AI That Forgets, Heals, and Speaks Dutch

The Netherlands built its own large language model, trained on Dutch text and governed by Dutch institutions, instead of renting that capability from a US company. If your government or employer handles sensitive data, this is the model for what 'keeping AI local' actually looks like in practice.

Researchers found that when you compress a language model to make it cheaper and faster, it can still ace multiple-choice benchmarks while completely failing to answer the same question in plain text. This matters because a lot of AI products are sold on benchmark scores, and this shows those scores can quietly lie to you.

A study on AI-rewritten radiology reports found that when a model 'cleans up' a doctor's notes, it quietly throws away clinically meaningful details in the process. If you work in healthcare, or are just a patient, this is a concrete reason to ask what happened to the original report.

LLMs can correctly identify that a user is from a different culture, but then respond as if they are not. Knowing something and using that knowledge are two different skills, and these models have one without the other.

Neural Cellular Automata can now grow, damage, and heal detailed textures in real time at HD resolution, which is fascinating to watch and a genuine step toward AI that maintains structure instead of just generating it once and moving on.

Tim Ferriss and others are asking whether AI has already replaced the self-help book, and the Hacker News crowd has a lot of feelings about it. Worth reading less for the answer and more for what it tells you about how people are actually using these tools day to day.