Inscription Architecture Map

Inscription Architecture Map

The Full System: From Body to Bitcoin and Back

This document maps every file, specification, and implementation related to the N'Ko inscription system, sigils, tokenization, the EPOCH protocol, Stacks/Clarity contracts, PsiChain, the cognitive twin, and anticipation geometry. It traces how they interconnect to form a single pipeline that encodes a life's computational dynamics as hash-chained N'Ko inscriptions settled on Bitcoin.

---

1. The Pipeline (High-Level)

  Physical Body / Agent Session
           |
           v
  z-trajectory (63D sensor / 768D embedding)
           |
           v
  Claim Detector (10 claim types, Rust)
           |
           v
  Typed Claim IR (StabilizeClaim, EchoClaim, ...)
           |
           v
  Lexicon (versioned basin tokens + place tokens)
           |
           v
  Surface Renderer --> N'Ko Line  (e.g. "ߛ [[100-200]] : z(sigma) down ; home ; c=0.85")
           |
           v
  Provenance Witness (InscriptionId = SHA-256 of IR + evidence + lexicon + surface)
           |
           v
  PsiChain Link (prev_hash chain, combining marks for visual depth, zero-width steganography)
           |
           v
  Clarity Contract (nko-inscription.clar on Stacks testnet)
           |
           v
  Bitcoin (anchored via Stacks Proof of Transfer)

The reverse path:

  Bitcoin-anchored inscription
           |
           v
  Clarity read-only query (get-inscription)
           |
           v
  N'Ko surface text + metadata
           |
           v
  Cognitive Twin context injection (style signature + trajectory scalars)
           |
           v
  LLM response generation (personality from LoRA, knowledge from Graph Kernel, dynamics from Anticipation Geometry)

---

2. File Inventory by System

2.1 cc-inscription (Rust crate) -- THE CORE ENGINE

Location: `Desktop/Comp-Core/core/semantic/cc-inscription/`
Lines: 22,881 Rust across 48 source files
Status: BUILT, compiles, deployed as part of cc-mcs-daemon

src/
  lib.rs                          -- Main entry, re-exports all modules
  claims/
    mod.rs                        -- Claim enum + ClaimId, ClaimType, BasinId, etc.
    stabilize.rs                  -- Sigil: STABILIZATION
    disperse.rs                   -- Sigil: DISPERSION
    transition.rs                 -- Sigil: TRANSITION
    return_.rs                    -- Sigil: RETURN
    dwell.rs                      -- Sigil: DWELL
    oscillate.rs                  -- Sigil: OSCILLATION
    recover.rs                    -- Sigil: RECOVERY
    novel.rs                      -- Sigil: NOVELTY
    place_shift.rs                -- Sigil: PLACE-SHIFT
    echo.rs                       -- Sigil: ECHO
  basin/
    mod.rs                        -- Basin state machine
    proto.rs                      -- ProtoBasin (pre-graduation)
    graduation.rs                 -- Proto -> Basin criteria (3 signals)
    lifecycle.rs                  -- Split/merge/retire operations
    constitution.rs               -- Basin invariants
  lexicon/
    mod.rs                        -- Lexicon container
    version.rs                    -- LexiconVersion semver
    tokens.rs                     -- BasinToken, PlaceToken
    changelog.rs                  -- LexiconChange enum
    epoch.rs                      -- LexiconEpoch (O(1) verification anchors, auto-commit at 1000 inscriptions or 24h)
    reinterpret.rs                -- Derived view re-rendering
  surface/
    mod.rs                        -- SurfaceRenderer, NKoLine
    renderer.rs                   -- Claim -> N'Ko line conversion
    grammar.rs                    -- Grammar skeletons per claim type
    slots.rs                      -- Slot renderers (time, basin, place)
    normalize.rs                  -- NFC normalization
  phrase/
    mod.rs                        -- Phrase system (e.g. "arrival turbulence")
    detection.rs                  -- Sequence mining
    compression.rs                -- Compression ratio testing
    registration.rs               -- Phrase registry
  detector/
    mod.rs                        -- ClaimDetector + DetectorConfig
    dynamics.rs                   -- z-trajectory metrics
  integration/
    mod.rs                        -- External bridges
    graph_kernel.rs               -- Slice boundary enforcement, admissibility tokens
    rag.rs                        -- Evidence retrieval for Echo claims
    dell.rs                       -- z-trajectory source (DELL subscription)
    sensor.rs                     -- Jitter alignment for deterministic digests
  canonical/
    mod.rs                        -- CBOR RFC 8949, SHA-256, NFC, QuantizedFloat
  provenance/
    mod.rs                        -- ProvenanceWitness, PureVerificationContext, ArchiveRef
  ontology/
    mod.rs                        -- OntologyOperation pipeline (split, merge, graduate, retire)
  chain_link/
    mod.rs                        -- ChainLink struct, ChainState, verify_chain_integrity
    combining.rs                  -- 9 N'Ko combining marks (U+07EB..U+07F3) for visual depth
    steganography.rs              -- Zero-width Unicode metadata embedding (chain height, epoch, device hash)
  types/
    mod.rs                        -- Deterministic types umbrella
    time.rs                       -- WallTime, MonoTicks, Timestamp (dual time model)
    quantized.rs                  -- QuantizedFloat (i64 mantissa, 10^-6 scale, no IEEE-754)
    evidence.rs                   -- Evidence sum type (Graph | Sensor | Hybrid, with type tag bytes)
    basis.rs                      -- BasisId, BasisRef
    session.rs                    -- SessionId, SessionContext

#### Supporting files:
- `lexicons/v1.0.json` -- Initial lexicon
- `docs/00-PROJECT_CHARTER.md` -- Project purpose
- `docs/01-GLOSSARY.md` -- Term definitions
- `docs/02-INVARIANTS_LEDGER.md` -- System invariants
- `docs/03-IMPLEMENTATION_GUIDE.md` -- Build guide
- `docs/DESIGN.md` -- Complete design document

---

2.2 The 10 N'Ko Sigils

File: `Desktop/Comp-Core/docs/nko-ecosystem/SIGIL_SEMANTICS.md` -- Full 234-line document mapping each sigil to visual representation, semantic meaning, UI affordance, motion signature, and N'Ko form.

Sigil stability: The 10 operator sigils are LOCKED. They will not change. (NIP-0001, NIP-0002)

---

2.3 The 7 Anticipation Geometry Scalars

Location (Rust): `Desktop/anticipation-geometry/crates/anticipation/src/scalars/`
Location (Python): `Desktop/anticipation-geometry/python/anticipation_geometry/`
Location (Package): `Desktop/Comp-Core/packages/anticipation-geometry/generalized_anticipation.py`
Location (Crate): `Desktop/Comp-Core/core/motion/cc-anticipation/`
Paper (standalone): `Desktop/anticipation-geometry/paper/paper.md`
Paper (Comp-Core copy): `Desktop/Comp-Core/papers/anticipation-geometry/paper.md`

Key finding: Transition pressure sign predicts conversation convergence at 71.8

---

2.4 How the 10 Sigils Map to the 7 Scalars

The sigils and scalars operate at different abstraction levels but share the same z-trajectory input. The 7 scalars are computed continuously from the trajectory. The 10 sigils are discrete claims fired when scalar thresholds are crossed.

                    z-trajectory (continuous stream)
                           |
              +------------+------------+
              |                         |
    Anticipation Kernel            Claim Detector
    (7 scalars, every frame)     (10 types, event-driven)
              |                         |
    AnticipationPacket            Claim IR
    {commitment, uncertainty,     {StabilizeClaim, ...}
     transition_pressure,                |
     recovery_margin,              Surface Renderer
     phase_stiffness,                    |
     novelty, stability}           N'Ko Line

Specific mappings:

The CognitiveTwin Bridge (`Desktop/Comp-Core/core/semantic/cc-cognitivetwin-bridge/`) performs the 5D<->7D coordinate mapping between RAG++ (5D trajectory coordinates) and Echelon (7 anticipation scalars + 3 vectors). See architecture doc: `Desktop/Comp-Core/docs/architecture/18-COGNITIVETWIN_BRIDGE.md` (818 lines).

---

2.5 EPOCH Protocol -- Stacks/Clarity Contracts

Location: `[home-path]` (Python services + SDK)
Location: `Desktop/crypto-gpu-fund/` (bots, scripts, logs)

2.5.1 Clarity Contracts (11 total, 8 deployed to testnet)

2.5.2 The nko-inscription.clar Contract

Full Clarity source in: `Desktop/psichain-full-specification.md` (lines 78-300+)

Key data structures:
- `inscriptions` map: nko-text (UTF-8 1024), inscription-hash (32 bytes), prev-hash (32 bytes), claim-type, sigil, confidence, block-height, timestamp, lexicon-version, density, basin-id, inscriber principal
- `lexicon-tokens` map: versioned vocabulary tokens with compression ratios
- `basins` map: basin lifecycle (proto, graduated, split, merged, retired)
- Chain statistics: total-inscriptions, total-tokens-minted, total-basins, avg-density, claim-counts

Public functions:
- `inscribe` -- Add N'Ko inscription to chain, emit inscription-committed event
- `mint-token` -- Register new vocabulary token (phrase compression)
- `advance-lexicon` -- Record lexicon epoch boundary
- `register-basin` / `graduate-basin` -- Basin lifecycle on-chain

2.5.3 Off-Chain Services (all Python, stdlib-only)

2.5.4 Epoch SDK

Location: `[home-path]` (`__init__.py`, `client.py`, `wallet_loader.py`)
Key feature: Implements c32check address encoding from scratch (Crockford base32, SHA256 double-hash checksums). Encodes/decodes all Clarity value types.

2.5.5 KARL-Chain Bridge

Location: `[home-path]`
- Synchronizes KARL trajectory data with on-chain karl-intelligence contract
- Computes Merkle roots of trajectory batches
- Monitors on-chain promotion events

---

2.6 PsiChain -- The Self-Referential Inscription Chain

Specification: `Desktop/psichain-full-specification.md` (comprehensive, extracted from session transcript)
Narration: `Desktop/psichain-architecture-narration.txt` (100+ lines, dense technical walkthrough)
Implementation: `Desktop/Comp-Core/core/semantic/cc-inscription/src/chain_link/` (Rust, BUILT)

Core Formula

Psi(t+1) = inscribe(z(t), Psi(0..t), L(t), B(t))

Each inscription depends on:
- Current z-trajectory input `z(t)`
- ALL previous inscriptions through hash chain `Psi(0..t)`
- Evolved lexicon `L(t)`
- Current basis `B(t)`

The system is computing its own continued existence.

Key Extensions Over cc-inscription

1. ChainLink (`chain_link/mod.rs`): Connects every inscription to its predecessor via SHA-256 hash chain. Fields: inscription_id, prev_hash, chain_height, surface_hash, epoch_ref, link_hash.

2. Combining Marks (`chain_link/combining.rs`): 9 N'Ko combining marks (U+07EB..U+07F3) encode chain depth visually. Deeper positions = darker, more ornate characters. Injected AFTER rendering but BEFORE surface hash computation.

3. Zero-Width Steganography (`chain_link/steganography.rs`): Binary alphabet using U+200B (=0) and U+200C (=1). Encodes chain_height (8 bytes), epoch_number (8 bytes), device_hash (4 bytes), checksum (2 bytes). Total: 178 invisible chars between visible N'Ko chars. Included in surface hash.

4. Computational Opacity: Data is not encrypted but compressed through accumulated context. Only the model that performed the full computation chain can reconstruct semantics.

Four Properties

1. Irreversibility -- Cannot alter without breaking all subsequent hashes
2. Density monotonicity -- Information density rho(t) = Shannon_entropy / stroke_count increases monotonically
3. Self-reference -- Each inscription depends on the full chain history
4. Selective transparency -- Inscriptions prove without revealing (semantic privacy via contextual depth)

---

2.7 N'Ko NLP Stack

Location: `Desktop/NKo/nko/` (Python package)

Tokenizer

Location: `Desktop/NKo/core/pipeline/tokenizer.py`
- Script-Aware Tokenizer (handles N'Ko, Arabic, Latin with bidi awareness)
- Token types: WORD, PARTICLE, PUNCTUATION, NUMBER, WHITESPACE, COMBINING, MIXED
- Combining mark metadata: tone marks, vowel marks, nasalization, gemination
- Recognizes N'Ko particles (postpositions like BE, TE, KA)

BPE Tokenizer (referenced in PsiChain spec): 614 vocab standard, 206 morpheme-constrained. Custom BPE trained on N'Ko text corpus. The 4-state FSM validates any N'Ko string: consonant -> vowel -> (optional nasal coda) -> (optional tone mark). This property makes N'Ko ideal for structured data encoding.

---

2.8 Cognitive Twin

Papers:
- `Desktop/cognitive-twin-research-paper.md` -- 120+ page research paper (RPS framework, 379K turns corpus, 6-layer Living Executor)
- `Desktop/cognitive-twin-research-paper.tex` -- LaTeX version (compiled to PDF)
- `Desktop/cognitive-twin-theorems.tex` -- Formal proofs (Banach contraction, coherence energy, identity drift bounds, autonomy ratchet)
- `Desktop/cognitive-twin-architecture.md` -- Practical architecture (LoRA + Graph Kernel + Anticipation Geometry)
- `Desktop/cognitive-twin-summary-for-mo.txt` -- Plain English summary

Implementation:
- `Desktop/Comp-Core/packages/cognitive-twin/` -- Full Python package
- `living_twin.py` -- Zero-training cognitive twin (7-layer prompt assembly: identity, few-shot, corrections, RAG, graph, RLM, real-time state)
- `twin_server_v2.py` through `twin_server_v6.py` -- 5 versions of inference server
- `twin_rag_server.py` -- RAG-augmented server
- `cognitive_twin/` -- Package module
- `identity/MO_IDENTITY.md` -- Static personality document
- `scripts/eval_twin.py` -- Evaluation harness
- `notebooks/train_twin.py` -- Training notebook
- `tests/test_living_twin.py` -- Tests

CognitiveTwin Bridge:
- `Desktop/Comp-Core/core/semantic/cc-cognitivetwin-bridge/` -- Rust/Python FFI bridge
- `Desktop/Comp-Core/docs/architecture/18-COGNITIVETWIN_BRIDGE.md` -- 818-line deep architecture doc

Key Architecture (3-component separation):
1. Personality = LoRA adapter on Qwen2.5-3B (all 36 layers, 2,923 training examples, 2.5:1 correction-to-affirmation ratio)
2. Knowledge = Live Graph Kernel (71,130+ triples, queried at inference time)
3. Trajectory = 7 Anticipation Geometry scalars (injected as prompt context, zero training)

5 Cognitive Modalities (RPS framework):
- V_L: Linguistic (voice, tone, vocabulary distribution)
- V_D: Decision (approval/rejection patterns, priority rankings)
- V_K: Knowledge (domain expertise across 11 KARL domains)
- V_V: Values (ethical boundaries, aesthetic preferences, quality thresholds)
- V_T: Temporal (work rhythms, urgency patterns, circadian phase)

Coherence Energy:

Phi(x; A, T) = (1/2) * sum_n || x_n - sum_m a_nm * T_{n<-m}(x_m) ||^2

Convergence via Banach contraction when cross-cognitive translators satisfy spectral norm bound sigma_max < 1.

---

2.9 NIP Documents (N'Ko Improvement Proposals)

Location: `Desktop/Comp-Core/docs/NIP/` (12 files)

---

2.10 Supporting Architecture Documents

---

2.11 Narratives and Blog Posts

---

3. Connection Map

                        +-----------+
                        | KARL      |
                        | Trajectory|
                        | Intel     |
                        +-----+-----+
                              |
                  trains LoRA | Merkle roots
                              |
    +------+          +-------v--------+           +----------+
    | NKo  |          | Cognitive Twin |           | EPOCH    |
    | NLP  |<-------->| (personality + |---------->| Protocol |
    | Stack|  sigils  |  knowledge +   | revenue   | (Stacks) |
    +--+---+  render  |  trajectory)   | funds GPU +----+-----+
       |              +-------+--------+                |
       |                      |                         |
       | syllable     injects | trajectory         nko-inscription
       | codebook     context | scalars            contract stores
       | tokenizer            |                    N'Ko on-chain
       |              +-------v--------+                |
       |              | Anticipation   |                |
       +------------->| Geometry       |                |
       |  phonetics   | (7 scalars)    |                |
       |              +-------+--------+                |
       |                      |                         |
       |              computed| from                    |
       |              z-traj  |                         |
       |              +-------v--------+                |
       +------------->| cc-inscription |<---------------+
          N'Ko        | (Rust, 22.8K   |  PsiChain
          surface     |  lines)        |  settles hashes
          rendering   +----------------+  on Stacks

Key Dependency Flows

1. cc-inscription depends on:
- `cc-anticipation` (7 scalars trigger claim detection)
- Graph Kernel (slice boundary enforcement, admissibility tokens)
- RAG++ (evidence retrieval for Echo claims)
- DELL (z-trajectory source)

2. PsiChain extends cc-inscription with:
- `chain_link` module (hash chain, combining marks, steganography)
- Stacks settlement via `nko-inscription.clar`

3. EPOCH Protocol provides:
- On-chain infrastructure (11 contracts, chainhooks)
- Revenue engine (DEX fees, arb, compute, staking)
- `karl-intelligence` contract for trajectory Merkle roots
- Bitcoin-clocked epoch rotation

4. Cognitive Twin consumes:
- Inscription corpus (training data, context)
- Graph Kernel (live knowledge at inference)
- Anticipation Geometry (7 scalars as prompt conditioning)
- KARL (skill routing, trajectory evaluation)

5. N'Ko NLP Stack provides:
- Phonetics (IPA mappings, char classification)
- Syllable codebook (3,024 entries for ASR retrieval target)
- Transliteration (N'Ko <-> Latin <-> Arabic)
- Tokenizer (script-aware, bidi, combining mark metadata)
- Sound Sigils (audio synthesis for the 10 sigils)

---

4. What's Built vs What's Documented

### BUILT AND RUNNING
- [x] cc-inscription Rust crate (22,881 lines, 48 files, compiles)
- [x] All 10 claim detectors with typed IR
- [x] Surface renderer (claim -> N'Ko line)
- [x] Canonical serialization (CBOR, SHA-256, NFC)
- [x] Provenance system (InscriptionId, ProvenanceWitness, PureVerificationContext)
- [x] PsiChain chain_link module (ChainLink, combining marks, steganography)
- [x] Ontology operations pipeline (split, merge, graduate, retire)
- [x] Basin lifecycle (proto -> graduated -> split/merge/retire)
- [x] Lexicon versioning with epoch boundaries
- [x] Phrase detection and compression
- [x] Anticipation Geometry (Rust crate + Python bindings, evaluated on 3 domains)
- [x] cc-anticipation kernel (7 scalars, <2ms latency, 50 tests)
- [x] CognitiveTwin Bridge (Rust FFI, 5D<->7D, style signature)
- [x] N'Ko phonetics, syllable codebook, transliteration, tokenizer
- [x] Sound Sigils (audio synthesis for all 10 sigils)
- [x] Living Twin (zero-training, 7-layer prompt assembly)
- [x] Twin servers v2-v6 (inference endpoints)
- [x] EPOCH contracts 1-8 deployed to Stacks testnet
- [x] EPOCH off-chain services (all Python, stdlib-only, Docker)
- [x] Epoch SDK (Clarity value codec, c32check from scratch)
- [x] KARL-chain bridge
- [x] cc-mcs-headless daemon (cloud-deployed, receives phone sensor data, runs inscription detection at 6Hz)
- [x] 10 NIP specifications

### DOCUMENTED / SPECIFIED BUT NOT FULLY WIRED
- [ ] nko-inscription.clar contract (full Clarity source written but not yet deployed)
- [ ] karl-intelligence.clar (specified, not deployed)
- [ ] Chainhook events firing to inscription pipeline (7 hooks specified)
- [ ] End-to-end: cc-inscription output -> nko-inscription.clar settlement
- [ ] Cognitive Twin consuming inscription corpus as training data
- [ ] V2 revenue split (50
- [ ] Genesis inscription (47K records, 635MB corpus bootstrap)
- [ ] Live Stacks mainnet deployment (testnet proven, mainnet pending legal review)
- [ ] Graph-native version control (GVC schema in Supabase, 133 records seeded)
- [ ] Anticipatory Transformer (recovered crate, needs integration)

---

5. The Full Round-Trip Pipeline

English -> Inscription

1. Human moves body / operates session / agent generates trajectory
2. Sensors capture motion OR conversation embeddings stream
3. z-trajectory computed (63D sensor OR 768D embedding)
4. cc-anticipation computes 7 scalars continuously
5. ClaimDetector fires when thresholds crossed
6. Typed Claim IR produced (e.g. StabilizeClaim)
7. Lexicon resolves basin/place tokens (versioned)
8. SurfaceRenderer produces N'Ko line
9. Canonical serialization (CBOR, NFC, QuantizedFloat)
10. ProvenanceWitness computed (InscriptionId = SHA-256 of all)
11. ChainLink created (prev_hash chain)
12. Combining marks injected (visual depth)
13. Zero-width steganography embedded (chain height, epoch, device)
14. Surface hash computed (includes marks + steganography)
15. Inscription submitted to nko-inscription.clar on Stacks
16. Bitcoin anchoring via Proof of Transfer

Inscription -> Twin -> Response

1. Query arrives (user question or agent pause)
2. Graph Kernel provides knowledge slice (live triples)
3. RAG++ retrieves relevant memories (pgvector, 768-dim embeddings)
4. Inscription corpus queried for relevant historical patterns
5. Anticipation Geometry computes current conversation trajectory scalars
6. CognitiveTwin Bridge maps 7D scalars to 5D RAG coordinates
7. Style signature retrieved (256-dim EMA of trajectory segments)
8. Living Twin assembles system prompt:
   - Identity document (static personality)
   - Few-shot examples (style calibration)
   - Accumulated corrections (feedback loop)
   - RAG context (relevant memories)
   - Graph context (relational knowledge)
   - RLM decomposition (multi-hop reasoning)
   - Real-time state (current trajectory scalars)
9. LLM generates response (personality from LoRA weights)
10. Response assessed by Autonomy Ratchet quality function Q(a)
11. Auto-pass if Q >= 0.85, human review if 0.60-0.84, auto-reject if < 0.60
12. Response becomes new trajectory data, feeding back to step 1

---

6. Key Invariants

From NIP-0001 (constitutional):
1. Provenance Law: For any InscriptionId, given archived evidence + lexicon + basis + config, the claim IR and N'Ko surface are deterministically reproducible and the InscriptionId recomputable.
2. Non-retroactive corpus: Old inscriptions are never rewritten. Reinterpretation is a derived view.
3. Operator sigils are LOCKED: The 10 sigils will not change.
4. No IEEE-754: All serialized values use QuantizedFloat (i64 mantissa, 10^-6 scale).
5. NFC-safe: All N'Ko characters confirmed NFC-normalized.
6. Chain integrity: Breaking any link invalidates every subsequent inscription.

From Anticipation Geometry (INV-001 through INV-007):
- Deterministic replay (no random seeds)
- Coverage threshold (reject < 0.9)
- All scalars validated and bounded
- No hot-path allocation
- Schema version enforced

From PsiChain:
- Density monotonicity: information density increases monotonically with chain height
- Self-reference: each inscription depends on the full chain history
- Combining marks are part of the cryptographic commitment
- Zero-width metadata is part of the surface hash

---

Generated 2026-03-23 by Meta-Recursive Explorer from codebase analysis across 7 major directories and 100+ source files.