🜂 SZL Holdings estate artifact — the estate that receipts its own thinking.

MiniEmbed-Nano

A deterministic 64 × 12 NumPy table for inspecting token hashing, pooled vectors, and small retrieval examples.

Artifact: Atelier NumPy embedding table · Stage: Software / reference

Explore in Command Lab · Build · Evidence

Before you use it

• This table is not a trained neural embedding model; it does not establish a comparison with a foundation embedding model.
• The five-pair retrieval result is SAMPLE. Historical MEASURED and reproducibility labels below are retained source claims, not new independent validation.
• Bind the exact archive, seed, normalization, loader, and revision before use. Khipu's reference card and this Atelier export have separate evidence.

Technical details and evidence

The original Atelier description is preserved below as historical source. Its architecture, comparison, measurement, and signing language does not qualify another archive or establish a new result.

MiniEmbed-Nano

A 64×12 embedding whose rows are a function of SHA-256. The vector is the receipt of the token.

Family. nano · Evidence. MEASURED · Weights. numpy · Params. 64 × 12

Hub: SZLHOLDINGS/MiniEmbed-Nano

The cut

Nobody ships an embedding where retrieval is provenance. MiniEmbed is not BGE, not NV-Embed. Mean-pool of L2 rows, seed 20260721, CPU NumPy.

An embedding you can re-derive bit-exact from the seed. No SGD theater. Cosine is a claim about the table, not a vibe.

Silhouette → leave → SZL

Leader	Take, then tweak
Anthropic	Honest about what it is not — not a foundation embed.
NVIDIA	Tiny table instead of NV-Embed / NeMo retrieval stacks.
Unsloth	No LoRA. Construction is build(seed) — Unsloth is the wrong tool and we say so.

Nobody else ships this combination. That is the point of a one-of-one.

Intended use

Silhouette of receipted retrieval. Teaching and tests.

Limitations

• Not a neural embed.
• Not comparable to BGE-base 768-d.
• Hit@2 is on five doctrine pairs — SAMPLE.

Honesty

Claim	Label
This card's numbers	MEASURED
Energy / joules	UNAVAILABLE unless a signed meter says MEASURED
Λ uniqueness	Conjecture 1 OPEN — not a theorem
GGUF as the signed object	FALSE

Doctrine v11 LOCKED · 749 declarations · 14 axioms · 163 sorries · locked-proven 8.

Apache-2.0. Copyright 2026 SZL Holdings · Stephen P. Lutar Jr. · ORCID 0009-0001-0110-4173.

---

Evidence binding (atelier card alignment, 2026-11-14)

• This card is aligned to the atelier source card cards/MiniEmbed-Nano.md at szl-holdings/szl-atelier revision 5054423228e500dc29bf3d6b719288607c3a4a8d; source card sha256 bd0cfa811f7005e83430d5ecf365f6dc21bf68674d223f70480010a2112ec8ed.
• Observed Hub revision immediately before this edit: 70f95abc63fd06625ef499cee5c0a3df717fc6fd. The prior card remains in revision history.
• Atelier class: MEASURED (source-side input, not a Hub-observed fact). No new measurement, evaluation, or weight change was performed by this edit. Any metric appearing above carries its own receipt binding in the source card; absent evidence is stated as absent. Failures and dated results are preserved at their declared scope.

Signed-off-by: Stephen Lutar stephenlutar2@users.noreply.github.com

Measured evidence: TRAINING RECEIPT BOUND (2026-11-14)

• Receipt file: TRAINING_RECEIPT.json (sha256 588a4a023e66…)
• Receipt facts: {"conjecture_1": "OPEN", "cpu_numpy": "LIVE", "cuda": "UNAVAILABLE", "doctrine": "v11 LOCKED", "energy_status": "UNAVAILABLE", "honesty": "REPORTED", "id": "szl-khipu.training.v0.1.0", "proven_trust": false, "seed": 20260721, "ts": "2026-08-29T17:11:32.518042+00:00"}
• Note: receipt file already present in-repo at audit; card binding added 2026-11-14. Re-public earned per estate doctrine (receipts earn visibility).