Introduction
What's this all about?
Start here: Escaping the flatland
Ideas? Soon you can join the conversation. Hold tight!

Protecting and promoting public good as a commercial entity

What's the most effective way to protect and promote public good as a commercial entity?

Studio Kare is currently focused on one question: what’s the most effective way to protect and promote public good as a commercial entity?

The concrete shape of this right now is a business plan for translating the computational advantages of LLMs to entities that serve the public good, ANBIs (Algemeen Nut Beogende Instellingen) and related public-benefit organizations.

Why this is a difficult problem

Public good and commercial incentives are often pulling in different directions. The hard part isn’t arguing that public good matters — it’s designing a structure where serving the public good is the path of least resistance for a commercial entity, not a tax on it.

LLMs shift the cost of computation. That changes which models of public service are even feasible. The question is which of those newly-feasible models survive contact with a commercial legal form.

Open questions

  • Which legal forms (ANBI, stichting, BV with a statutory public-purpose clause) actually align incentives versus just signal intent?
  • Where does the computational advantage of LLMs land for a public-benefit entity — cost reduction, reach, new service categories?
  • How do you keep the public-good commitment load-bearing under commercial pressure, instead of decorative?

Please get in touch if your thinking about this, have ideas, built something.

Extracting information that feels 'alive'

AI tools that provide information can be soul-numbing.

AI tools that provide information often feel “flat.” (Inert? Emotionally unlit?)

This makes learning harder, it doesn’t grip your attention.

The sense of “casting information in 3D” (making it rich with detail and emotionally awake) I’m after is a very subjective sense. That subjectivity is a hint that the last mile belongs to humans: teachers, friends, writers.

I miss LLM-based tools that can support that last mile without pretending to replace it.

  • LLM-based tools often try to “encourage curiosity,” but the encouragement can itself fall flat. It doesn’t kindle curiosity; at best it’s oxygen to a fire that’s already burning.

Log

2026-06-24 Moving to backburner. After imagining and experimenting with solutions for N=1, I’ve run into dead ends when it comes to going to N=10.

Augmenting memory using our digital footprint

[Andy Matuschak got this covered].