Research track

Four domains draw the same loop 네 영역, 하나의 이상한 고리

Escher's drawings, Bach's canons, Gödel's theorems, the self-attention of transformers — four seemingly unrelated domains share a single three-stage formal model: reduction → self-reference → emergence. The claim that emergence occurs once self-reference crosses a critical complexity, and that this structure is formally isomorphic across domains, is the Threshold Recursive Isomorphism Hypothesis (TRIH).

109k thesis characters · 70 references
One way of thinking

Reduction → self-reference → emergence

A picture that contains itself, recursively

In Drawing Hands two hands sketch each other, and in Print Gallery a figure inside the picture is looking at the very picture that contains him. The reduced units of line and plane are recombined into a structure that draws itself. Beginning from the symmetry rules of wallpaper groups and hyperbolic distance, an irreducible visual infinity emerges.

§4 Ricercar ↔ Transformer is treated as a formal parallel — the stronger claim of isomorphism is reserved for the visual and logical domains.

Isomorphic figures

Three domains, one self-referential structure

§4 Bach · §5 Gödel · §6 Transformer — self-made conceptual figures (SVG) of each domain's self-referential structure.

axis forward retrograde Crab Canon — forward / retrograde
§4 · Bach · Crab Canon One voice forward, the other in retrograde — a crab's walk across the axis of symmetry.
¬∨∀ symbols 2ᵃ·3ᵇ·5ᶜ prime encoding G the number names its own proposition ¬ ∨ ∀ → 2ᵃ·3ᵇ·5ᶜ → G
§5 · Gödel · numbering A string of symbols is encoded as prime exponents into a single number G, which loops back to name its own proposition.
layer 1 · previous-token head layer 2 · induction head match prev A → copy its B predict ? = B A B C D A ? induction head — [A][B]…[A] → ?
§6 · Transformer · induction head A two-layer circuit where attention looks back to the B after a prior A to predict the next token — self-attention's self-reference.
Publications & public records

Verifiable scholarly records

  • 2026

    Self-Reference and the Possibility of Mind — Proposing TRIH

    Self-Reference and the Possibility of Mind

    유승훈 (Seunghun Yu) · Published on Zenodo (Jun 2026) · preparing journal submissions (Leonardo, Cognitive Science)

    Across four domains — the visual (Escher), the auditory (Bach), the logical (Gödel) and the computational (Transformer) — self-referential structure is treated under a single three-stage formal model: reduction → self-reference → emergence. The Threshold Recursive Isomorphism Hypothesis (TRIH) holds that emergence occurs once self-reference crosses a critical complexity, and that its structure is formally isomorphic across domains. In the Transformer's self-attention, a triple line of circuit-level evidence — induction heads, monosemantic features and attribution graphs — makes the self-referential circuit directly observable. This lifts a discourse once confined to analogy into a claim demonstrable at the circuit level.

Research interests

An interdisciplinary program around TRIH

AREA 01

Self-reference · strange loops

A formal system pointing to itself through Gödel numbering. Mathematical logic & computability (Gödel, Turing, Church).

AREA 02

Emergence · criticality

The point where self-reference, crossing a critical complexity, gives rise to a new hierarchy. Complex systems & the free-energy principle.

AREA 03

AI / LLM interpretability

Observing the self-referential circuits of transformer self-attention at the circuit level. Mechanistic interpretability.

AREA 04

Consciousness · cognitive science · Eastern thought

Bridging the Hard Problem and Global Workspace Theory with the interdependence of Buddhist dependent-origination and Huayan, through self-reference.