Florian's Filter — W40 2026
The gap between what AI promises and what it actually delivers creates a strange psychological oscillation. I move from the awe of "this is the future, and we are living it!!!" to the skepticism of a hype cycle, often within the same hour. Finding stability in this landscape requires both critical distance and a focus on specific, sovereign utility.
Articles
- Don’t be fooled by this summer of AI hype
Timnit Gebru and her colleagues provide much needed voices in the AI space. This piece for MIT Technology Review examines how breathless claims about AGI often collapse under scrutiny. - Kolibri Has Landed: A Sovereign Open-Weight Model — Aleph Alpha
Aleph Alpha introduces Kolibri, an English-German Mixture-of-Experts model. This approach to transparency and EU regulation is a welcome addition to the current GenAI landscape. - How the Sovereign German LLM Works
Tejas Kumar did a great job of summarizing what Kolibri is about - what it consists of, how it came to be, what it is good at, and where it is behind its competition. If you are not familiar with GenAI Model "Whitepapers", read this before you use the Aleph Alpha documentation.
Videos
- The Race For AGI (creative commons)
Darren P. Orr presents a short and (to me) very funny visualization of the AI race. At the same time, this video serves as a testament to the current quality of AI-generated deepfakes. - Opus 5.5 is disturbing
Less Bitter offers an honest and relatable take on the shifting capabilities of GenAI. The video captures the oscillations of feeling both impressed and unsettled by new model iterations. I have tried to go "above and beyond" in terms of building websites with animated visuals etc. this week, inspired by this video - and I was seriously blown away by what I got out of that.
Tools
- Sixteen years of using one percent of an app
Reflecting on a long-term relationship with Alfred, an app where a tiny fraction of the feature set provides all the value. A short declaration of my love for utility over complexity.