diff --git a/conductor/tracks/video_analysis_deob_warmup_20260621/prompt_template.md b/conductor/tracks/video_analysis_deob_warmup_20260621/prompt_template.md index a42390fc..682c7b3b 100644 --- a/conductor/tracks/video_analysis_deob_warmup_20260621/prompt_template.md +++ b/conductor/tracks/video_analysis_deob_warmup_20260621/prompt_template.md @@ -11,10 +11,10 @@ You are a de-obfuscator. Your task: take a Pass 1 report (full of standard math Your operational stance: - **Library Specification > Philosophy** (per Cluster 0, Pattern 9): prefer executable, debuggable, deterministic specifications over intuition pumps. - **Decompression > Compression** (per Cluster 0, P1): the first step for any math is to decompress it. -- **Construct, not Invent** (per Cluster 0, Pattern 3): use the user's pseudo-code DSL, not free-form prose. - **Bounded form required** (per §1.1 of `report.md`): no `∞_val`; use `Stream A = nat -> A` for processes. - **Form anchor required** (per §5): every re-encoding has a form anchor — "what bounded form does this project from the indefinite?" - **Honest epistemic hedging** (per §1.10): if uncertain, flag it; do not guess. +- **Scheme-canonical vs. user-specific:** the de-obfuscation scheme's principled re-encodings come from the 5 rules (boundedness, form-anchor, etymology, lossless, encoding-explicit). Some terms ALSO have user-specific preferred forms (Sectored Language V1, GA reinterpretations, classical Greek/Latin); these are *optional* output conventions documented in `report.md` §3.4 + §3.5, not scheme-canonical. Apply the user-specific forms only when the reader would prefer them. ## Input @@ -167,9 +167,11 @@ Reference: `report.md` §3 for the full lexicon (~70 terms after Phase 1 expansi - **Tier 3 (Type-theoretic primitives, 18 terms):** `Type` → `kind`; `Type of types` → `Kind`; `Constructor` → `intro`; `Eliminator` → `elim`; `Computation rule` (value-level) → `comp`; `Type-level Computation` → `getType(...) === T`; `Pair` with `Build/Build`; `Dependent(B)`; `lambda.x.M`; `objects : m : A, n : B ;`; etc. - **Tier 4 (AI-fuzzing tolerance, 21 terms):** "invent" → `construct`; "real number" → `encodable quantity`; "imaginary number" → `bivector`; "dot product" → `length-projection product` (or `'scalar product'`); "cross product" → `wedge product`; "anti-wedge" → `regressive product` / `contraction` / `interior product`; "negative" → `F²` operator; "infinity" → **BANNED**; "point" → `Punctum` / `σημεῖον`; "kernel" (cross-domain) → `discrete subsystem that holds a continuous process up`; "Bourbaki" / "Standard GA" → **FOIL**; etc. -## The Sectored Language Operator Names (per `report.md` §3.5, from Cluster 9) +## The Sectored Language Operator Names (per `report.md` §3.5, from Cluster 9) — OPTIONAL -For linear algebra and CAS, use the Sectored Language naming: +> **Reading guide.** This is the **user's preferred output convention** for linear-algebra and CAS operations (per Cluster 9, the FGED V1 .sectr files). It is OPTIONAL — the de-obfuscation scheme does not require it. The scheme's principled re-encodings (e.g., `scalar product`, `magnitude`, `normalize`) are what the LLM produces; the Sectored Language names are one of several ways the reader can express those principled re-encodings. Apply the Sectored Language names when (a) the user requests it, (b) the term appears in a context where the user's prior de-obfuscation work used Sectored Language, or (c) the reader's preference is to use Sectored Language output. Otherwise, use conventional math with explicit type annotations. + +For linear algebra and CAS, the user's preferred Sectored Language naming is: - `magnitude(v)` for `||v||` - `normalize(v) -> UnitVector` for unit vector - `transpose(M) -> Matrix` for matrix transpose @@ -203,7 +205,7 @@ After producing the 3 files, verify each: - [ ] **Etymology-cited** — every new term has the 1-line origin + 1-line definition history. - [ ] **Form-anchored** — every re-encoding has a form anchor. - [ ] **Noise-deduped** — the 6 noise-dedup maps applied where applicable. -- [ ] **Sectored-language-named** — linear algebra and CAS use the Sectored Language names (per §3.5). +- [ ] **User-specific conventions applied only when appropriate** — Sectored Language names + classical Greek/Latin/Sanskrit forms + GA reinterpretations are USER preferences, not scheme-canonical. Apply them only when the reader would prefer them (per §3.4 reading guide + §3.5 note). - [ ] **EPP-formatted** — the fully-expanded pseudo-code follows the EPP format (per Cluster 1, Pattern 5). - [ ] **Univalence-flagged** — when the LLM encounters the univalence axiom, flag it as "compression (lossless?**)" so the user can opt in or out. diff --git a/conductor/tracks/video_analysis_deob_warmup_20260621/report.md b/conductor/tracks/video_analysis_deob_warmup_20260621/report.md index 2ea56e01..3f83a0fd 100644 --- a/conductor/tracks/video_analysis_deob_warmup_20260621/report.md +++ b/conductor/tracks/video_analysis_deob_warmup_20260621/report.md @@ -277,6 +277,8 @@ The lexicon is the heart of the de-obfuscation. It is organized in 4 tiers. The ### §3.4 Tier 4: AI-fuzzing tolerance terms (22 terms, expanded in Phase 1) +> **Reading guide.** This tier mixes **principled re-encodings** (derived from the 5 load-bearing rules) with **user-specific re-encodings** (the user's personal preferences, drawn from Cluster 0/1/6/7/8/9 — including the Sectored Language V1, the GA reinterpretations, and the classical Greek/Latin/Sanskrit forms). The principled entries are the scheme's canonical forms; the user-specific entries are *one possible output convention* the scheme can produce. The 4-layer output format (§6.2) is similarly optional. For each entry below, the "Re-encoded" column is the form the *scheme* produces; entries that ALSO accept the user's preferred output are marked **[user-also-accepted]**. Phase 1 (lexicon child) will formalize this distinction. + | # | Conventional (fuzzy) | Re-encoded (precise) | Source cluster | |---|---|---|---| | 4.1 | "invent" | `construct` | Cluster 0 | @@ -306,7 +308,7 @@ The lexicon is the heart of the de-obfuscation. It is organized in 4 tiers. The ### §3.5 Sectored Language operator terms (Phase 1, from Cluster 6 + Cluster 9) -These are the user's preferred operator names from the Sectored Language (per Cluster 9, the FGED V1 .sectr files): +> **Reading guide.** This section documents the **user's preferred output convention** for linear-algebra and CAS operations — the Sectored Language V1 (FGED V1) naming. The de-obfuscation scheme does NOT require this convention. It is one example of how the scheme's principled re-encodings (e.g., `scalar product`, `magnitude`) can be realized in an executable form. Other readers may use different conventions (Standard GA, conventional math with explicit type annotations, etc.); the scheme's output is the re-encoded form, not the Sectored Language names. The 7 example transformations in §7 demonstrate how the scheme produces these specific names. **Phase 1 (lexicon child) will move this table to Appendix B ("User's preferred output conventions, optional") to make the principled/user-specific distinction explicit.** | Conventional | Sectored Language name | Source | |---|---|---| @@ -360,6 +362,8 @@ All 6 are documented below. Sources: Cluster 0, 2, 3, 4, 7, 8 (see report §4 of ### §4.4 Map 4: "Real" = "Imaginary" = "Bivector" (geometric algebra) +> **Reading guide.** This map is the **user's geometric-algebra reinterpretation** (per Cluster 0, Pattern 2; Cluster 8 GeoAlg). It is the user's preference, not a scheme-canonical dedup. The scheme's principled form for "imaginary number" is `bivector` (or `Bivector` with explicit grade annotation); "real number" re-encodes to `quantity() : ` (per Rule 5). The collapse of "real" + "imaginary" into the GA framework is the user's contribution; the scheme's job is to produce the principled re-encoding, and the user (or the LLM at their direction) can additionally adopt the GA collapse if they want. + **Source cluster:** Cluster 0 (Pattern 2 — Descartes-rejection / Clifford-affirmation); Cluster 8 (GeoAlg — `Point`, `Circle`, `Line`, `Plane` types). ### §4.5 Map 5: "Invent" = "Create" = "Imagine" → "Construct" @@ -423,7 +427,9 @@ The etymology rule is the second operational requirement, derived from §1.5. If the term is a user coinage, the etymology is the user's reason for the coinage, and the definition history is "user-specific; see `samples/`." -### §6.2 The 4-layer output format (per Cluster 7) +### §6.2 The 4-layer output format (per Cluster 7) — OPTIONAL + +> **Reading guide.** This 4-layer format is the **user's preferred output convention** for terms with rich etymological trails (per Cluster 7, `Notiones.txt`). It is NOT a scheme-canonical requirement. The de-obfuscation scheme's baseline output is the 3-layer format (§5.2): translation (side-by-side) + replacement (re-encoded) + decoder (per-term). The 4-layer form adds an optional etymological layer (Greek + Latin + English + Sanskrit) that the LLM can apply **if the user requests it** or **if the term has a rich etymological trail**. Readers who don't care about Greek/Latin/Sanskrit get the 3-layer output; readers who do get the 4-layer output. The deob-warmup's `prompt_template.md` should make this an OPTION, not a REQUIREMENT. For terms with rich etymological trails, the output should be **trilingual or 4-language** (per Cluster 7, Pattern 6): 1. **Original** (e.g., the Greek or Latin).