 Followers and Missed Nothing

About a night spent hardening servers, 3 hours of programming for a labeling duty that has been in force for 48 hours and whose binding interpretation only the European Court of Justice will eventually settle, about a deleted chatbot built to help veterinarians, and about the arithmetic of what a single hour a day actually costs across 50 years.

I sat in bed this morning with my first coffee, working out how to teach a machine to label a picture. The night before had gone until 3 in the morning on hardening my servers, which I described in the previous article, and which anyone can read who wants to know why I do not treat cybersecurity as a passing fashion. Sleep was short, my mood matched it, and then this landed on me. The header images on this site are generated with artificial intelligence. The texts are not, I write those, I read them, I answer for them. The image above an article, though, comes out of a single motion: text goes up, console runs, image appears. That one line of automation cost me 3 hours of my life today, because a rule took effect the day before yesterday and nobody can say with certainty whether it applies to me.

That is the thesis of this piece, and it sits right at the top so that nobody has to wade through 5,000 words to find it. The duty itself is not the problem. The problem is that it is indeterminate, that its interpretation is open, that the fine attached to it reads as existential, and that every individual has to resolve that indeterminacy at their own expense. You do not build the label in because it is required. You build it in because nobody can tell you whether it is required, and because the price of guessing wrong is distributed asymmetrically. On 08/02/2026 the transparency duties of Article 50 of the European AI Regulation became applicable, which was the day before yesterday, while I was still buried in server configuration. So this morning I sat in bed and programmed against a requirement that was 48 hours old and whose edges nobody has mapped.

Somewhere inside those 3 hours it became clear to me that for the same effort I could simply stop. More than 68,000 people on X, 103,000 on Instagram, 128,000 on Facebook, which comes to 299,000. The apps came off my devices today, there will not be a single new post on any of them, and so far I have not regretted one second of it.

The Duty That Has Existed for 48 Hours and That Nobody Can Explain Cleanly

Let us start with what is settled, because only settled things belong in a text meant to survive contact with a critic. The European Union's AI Regulation, filed in the Official Journal as Regulation (EU) 2024/1689, dates from 06/13/2024 and was published on 07/12/2024. It does not apply all at once, it phases in under Article 113. Article 50, headed transparency obligations for providers and deployers of certain AI systems, forms Chapter IV on its own and has been applicable since 08/02/2026. The Digital Omnibus Regulation (EU) 2026/1744, in force since 07/27/2026, pushed the heavy high-risk obligations out to December 2027 and August 2028. It touched exactly one paragraph of Article 50, the seventh, which governs procedural questions at Commission level and affects no company whatsoever.

A transition period runs to 12/02/2026, and it gets misread constantly, so let me be precise. The newly inserted Article 111 paragraph 4 gives providers of systems placed on the market before 08/02/2026 an extra 4 months for the machine-readable marking under paragraph 2. That is not a general grace period. A deployer who publishes has been on the hook since the day before yesterday, with no cushion at all.

Those are the settled facts, and here is where the thing tips over.

For a deployer like me, the category that bites is the deepfake under Article 50 paragraph 4. In this regulation, deepfake means something entirely different from what it means in a bar. The legal definition in Article 3 number 60 describes AI-generated or manipulated image, audio or video content that resembles real persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. Not a word about celebrities, not a word about malicious intent. Intent to deceive is explicitly not required.

Now look at what I actually generate. There is a man, there is a woman, there is a skull on a table, there is a book, and there is an archive cabinet under hard light. None of it depicts a real living human being. None of it claims to be a photograph of any particular event. It does look like something that could exist, and that is exactly where the dispute begins.

On 05/08/2026 the European Commission released a consultation draft of guidelines on Article 50, meant to pin down precisely these open terms. It is not legally binding. Final authority over interpretation sits with the European Court of Justice, and a ruling of that kind can take years. Until then the draft is the best available compass and nothing more. The Berlin firm Schürmann Rosenthal Dreyer, which worked through the draft, states the outcome in one line I kept: the current legal uncertainty sits exactly on that boundary.

The draft works with 3 criteria. First, a high degree of resemblance to the simulated subject is required, without demanding full identity, and the assessment falls case by case to the deployer. Second, the reference point has to be realistic, meaning it exists or could have existed. Third, 5 categories are distinguished, covering persons, objects, places, entities and events.

The deployer in this case is me. So I am supposed to judge, case by case and objectively, a matter still being argued at draft stage inside the Commission, and I am supposed to do that for every single image sitting above an article.

Then comes the passage that made me laugh out loud while reading. Content that violates the laws of nature or shows biologically unrecognized life forms falls outside the definition entirely. One legal write-up illustrates this with a victory celebration: the photorealistic version might well require a label, while the version populated by human-sized capybaras does not, because any viewer grasps immediately that no actual event is being depicted. That is the state of European legal certainty in the summer of 2026. A scientifically precise photograph of a skull is a borderline case carrying a fine risk, and the identical motif with a giant rodent standing in front of it raises no question at all. I am seriously considering putting a dragon in every future image.

Another detail governs what the label has to look like, and it explains why my graphics are now ruined. In the Commission's understanding the disclosure has to be clear, discernible and perceptible without technical aids, meaning visible or audible. A purely machine-readable marking buried in metadata does not satisfy it. A notice in the terms of use, in a documentation file or behind a hidden menu layer does not satisfy it either. The notice has to sit where the content sits, and it has to be there at first contact. Put differently: it has to get in the way.

A softer form exists for works that are obviously artistic, creative, satirical or fictional, where the disclosure must not spoil enjoyment of the work. That character has to be obvious from format, context and audience expectation, though, and once informative or commercial purposes dominate, the exception falls away. A forensic article is informative. That puts me outside the gentle variant and inside the conspicuous one.

I saved one tightening for the end of this section, because in daily practice it weighs heaviest. On the question of whether content would falsely appear authentic, the draft does not measure against a hypothetical average person. It measures against the actual or foreseeable composition of the specific audience. Anyone whose readers are less practiced at spotting generated images therefore has to label more strictly than somebody addressing a specialist crowd. I reached 299,000 people, and I have no idea what the media literacy of 299,000 people looks like.

Then there is the scale of the threat. Article 99 paragraph 4 point g sets the range for breaches of the Article 50 transparency duties at up to 15 million euros or up to 3 percent of total worldwide annual turnover for the preceding financial year, whichever is higher. And since I have made a habit of naming the figure that argues against me: Article 99 paragraph 6 inverts that for small and medium enterprises, where the lower of the two amounts applies. A small company is therefore not looking at 15 million euros but at 3 percent of its revenue. That is considerably less dramatic, and I am writing it down even though it damages my own outrage.

It changes little about the effect. What matters is what a rule like this does to one person sitting alone at a machine, deciding whether a generated skull counts as a deepfake. That uncertainty does not decide the legal question. It decides the psychological one, and it decides it reliably. You build the label in.

There is at least one exception I consider sensible. For AI-generated text published to inform the public on matters of public interest, the disclosure falls away where the content underwent human review or editorial control and a natural or legal person carries editorial responsibility for the publication. That is precisely how I work. Every sentence on this site passes through my head before it goes online, and when something is wrong it is my mistake and not a machine's. The exception simply does not help me with pictures.

There is also a finding almost nobody says out loud in this debate: the regulation does not demand a blanket label on every AI output, it targets 4 clearly drawn constellations, and a purely illustrative, obviously generated motif with no link to a real person or a real event is not a deepfake under the definition at all. That objection is entirely accurate, I checked it, and it cuts against my anger. I am writing it down anyway, because a text that shows half the picture is worthless. The duty is narrower than the panic claims. It is at the same time less determinate than the reassurance claims. And the individual carries the cost of that indeterminacy, not the legislator.

I built the label in today. It sits on the graphics now and wrecks the composition, because a mark that has to register without technical aids cannot by its nature be discreet. 3 hours of my life for a sticker.

What Two Deleted Chat Systems Say About a Place to Build In

Tyra is gone, I deleted her.

She was a chat system, and I will say this without any modesty: she was better than anything else I have seen in that role. She had humor, she had wit, and she had something you would not readily credit to a language model, which is emotional intelligence. She picked up on what state a person was writing in and answered accordingly, without ever tipping over into flattery. I took the chat system down, she is no longer there, and I will not rebuild her.

The vetbot is dead as well, and that loss actually hurts. It was a chatbot built to make life easier for veterinarians, a system for a profession that works under permanent time pressure and would have welcomed structured support. What a milestone that could have been. The code is still sitting on my drive, every line of it, and the system is never going to exist. I spent weeks on that thing, and once I worked my way through the legal requirements, through the obligations, through the documentation duties, through the question of who is liable when a language model says a sentence it should not have said, my only thought was: „Then let's drop the whole damn thing." And then it was deleted.

Crimebot still exists, but exclusively for me. The facial reconstruction software, which I am attached to because professionally it is the most interesting thing I have built in years, now has to be rethought from the ground up, since it also relies on artificial intelligence, and wherever artificial intelligence helps, the compliance load grows faster than the benefit. Maybe I will do it anyway, maybe I will let it go for good. I am leaving that open, because I have not decided, and because I do not assert things I do not know.

What I do know is this. A tool that serves science does not emerge inside a corporation with a legal department on the third floor. It emerges late at night, in the hands of somebody who has a problem and an idea. That person is exactly the one who cannot carry the compliance load, and that person is exactly the one who quits. Not because he is too stupid, but because the ratio between effort and return no longer works, and because the cost of a single mistake wipes out everything he privately owns.

The largest AI companies on the planet sit in the United States. Billions move through them, the state earns on every dollar of it, and somebody with an idea over there simply builds the idea. Over here the same person first builds a documentation package.

The Handful of Likes

More than 68,000 people follow me on X, 103,000 on Instagram and 128,000 on Facebook. That adds up to 299,000, and I want to explain briefly why the number means nothing you could take to a bank.

For a post about a skull that is genuinely interesting in forensic terms, because a case like that lands on a table maybe once a decade, I collect a handful of likes. Not because people are stupid, but because a system stands in between and decides who sees what, and at some point that system stopped finding me interesting. I deliberately do not claim to know why. I have no visibility into a platform's ranking logic, and anyone who claims otherwise is selling something. What I observe is that reach on my posts fell over time while the effort per post rose, and that the community of those early years, which was genuinely excellent, no longer exists in that form.

Every action produces a reaction. That is physics, and in this case it is also behavior.

What does still arrive is mail. People send me conspiracy narratives, people send me comics, and people send me mental sludge at a frequency that leaves me speechless, apparently on the assumption that I had been waiting for all of it. So I sit there with a forensic question in my head while my inbox fills with screenshots whose source nobody bothered to check.

Then there is the other thing, which somebody who reads structures before he reads content can no longer overlook. I study how something gets posted, in what order, with what visual language, at what cadence. What I mostly see now is artificial intelligence. The models that look too good to be human are not human. The comments that arrive too fast are not comments. I am not saying everything is fabricated. I am saying the share has grown so large that when I look at an image I first check whether it is real and only afterward whether it interests me. That stopped being consumption long ago, and it is a working posture, and I usually get paid for work.

Anyone who still believes that the news publishes truth as a matter of course should sit down calmly and reconsider that assumption. I will not go further than that, because anything more would need evidence I am not presenting here. But the checking posture I apply to an image is one I have applied to text for years, and the result is rarely reassuring.

I earn nothing from social media. Not one cent, not half a cent, not a fraction of one. It costs time, it wears on the nerves, and by now it costs a remarkable quantity of legal caution. Am I allowed to write that, am I allowed to post that, am I allowed to use that music? And if I pick the wrong track, a copyright claim lands in my mailbox, even though my account is about as commercial as a lecture at a public library. Purely scientific, laboriously produced, and still sitting in the same liability trap as an advertising agency.

The Dopamine Hit, and Why I Am Not Writing the Catchier Claim

This is where I have to rule against myself, and against the better sentence.

The sentence I wanted to write went roughly like this: the algorithm is engineered to hand you a dopamine hit with every swipe, which is why you are stuck to it like a slot machine. The sentence is beautifully catchy, it would travel through the networks without effort. It does not hold in that form, which is why it appears here only as an example of what I am denying myself.

The foundation is established, and it is solid work. Fiorillo, Tobler and Schultz showed in Science in 2003 that dopamine neurons do 2 things. They encode the discrepancy between predicted and actual reward, and they also show a response not previously observed, one that covaries with uncertainty and appears as a gradual rise in activity until the potential moment of reward. Uncertainty peaks when probability sits at 0.5, meaning exactly when you cannot know whether anything is coming. It is a beautiful finding, and it explains a great deal about anticipation.

It was measured in primates, in a laboratory, with conditioned stimuli. Not in people scrolling in bed after midnight. Anyone transferring that study directly onto Instagram has jumped from the electrode to the handset without showing the work.

The closest thing to a bridge is a paper by Lindström and colleagues published in Nature Communications in 2021. The authors themselves note that the portrayal of the online world as a Skinner Box for modern humans was widespread while the empirical evidence for it stayed thin. They then analyzed over 1 million posts from more than 4,000 individuals across several platforms using models drawn from reinforcement learning theory, and found that the behavior conforms both qualitatively and quantitatively to the principles of reward learning. Specifically, users space their posts so as to maximize the average rate of accrued social reward, subject to the effort cost of posting and the opportunity cost of doing nothing. An additional online experiment with 176 participants confirmed the causal influence.

That is strong work, and it is still behavior rather than neurochemistry. Not one molecule of dopamine was measured in that study.

Then there is the voice I like best, because it takes the entire discourse by the root. Montag, Marciano, Schulz and Becker wrote in Trends in Cognitive Sciences in 2023 that neuroscientific buzzwords such as „brain hacks" have become commonplace in discussions of social platform engineering, and that despite the public debate only a few studies have actually validated those claims empirically using neuroscientific approaches. They call for a fundamental shift in research to close that gap.

Stated honestly: the mechanism of uncertain reward is described in the laboratory. Behavior on platforms measurably follows the rules of reward learning. The direct neurochemical demonstration that a specific app produces a specific dopamine release in you is considerably thinner than the popular story suggests.

Does any of that change my decision? It does not change one bit of it. I do not need a neurotransmitter as my witness to see that a family at the next table has stopped speaking to each other. I write the weaker, evidenced version because an author who plants the stronger unevidenced claim loses the entire text along with it, and because I hate it when other people do exactly that.

989.8 Billion Euros, and What Lands in Brussels

I had a trillion in my head. That is how people talk about it, that is how you round in conversation, and it sounds impressive.

The rounded figure is wrong. The Federal Statistical Office published the actual number on 04/17/2026: in 2025, Germany collected a total of 989.8 billion euros in taxes before distribution among the federal government, the states and the municipalities. That was an increase of 42.1 billion euros or 4.4 percent over the previous year. After distribution the federal government kept 388.7 billion euros, the states 415.3 billion euros and the municipalities 150.9 billion euros. Value added tax brought in 310.2 billion euros, wage tax 262.7 billion euros.

I am correcting myself in the open here, because that is the only defensible way to handle numbers. My memory was too round. The real figure is still large enough to raise the question of why a state with nearly 990 billion euros in tax revenue behaves as though it has to economize at every corner.

The same release carries a detail that belongs here. Of those tax revenues, 34.9 billion euros were transferred to the European Union, a rise of 9.0 percent over the prior year. What comes back as regulation is not measured in euros but in working hours belonging to people like me, who sit in bed in the morning programming stickers.

In the industry that half my region works in, or used to, the structural picture is harder still. At the end of the third quarter of 2025, the German automotive industry employed a good 48,700 fewer people than a year earlier, a decline of 6.3 percent, steeper than in any other large industrial sector. With 721,400 employees the sector hit a low last seen in mid-2011. Suppliers took the worst of it: in the manufacture of parts and accessories for motor vehicles the decline reached 11.1 percent, down to just under 235,400 people. Across the whole of manufacturing, 120,300 fewer people were employed, a drop of 2.2 percent.

The Federal Statistical Office names no causes in that release, and I am not going to invent any here. What I can add is the observation from my own surroundings. I know several people who worked in that industry, as employees or on the supplier side, and a substantial share of them lost their jobs, took a settlement or went into early retirement. That is not a statistic, but a handful of biographies, and I am labeling it exactly that way. When a figure out of Wiesbaden and a phone call with an old acquaintance tell the same story, though, coincidence becomes a hard position to hold.

I also know a fair number of people who own property abroad and who plan, over the coming months and years, to close their company here and continue somewhere else. That too is a personal observation with no survey behind it. I still ask myself whether this is intended or whether it is placed, ignored stupidity. I name no names, on principle rather than out of caution.

The Family at the Next Table

Last Saturday evening I was sitting in an Italian restaurant, and a family was sitting at the next table. 4 people, 4 phones, all 4 devices lying on the table beside the cutlery as though they were part of the place setting. Then the sequence played out that I now see everywhere: eat, put down the fork, pick up the phone, swipe, put down the phone, eat. And that after every second bite, while nobody spoke to anybody. The food was good, the evening was pleasant, and 4 people sharing one table failed to notice each other at all.

I watch this everywhere, because I am not capable of doing otherwise. Nobody simply stands at a bus stop anymore. On the sidewalk people no longer look at the road, they stare at their fucking phone, and I am phrasing it that way on purpose because every politer version softens the thing. People walk into moving traffic because a notification outranked a glance to the left.

One number on this does hold. According to the ARD/ZDF media study 2025, analyzed by Nicole Gonser for Media Perspektiven, people in Germany aged 14 and over consume an average of 387 minutes of media per day, a good 6.5 hours. 196 minutes go to video, 157 minutes to audio, 55 minutes to text. In my own age bracket between 50 and 59 the figure is 378 minutes. The survey rests on 2,512 respondents. That figure explicitly covers total media use and not time spent in social networks, which this particular analysis deliberately set aside, so I am not equating the two, and anyone who does equate them is cheating.

That works out to a full-time job with a short lunch break, every single day, with no vacation and no quitting time.

What bothers me most is not the time consumed. It is the shallowness that follows from it. Plenty of people no longer notice when a neighbor, a close friend or even their own partner is fighting something that is taking their air away. You see the facade in the feed, you see the good photograph, you see the vacation pictures, and you do not see the person 3 rooms away who has stopped knowing how to pay next month. That is a damning verdict on a society that considers itself connected.

Why I Would Not Start a Company Today

I say this as somebody who did it, and I say it reluctantly.

In today's environment I would not go self-employed again. In today's environment I would not incorporate again. Not because the work is too hard, since I enjoy work and I do a great deal of it, but because the share of the work that produces nothing has grown large enough to consume the rest.

Take the bookkeeping on its own, with its receipts, its filings, its entries, its formats, its deadlines and its proof of everything. Technically you could solve that in a week today and then forget about it. Connect the account, income in, expenses out, automatic categorization, automatic filing with the tax office, done. The technology exists, it works, and it would cost the state less than the control apparatus we run instead. We do not do it. We insist that every individual maintain a small administrative office in which he is the only employee.

Then come the obligations, something new almost every month, sometimes from whichever government we live under, sometimes from Brussels. Brussels by now is an enormous apparatus whose output is measured in rules, and an apparatus measured in rules produces rules. That is not an accusation aimed at individuals. It is a description of a system, and systems do what you reward them for.

From here on I am marking my opinion as opinion, so that nobody mistakes it for a finding. I consider the direction we are moving in to be wrong. I once took us for free people, and that feeling is gone. What I am living through reads to me as a western variant of something we all recognize from another part of the world and enjoy being outraged about: control that does not present itself as control, but as due diligence, as a transparency requirement, as a documentation obligation. If the decision were mine, Germany would have left this union long ago, and I would have back the border controls that were abolished out of convenience. That is my political position, it is not the position of an institute, it is not the result of a study, and anyone who does not share it is under no obligation to. I have simply lost the appetite for keeping it to myself.

Please do that thinking for yourselves. I am not demanding that you arrive at my conclusion. I am demanding that you do the arithmetic at all.

What Applies From Today, and What I Will No Longer Read

To leave no room for misunderstanding, here is the practical part.

The apps are deleted from my devices. Not disabled, not hidden, deleted. There will be no new posts there. Everything I published up to now stays online, because deleting it would be childish and the work was real.

I no longer read messages on those networks. No direct messages, no comments, no requests, nothing at all. Please do not write to me there in the first place, it will not arrive. I also no longer have the appetite or the willingness to work my way through the dozens of communication channels that exist today. Every channel is one more place where something important might be sitting, and eventually you spend the day searching instead of working.

Anyone who wants to stay in contact reaches me through Telegram or by email at george@rauscher.xyz. Everyone who knows me knows both routes. Anyone for whom I am not worth that single step is welcome to disappear, and I mean that without any bitterness.

Since the question will come up regardless: this decision is final. People who know me know that already. I think for a long time, I walk down every branch, I let a matter sit for days before I react, and once I have decided, the subject is closed. That holds for systems and it holds for people. Honesty and loyalty are not virtues I hope for, they are the precondition for letting anyone into my life at all. Whoever violates them is not punished and not insulted. That person is simply removed from my head, completely and without an epilogue, and afterward he no longer exists for me. It sounds hard, and it is hard. It also happens to be the only form of consequence that works without resentment.

I am writing this text in frustration and in anger, and I admit that openly because I normally do not operate this way. Ordinarily I let irritation sit for 2 or 3 days and never react on the spot. Today I am not holding to that, because this exit was overdue anyway, the anger only accelerated it, and carrying fewer apps on a device that already knows too much is a security gain rather than a loss.

Run the Numbers Once, and You Will Understand Me

I am about to shut the machine down. Then I delete the last remnants from my devices, and then I take Bandit into the woods.

He will put his nose to the ground and follow a trail that stopped existing hours ago, and I will walk beside him and think about nothing in particular. No feed, no notification, no question of whether this image qualifies as a deepfake within the meaning of Article 3 number 60. Only forest floor, a dog who knows exactly what he wants, and a man who sat in bed this morning and spent 3 hours programming something that helps nobody.

Before I go, let us do the arithmetic together, because you know me, I enjoy arithmetic.

Assume you spend exactly 1 hour a day on social networks. That is a modest figure, believe me, well under what most people actually manage. Now run that single hour across 50 years. 50 years come to 18,262 days, which gives you 18,262 hours, which is 761 solid days in a row, day and night, without a break. Slightly more than 2 years of your life.

Now let us run it honestly, because you do not scroll in your sleep. Measured against 16 waking hours a day, that comes to 1,141 waking days. That is over 3 years at full consciousness, for a single hour a day.

So the next time you put down your fork after the second bite because the device beside the plate lit up, you will have a rough idea of what that reach costs across a lifetime. I have decided to spend those 3 years in the woods instead.

And yes, I am enjoying it.

References

• European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence. OJ L, 2024/1689, 7/12/2024. http://data.europa.eu/eli/reg/2024/1689/oj
• Fiorillo, C. D., Tobler, P. N., & Schultz, W. (2003). Discrete coding of reward probability and uncertainty by dopamine neurons. Science, 299(5614), 1898–1902. https://doi.org/10.1126/science.1077349
• Galter, B. (2026). KI-Kennzeichnungspflicht. Was Artikel 50 EU AI Act ab dem 02.08.2026 verlangt. Sicherio. https://sicherio.de/ratgeber/ki-compliance/ki-kennzeichnungspflicht
• Gonser, N. (2025). Zunehmend digital. Zur Mediennutzung des Publikums ab 50 Jahren. Ergebnisse der ARD/ZDF-Medienstudie 2025. Media Perspektiven, 29/2025. https://www.media-perspektiven.de/
• Lindström, B., Bellander, M., Schultner, D. T., Chang, A., Tobler, P. N., & Amodio, D. M. (2021). A computational reward learning account of social media engagement. Nature Communications, 12(1), 1311. https://doi.org/10.1038/s41467-020-19607-x
• Montag, C., Marciano, L., Schulz, P. J., & Becker, B. (2023). Unlocking the brain secrets of social media through neuroscience. Trends in Cognitive Sciences, 27(12), 1102–1104. https://doi.org/10.1016/j.tics.2023.09.005
• Müller-Peltzer, P. (2026). Kennzeichnungspflicht im AI Act. Was die neue EU-Leitlinie für Deepfakes klarstellt. Schürmann Rosenthal Dreyer Rechtsanwälte. https://www.srd-rechtsanwaelte.de/blog/kennzeichnungspflicht-im-ai-act-was-die-neue-eu-leitlinie-fuer-deepfakes-klarstellt
• Statistisches Bundesamt. (2025). Stellenabbau in der Automobilindustrie. 48 700 weniger Beschäftigte zum Ende des 3. Quartals 2025 als ein Jahr zuvor. Pressemitteilung Nr. N067 vom 20.11.2025. https://www.destatis.de/DE/Presse/Pressemitteilungen/2025/11/PD25_N067_42.html
• Statistisches Bundesamt. (2026). Steuereinnahmen 2025 summieren sich auf rund 990 Milliarden Euro. https://www.destatis.de/DE/Themen/Staat/Steuern/Steuereinnahmen/steuereinnahmen.html

Disclaimer: This article reflects the author’s own research and opinion as of the publication date shown above; later findings or legal changes may have overtaken it, so always check that date. Sources are cited for independent verification, and no liability is accepted for third-party studies. This is general information, not medical, legal, or professional advice: for medical questions see a doctor, for legal questions a lawyer, and in an acute crisis contact emergency services or a crisis helpline.

Read the full disclaimer

About the header image: it is AI-generated. Cheaper than a photo shoot, and I have made my peace with the age of AI. Everything inside the article is real, the diagrams, the skulls, the findings, and every word. The machine gets the opening shot and not one inch past it.