N'Ko Program — Master Consolidation Index

N'Ko Program — Master Consolidation Index

> Cross-repo inventory of the entire N'Ko line of work. Built 2026-05-31 by
> sweeping every N'Ko artifact modified in the last week. This is the front door
> ABOVE `nko-brain-scanner/PROGRAM.md` (which only covers the research papers).
> PROGRAM.md = the science. This file = the whole program (papers + protocol +
> apps + synthesis + audio).

## One-line thesis (unchanged across every artifact)
N'Ko is computational infrastructure — a designed, bijective, tone-marking
script that changes what models can represent, how acoustic evidence aligns to
symbols, whether an error rate is meaningful, and how sound/motion can be encoded.

---

The surface, organized into 4 layers

Layer A — The science (papers + the codebook)

### Layer B — The protocol / economic layer
| Artifact | Path | What it is | Status |
|---|---|---|---|
| N'Ko Compute Network whitepaper | `Desktop/NKo-Compute-Network-Whitepaper.md` | Proof-of-Linguistic-Competence on Bitcoin L2 (Stacks/EPOCH, 8 Clarity contracts, $21.64 testnet). Workers' N'Ko fluency IS the compute. | drafted (March 2026) |

### Layer C — Applications
| App | Path | What it is | Status |
|---|---|---|---|
| NKoScribe | `Desktop/NKoScribe/` | On-device N'Ko ASR iOS app (whisper.cpp xcframework). v1.0 on TestFlight w/ stub. | v1.1 mid-task, BLOCKED on ASR checkpoint (Vast/HD1) |
| NKoMathLab | `Desktop/NKoMathLab/` | SwiftUI math-lab learning the FAC/Anticipatory-Transformer/choreography math. | NEW; both builds pass 2026-05-29 |

### Layer D — The synthesis arrow + narration
| Artifact | Path | What it is |
|---|---|---|
| Computational Choreography ↔ N'Ko | `Desktop/LUME-CC/07-nko/` (overview, movement-notation, maoe-routing) + `computational-choreography-nko-audio/` | N'Ko as a MOVEMENT score (the SYNTHESIS arrow: N'Ko→body motion). A distinct generator, not a front-end swap. |
| Narration audio | `Desktop/FAC-NKo-Documentation-1.3x.mp3`, `FAC-paper-1.3x.mp3`, `Semantic_Theory_NKo.mp3`, `MAOE-NKo-Onyx-`, `computational-choreography--audio/` | OpenAI/Gemini TTS narrations of the docs. |

---

## The two arrows (the unifying frame from PROGRAM.md)
- ANALYSIS (signal → N'Ko): ASR / FAC-encoder + the z_t/AGP recognition stack. This is the brain-scanner program.
- SYNTHESIS (N'Ko → signal): FAC decoder (sound) + the choreography generator (N'Ko score → motion). The codebook is the shared notation; the generators are separate builds.

---

## Honest blockers (carry everywhere)
1. ASR checkpoint not local — trajectory-CTC `.pt` on Vast.ai `/workspace`; HD1 unmounted. Blocks end-to-end toned-CER AND NKoScribe shipping.
2. One read-speech clip — the only thing between FAC and its first real acoustic-tone number. Harness built (`run_read_speech_eval.py`).
3. OCR tone-gold unreliable — two SOTA models agree on only ~30–50
4. Tone-mark inventory bug — was fixed in the codebook + FAC paper (U+07EE = native falling tone, not a length mark). MUST verify the flagship doesn't still carry it.

---

## Consolidation decision (DECIDED + DONE 2026-05-31)
- Structure: Flagship + companions. The canonical paper is the umbrella; FAC,
MAOE-N'Ko, and the Compute Network stay separate companions.
- Name: "N'Ko as Computational Infrastructure" + subtitle "Script-Native Speech
Recognition, a Phonemically Interpretable Error Metric, and Admissible Tone Correction."

Flagship UNFROZEN + rebuilt (`paper/current/paper_canonical_nko_agp_20cer.tex`):
1. Retitled + subtitle + date (revised 31 May 2026).
2. Abstract: added the tone/reconstruction lever + a closing paragraph naming the
companions and the prior×evidence tone frame.
3. NEW section "The Reconstruction and Tone Pillar" (before Limitations):
tone = text prior × acoustic evidence; corrected 7-mark inventory (U+07EB–U+07F1,
U+07EE = native falling, designed extension is purely timbral); provisional
empirical tone prior (~75
bar); the decode→govern→correct convergence figure (ASCII, reuses z_t); two-arrow
analysis/synthesis frame; program-structure + companion-papers subsection.
4. Limitation #3 + Conclusion updated so tone is no longer "pure future work."
5. Builds clean: 31 pages, 0 undefined citations, 0 undefined refs, 0 missing chars.

Tone-mark bug: flagship was NEVER buggy (line 427 already used the correct
U+07EB–U+07F1 range); the new subsection makes the corrected inventory explicit.

⚠ Still untracked in git — flagship tex+pdf are `??` in `nko-brain-scanner`.
Versioning the flagship is the recommended next step.