A framework for describing the morphology of informational flows using 6‑bit sliding windows, 64‑state frequency profiles, and three fundamental metrics of form: entropy, anisotropy, and morphological tension. SUBIT‑morphs provide a universal, model‑independent way to characterize the shape of any data stream.
SUBIT‑morphs treat information not as text, symbols, or semantics, but as form.
Any stream of bytes can be converted into a sequence of bits, segmented into overlapping 6‑bit windows, and mapped into a 64‑state distribution. This distribution becomes a morphological fingerprint of the stream.
The result is a compact, comparable, model‑agnostic representation of informational structure.
A data stream is converted to bytes, then to bits.
A sliding window of 6 bits produces values in the range 0–63.
Each 6‑bit value increments one of 64 counters.
Normalized frequencies form a 64‑dimensional vector.
The 64 states are arranged as:
row = k // 8
col = k % 8
This produces a stable 8×8 grid representation.
An alternative geometric embedding:
x = (k >> 4) & 0b11
y = (k >> 2) & 0b11
z = k & 0b11
This reveals volumetric structure and directional patterns.
Measures randomness of the 64‑state distribution.
Measures directional structure using covariance eigenvalues.
Measures radial variation around the distribution’s center.
Together, these form the SUBIT‑address:
(E, A, T)
This coordinate locates any informational flow in morphological space.
SUBIT‑morphs provide:
- a universal morphological unit
- a model‑independent representation
- a compact invariant for comparing data streams
- a bridge between human‑readable information and latent geometries in AI systems
- a tool for interpretability, clustering, anomaly detection, and style analysis
They apply equally to natural language, code, DNA, audio, logs, and model outputs.
subit-morphs/
│
├── README.md
├── LICENSE
├── CHANGELOG.md
│
├── /spec/
│ ├── subit-morph-spec.md
│ ├── subit-address.md
│ ├── metrics-E-A-T.md
│ └── artifact-claude.md
│
├── /examples/
│ ├── pi.txt
│ ├── dna-sample.txt
│ ├── code-snippet.py
│ ├── natural-language.txt
│ └── model-output.txt
│
├── /analysis/
│ ├── example-morphs.md
│ ├── morphological-space.md
│ ├── comparison-table.md
│ └── interpretability-notes.md
│
├── /src/
│ ├── python/
│ │ ├── subit.py
│ │ ├── metrics.py
│ │ ├── grid.py
│ │ └── utils.py
│ └── pseudocode/
│ └── subit-algorithm.txt
│
└── /artifacts/
├── claude-subit-morphs.html
├── pi-morph.html
└── README.md
The /artifacts directory contains interactive and narrative materials generated during the development of SUBIT‑morphs. These artifacts illustrate conceptual foundations, demonstrate morphological analysis, and provide visual examples such as the π morph.
The /examples directory includes sample data streams used to demonstrate SUBIT‑morph computation:
- mathematical constants
- biological sequences
- programming code
- natural language
- model‑generated text
Each can be processed to produce a 64‑state grid and SUBIT‑address.
The /src/python directory contains a minimal reference implementation:
- convert data to bytes
- extract 6‑bit windows
- compute the 64‑state distribution
- generate the 8×8 grid
- compute E, A, T
This implementation serves as a baseline for experimentation and integration.
SUBIT‑morphs support:
- morphological search and clustering
- style and authorship analysis
- anomaly detection
- model interpretability
- comparison of architectures and outputs
- morphological control of generative systems
They provide a universal morphological layer that complements existing semantic and statistical methods.
MIT
- optimized implementations
- visualization tools
- morphological embeddings
- model‑level integration
- expanded artifact gallery
- research papers and benchmarks