Links, April 11, 2025

§Item Flow, Part 1: A new unified concept for layout

CSS Grid and Flexbox brought incredible layout tools to the web, but they don’t yet do everything a designer might want. One of those things is a popular layout pattern called “masonry” or “waterfall,”

This is good and I hope to see it come to fruiition.

§ffmpeg for Arists: A Cookbook for Audio and Video Processing on the Command Line … with Examples!

There’s a joke that ffmpeg is a programming language in its own right, and that’s not undeserved. That said, it’s an incredibly powerful tool, and I personally am a fan of the cookbook format.

§The Future is Federated

Great short film series about the Fediverse.

No I won’t stop dissing “AI”

§On correct but wrong responses

Improv-brain” relates to LLM systems’ tendency to not challenge input but just roll with it. Again, this stems from lack of knowledge, understanding and competence but has a bit of a different bend: LLMs try to go “yes … and” on whatever you tell them and rarely challenge prompts. Which is why a suboptimal prompt can lead them astray so easily.

When I say that, “what takes my time is not writing code, but understanding the problem”, this is in part what I’m talking about.

§Immaterial Sensations

The idea that you can “democratize artistic expression” at all only makes sense if you first believe that the capacity for art is innate and unchangeable, something you’re either born with or will never have.

I think one of the saddest features of our culture is how so many people are convinced many people that they simply are not capable of artistic expression.

§Most Americans think AI won’t improve their lives, survey says

Bender told Ars that pretending AI experts know best is likely problematic because “AI experts don’t know very much about how people who work in other fields do their jobs, and the people who do those jobs are the ones who know what that’s about.”

§Generative AI in Servo

Servo is an in-progress embeddable browser engine. There was a recent controversy about reversing its ban on LLM & probabilistic tool contributions, and this is a great summary about why they’re keeping the ban and places where machine learning tools like speech recognition may prove to be useful.

§A Normal Person’s Explainer on What Generative AI Is and Does

We make a couple bad mistakes when we interact with these giant piles of vector math and statistics, running on servers all over the world. The first is assuming that they think like us, when they have no human-like thought, no internal world, just mapping between words and/or pixels.

I recently sent this to an IRL personal connection whose boss wanted to use ChatGPT to help set state health policy. Meanwhile, legislation was introduced to make chatbots qualify as practitioners able to issue medication prescriptions.