Silicon–Àṣẹ

Critical comparative atlas • Local computation • No spiritual claims

The Silicon–Àṣẹ Question

What becomes visible when transformer mathematics, Ifá and Candomblé vocabularies, Kabbalistic histories, and the golem archive are placed beside one another—without declaring them identical?

Enter the first study gate
01

Comparison is not conversion. Shared shapes do not establish shared history or ontology.

02

Simulation is not consultation. Nothing here produces a sign, verse, remedy, or spiritual diagnosis.

03

Authority stays human. Communities, practitioners, scholars, and source holders govern meaning and access.

Original AI-generated interpretive artwork; no ceremony or sacred interior is depicted.

Gate I / Representation

Form before meaning

A computer can record sixteen binary orientations exactly. That does not make the record an odù, a consultation, or a spiritually efficacious cast.

Formal correction

One sixteen-object binary configuration has 216 = 65,536 microstates. Counting one orientation yields 17 categories, 0 through 16—not automatically sixteen. Two unrestricted counts therefore yield 17 × 17 = 289 ordered pairs.

When does 16 × 16 = 256 apply?

Only after a named, versioned, source-governed rule maps each result into exactly sixteen categories. If that premise is established, the ordered sign pair has 256 cells. This atlas supplies no religious mapping table.

Local learning simulation

Sixteen-object state editor

Never transmitted

Activate each object to cycle through unset, orientation A, and orientation B. These are neutral study states—not open/closed religious classifications.

0 orientation A 16 unset

Complete all sixteen states to record a set.

Set A count—
Set B count—
Raw pair—
17 × 17 index—

No sign name, verse, polarity, offering, or spiritual guidance is generated.

Study objects are visual abstractions; the interface does not reproduce a house-specific rite.

Gate II / Encoding

Language enters a model as tokens

Prompting changes the model’s current activations and probability distribution. Ordinary inference does not rewrite the trained parameter weights.

texttoken IDsembeddingsactivationslogits

This demonstration uses a transparent word-and-punctuation splitter plus deterministic hashes. It is not any commercial model’s tokenizer or embedding model.

Finite context demonstration

Watch a sentence become bounded units

0/24

Twenty-four empty context positions

Units found0
Loaded0
Outside window0

A token is not an incantation, sacred phoneme, or quantum of àṣẹ.

Original abstract artwork visualizes discretization; it contains no sacred or readable script.

Laboratory / Relation

Attention is a weighted mixture. A pair rule is a lookup.

Both can be written with matrices. Their matrices do different jobs.

Transformer self-attention

Attention(Q, K, V) = softmax(QKT / √dk)V

  • Every position produces a query, key, and value through learned linear projections.
  • Each row of scores is normalized by softmax.
  • The result is a weighted sum of value vectors.

Versioned ordered-pair resolver

result(a, b) = eaT G eb

  • G is a sparse or complete table approved for one named rule set.
  • Order may matter: (a, b) need not equal (b, a).
  • A missing cell stays unresolved; a model must not improvise it.

Inspectable computation

Two-vector attention demonstration

Lower values sharpen this demonstration’s row distributions; this is not a spiritual “energy” control.

Run the matrix to inspect four normalized attention weights.

Category boundary

Throw A is not Q. Throw B is not K. Guidance is not V. Query, key, and value are learned projections computed for every model position—not aliases for two divinatory results.

The split visual keeps the two mechanisms distinct; no correspondence table is implied.

Comparative atlas / Historical layers

Three geometries. No master key.

Numbers can organize comparison. They cannot manufacture a transmission history or a canonical crosswalk.

Try the exact map

Can 10 + 22 be paired one-to-one with 16 × 16?

The types and cardinalities have not yet been compared.

The artwork is an unlabeled comparative triptych, not a canonical Tree, odù diagram, or cosmogram.

Gate III / Authority

The limit is not a missing parameter

Àṣẹ and Brazilian axé are not generic energy scalars. The cited scholarship places them in relations among speech, persons, practices, materials, places, and authority.

A model has implemented structure and learned numerical parameters. There is no technical evidence that a model bears lineage, ritual standing, ancestor relation, embodied accountability, or spiritual efficacy. Saying so is a boundary of evidence—not a benchmark result.

Assemble the responsibility statement

Select every layer the software actually needs

Select all three. Computation alone cannot supply the missing governance.

The golem archive

The famous Maharal-of-Prague creation story is a nineteenth- and early-twentieth-century literary development, not a documented sixteenth-century event. Its productive modern parallel is ethical: makers remain responsible for delegated force.

Ordinary leaves, water, iron, and silicon are shown as material contrasts; no ritual arrangement is depicted.

Comparative lexicon / English • Español • Português • Italiano

A bridge with load limits

Every entry preserves the English source definition and adds a complete Spanish, Brazilian Portuguese, or Italian companion: documented meaning, situated comparison, permitted analogy, and the equivalence that must not be asserted.

The archive cards are intentionally blank; all linguistic content is human-readable HTML, not generated pseudo-script.

Source archive / Exit

Leave with the sources intact

The deepest reading is not that every system hides the same code. It is that form, language, secrecy, embodiment, and responsibility have different histories—and comparison becomes stronger when those differences remain legible.

Your local study path

The archive is open

Viewed sections are counted only for this browser session. No prompt, state pattern, or personal information leaves this device.

Read the expanded manuscript Download the editable DOCX Download the Python matrix simulator Download the eight visual prompt templates

Original silent short film · 25 seconds

Eight thresholds, one critical path

The eight commissioned artworks move from an abstract threshold to an open archive. The sequence depicts no ceremony, sacred interior, invocation, or ritual performance.

Download the journey film
Machine learning

Transformer attention

Vaswani et al. define scaled dot-product attention and the Transformer architecture.

NeurIPS paper
Living heritage

Ifá corpus structure

UNESCO describes sixteen principal signs, 256 odù, an open corpus, and trained diviners.

UNESCO record
Primary text

Sefer Yetzirah

The work’s thirty-two paths comprise ten sefirot belimah and twenty-two letters; later diagrams have their own histories.

Bilingual text
Diagram history

Kabbalistic ilanot

The University of Haifa project documents the varied manuscript genre behind later “Tree” diagrams.

Ilanot—Maps of God
Legend history

The Prague golem

The Maharal association enters print long after his lifetime and is elaborated in modern literature.

NEH source history
Cybernetics

God & Golem, Inc.

Norbert Wiener provides a documented modern bridge among machines, learning, religion, and responsibility.

MIT Press edition

The archive closes the journey with provenance rather than a simulated revelation.

Before the threshold

Enter as study, not initiation

This experience stages a disciplined comparison among formal structures, histories of language, and questions of authority. It does not reproduce restricted knowledge or make a spiritual claim.

Go directly to sources

Method

Analogy under discipline

The atlas uses three evidence labels throughout the companion manuscript:

Documented
A claim supported by a cited primary, institutional, or scholarly source.
Formal
A mathematical or computational statement that follows from declared definitions.
Interpretive
A clearly marked philosophical comparison that does not claim common origin or identity.

Where the requested premise conflicts with evidence—such as “prompts change weights,” “two shell counts automatically make 256,” or “the Prague golem is a sixteenth-century record”—the interface teaches the correction.

Source note

What this app uses—and refuses

The interface draws its technical account from the Transformer paper and current interpretability literature; its Ifá overview from UNESCO, museum records, and the manuscript’s cited linguistic and religious-studies sources; and its Kabbalah/golem history from Sefer Yetzirah, manuscript-history projects, scholarly publications, institutional archives, and Norbert Wiener.

The app intentionally does not include a universal sign dictionary, a house-specific dual-result rule, ritual speech, sacred-name operations, offering instructions, or a database that converts odù, sefirot, Hebrew letters, and model vectors into one table. Full references and limitations appear in Chapters 6 and 7 of the expanded manuscript.