#0006 Agents, Enchanted Knowledge Objects, AI is Worse is Better, A Roadmap To Alien Worlds

Welcome to Constant Flux, a weekly lens taking a systemic view on the polycrisis.

I promise, zero tariff talk. Less AI than last time, more delightful weirdness. Hope you enjoy this edition!

Agents

Agents is a speculative piece that imagines AI not as a single, central intelligence but as a set of small, focused helpers, each with a task, a context, and a reason to exist. A beautiful idea that shifts attention away from the all-knowing black box and toward something more modular, more human-scaled.

What I like here is how it leans toward resilience. In the systems sense. When functions are distributed, no single point of failure brings the whole thing down. It starts to look more like an ecosystem than a machine. More like mushrooms than mainframes. :-)

Honestly, a lot more interesting than the usual Smart City concepts. This concept leaves room a bit of weirdness.

MAKE AI STRANGE AGAIN!

Enchanted Knowledge Objects in LLM UI

This article reimagines how we interact with language models. I love that this is an esoteric take on user interfaces. Documents aren’t just inputs, they’re talismans, filled with intention, shaping the behaviour of the model like magical artifacts artyfacts.

It more magic than tech. You don’t load a file, you invoke it. Managing them like an RPG inventory, equipping, combining, casting, feels less like UI design and more like ritual work.

It wouldn’t feel out of place in The Laundry Files Series (hilarious btw, very recommended), where code and magic are two sides of the same coin. There’s too little of that kind of thinking.

AI is Worse is Better

I guess it’s easy to see the downside of fast, AI-driven software development. Sloppy code, hidden bugs, systems built on sand. I get it, there’s real risk in moving fast.

But there’s power in it too. Or at least I'd like to think so (since my coding skills is becoming obsolete by the day).

I'm convinced that fast, cheap tools open doors. A local government spins up a permit app in a week. A teacher builds a quiz tool with no code.

It’s sure is messy, but systems adapt by trying things, seeing what works, and building from there. Even imperfect tools can unlock new behaviors and tighter loops between need and response. The toy can win.

A Roadmap To Alien Worlds

This piece explores the idea of a planet’s “acquired memory”. The notion that its present state holds traces of everything that has happened on it. Using assembly theory, the authors suggest we can read this memory by looking at the complexity of molecules in a planet’s atmosphere. The more steps it took to make something, the more history it carries.

It’s a planetary systems view. Complexity isn’t just structure, it’s evidence of past interactions layered over time.

Instead of just asking “is there life here right now,” they invite us to ask “has this place been shaped by something that behaves like life?” Love the long view take. Odd but hopeful. Also very Avatar.

Ripples

WikiTok
WikiTok is a web app that lets you scroll through random Wikipedia articles with the same rhythm as TikTok. Endless curiosity!

I like the serendipity of it. Stumbling into knowledge you didn’t know you were looking for (well, it's WikiPedia after all...). All in all, a(nother) nice little detour for the brain.

Scientists Create Sound That Can Curve Through a Crowd and Reach Just One Person
Researchers at Penn State have developed “audible enclaves” using focused ultrasound to create little pockets of sound that only exist in specific spots. You could be in a busy room and still hear a message meant just for you, no headphones needed. A bit Duneish isn't?.

Makes me wonder if it could be flipped the other way. I’d love to carry around a cone of silence.

Piezoelectric Catalyst Destroys Forever Chemicals
A Swiss startup has come up with a way to break down PFAS in water using a piezoelectric catalyst. So no toxic leftovers, no secondary waste. Feels like an important signal that some of these long-term environmental headaches might actually be solvable.