Good interfaces help us see and express. Abstraction compresses reality into models we can hold in our heads. The best abstractions are necessarily lossy, but that's a feature, not a bug.
Podcast: Linus Lee - Engineering for Aliveness - Jackson Dahl
What makes a tool suitable for creative and exploratory work? Linus Lee defines good tools designed for creative and exploratory work around two core ideas: seeing and expressing.
A good interface lets you see information clearly from the right perspectives. It also enables you to express your intent as naturally and precisely as you desire. These two abilities define how human agency operates through tools.
Abstraction makes these two abilities possible because it gives us models we can fit in our heads and work with.
Think about different maps of the same city. Terrain maps. Transit maps. Road maps. Each abstracts differently. None is perfectly accurate. But each serves its purpose precisely because it's lossy in the right way. "All models are wrong, but some are useful."
This applies to programming languages as well. Python is easier than C because it hides more complexity. That makes it more powerful for most tasks. You can see what the system does and express what you want more clearly.
Innovation in tools often manifests as better abstraction and representation.
And that's why AI is such an interesting technology. LLMs think geometrically. They translate words into shapes in latent space. Most of their processing happens with ideas represented as shapes in high dimensional space, which can be a very powerful abstraction.
For Linus, the dream is to externalize this. Build interfaces where ideas are literally shapes. Incompatible concepts wouldn't fit together. Compatible ones would snap into place, and this might help us work with the complexity that our current tools can't handle.
Two further readings if you are interested: