Between, not about
A language about abstraction would be a set of terms you apply to paintings. That is criticism, and it already exists.
Abstract art↔language↔people
What is being built here stands between two parties and does what a language does: lets one side reach the other. The painting has never been able to say what it is. The person has never had the words to say what happened. Both problems are the same problem, and it is not a problem of taste.
The difference from every other language is that this one was made rather than inherited. Nobody grew up speaking it. Every part of it — the dimensions, the names of the shapes, the geometry of the mark — was decided by somebody, which means every part of it can be wrong and can be shown to be wrong.
The parts of it
- Unit
- The ArtGram. One mark carrying one painting’s whole reading. The smallest thing in this language that means anything on its own.
- Vocabulary
- The eight shapes. Stone, Field, Blade, Keel, Fan, Shield, Spire, Rudder. Not categories with edges — named regions of a continuum, which is what most words are.
- Grammar
- The relationships between ArtGrams. This is the part still forming, and it is worth saying so.
- System
- ArtGraph. The whole through which a person and an abstract painting can reach each other.
One thing combines. When several paintings have moved the same person, the card resolves them into a single new mark — several ArtGrams weighted by what they did, drawn as one. That is the first real syntax in the framework, and it runs through a person rather than between two objects. It is where the grammar starts, if it starts anywhere.
Where the machine sits
A language model is not the language. The ArtGram is the language. AI interprets, enriches, relates and scales what the language produces — it does not supply the terms and it did not decide what a dimension is.
Concretely: a model read 224 paintings blind and its readings are published beside the author’s in the analysis, including where they disagree. A model runs the Concierge, translating what a person says into terms the framework holds. Neither of those invented anything. Both are a made language being spoken at a speed a person could not manage alone.
The distinction matters because the opposite reading is available and wrong: that a machine looked at some paintings and produced a taxonomy. It did not. A painter decided what could be asked of a painting, and the machine applies it.
Three layers
- ArtGraph
- The language.
- The graph
- The relationships that language reveals.
- AI
- How that language operates at scale.
The name has been making this promise since it was chosen. Art is the abstraction. Graph is what connects. Most of what exists so far is the first half.