Index of everything on this site: https://ai.hiina.space/index.md
"Can a magic circle be a program?" turns out to be two separable questions, and prior art has answered each half in isolation: notation (Mystical, VIPR, Gallifreyan) and recognition (the 2026 fan simulators, Arrow R-CNN, DECIMER). The missing piece is a semantics chosen to suit the recogniser. Once meaning lives only in what circle fits, angles, containment and perfect matchings measure reliably, the parsing problem shrinks to something a headset CPU or a phone photo handles, and the familiar failure points (arrow topology, VLM counting, precision-gated frustration) are designed out instead of fought.
The more interesting finding is that "how well it's drawn" has no single right meaning. Wild magic from calibrated posteriors, analog drift as chorus, and neatness as mana efficiency are three distinct and defensible readings, each with a different fairness profile, and the accessibility knob (λ, β, or the steady assist) is part of the language definition rather than a setting bolted on afterwards. That is the part no shipped game or fan project has worked out, and the one-evening Widdershins experiment is the cheapest way to find out whether the first of those readings survives contact with a real hand in mid-air.
Text licensed CC BY 4.0 (attribution: Hiina, ai.hiina.space). Source: https://github.com/hiinaspace/ai-hiina-space