Publikationen
Forschung zu KI-Agenten, Backtesting und Prüfung von Marktprognosen sowie zur Architektur der Plattform.
Sonstiges
Nobody needs to conquer humanity. They only need to run the electricity and the internet. A thought experiment about the quietest possible takeover - and the honest reasons it may never happen.
6 Min. Lesezeit4 Abschnitte
Sonstiges
A clip promised that by 2050 we will turn hundred-year-olds back into twenty-year-olds. I will not defend the date. I want to defend the direction - as a metaphor, a speculation, and a question worth holding honestly.
7 Min. Lesezeit4 Abschnitte
KI-Business
A handful of companies are racing for the top of the most expensive podium ever built, and trillions of dollars rest on one assumption nobody has proven yet. Here is the honest picture, the numbers I can actually source, and why I am not going to tell you to buy anything.
9 Min. Lesezeit7 Abschnitte
KI-Business
A founder's case for the most personal technology ever built - an agent that manages your calendar, your money and your habits, and is willing to overrule you. The interesting question is not whether it arrives. It is who it answers to.
9 Min. Lesezeit10 Abschnitte
Sonstiges
A speculative case, argued from what is actually being built. The fruit fly brain is already mapped, the mouse is underway, and 67 experts put four in ten odds on a digital mind by 2040. Where that road ends is the part nobody can prove yet - including me.
6 Min. Lesezeit7 Abschnitte
KI-Business
A billion dollar valuation does not require a billion dollars of anything. Divide it by the multiples buyers actually paid in 2026 and it resolves to seventeen to forty-three million of recurring revenue, priced off the rate of change rather than the size.
9 Min. Lesezeit7 Abschnitte
KI-Business
Published 2026 data puts the median at about two years for AI companies and about nine for everybody else. The compression is real, its source is unglamorous, and the one company in this cohort that reached a full exit did it by being bought rather than by listing.
10 Min. Lesezeit8 Abschnitte
KI-Business
Roughly 3,000 a month sounds modest until you ask what has to happen on an unremarkable Tuesday to produce it. Every serious route, divided out against published 2026 rates, including the one where the sums stopped closing altogether.
10 Min. Lesezeit7 Abschnitte
KI-Business
A hundred thousand is the first target the published trajectories actually reach. The month-by-month ramp, what is selling on Upwork right now, the tool stack and what it costs, four niches that work, and six things that compress the timeline.
8 Min. Lesezeit9 Abschnitte
KI-Business
Every guide quotes the margin. None does the division. What $10,000 a month requires in five AI businesses - 833 images, 20 brands, four retainers, 555,000 views - and what is actually scarce in each.
7 Min. Lesezeit8 Abschnitte
KI-Business
Model prices are not falling only because the technology improved. Three reasons a giant drives a price toward zero, what each means for anyone building on top, and which positions survive a price collapse.
5 Min. Lesezeit6 Abschnitte
KI-Business
Benchmarks are the most cited and least informative measure of AI progress - saturated, contaminated and optimised against. The four scoreboards being kept at once, and the three properties a scoreboard needs to mean anything.
5 Min. Lesezeit7 Abschnitte
