conductor(deob_pass3): Initialize Pass 3 track scaffold + TIER2_STARTER.md
Pass 3 is the FINAL phase of the 3-pass research campaign: project the v2-deobfuscated outputs to C11 or Python code that conveys the subject video's content. Track scaffold: - spec.md: 14 sections, 11 videos, per-language default, 4 + 3 verification criteria - plan.md: 3 phases, 11 tasks, Tier 2 + 4 Tier 3 sub-agents - metadata.json: scope, per-language default, hardware target (up to ), risk register - state.toml: 3 phases, 11 tasks, verification flags - README.md: track index TIER2_STARTER.md (the dispatch prompt for Tier 2): - 15 sections, self-contained - The 4 PRIMARY inputs to read in order (v2 lexicon, C11 convention, Pass 1/2 content, manual_slop) - The 11 videos with per-language default (10 C11 + 2 Python + 1 synthesis) - The per-video deliverables (C11 .c/.h + 3 docs; Python .py + 3 docs) - The 4 + 3 verification criteria - The commit discipline (per-file atomic) - The 6 open questions answered - The 7 risks - The 4 Tier 3 sub-agent prompts (per cluster) Per-language default: C11 for math/algorithms oriented; Python for probability/information-theoretic; synthesis in Python. Tier 2 may override per video.
This commit is contained in:
@@ -0,0 +1,109 @@
|
||||
# Pass 3 — C11/Python Projection (Track Index)
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623`
|
||||
**Type:** Research-to-code (the FINAL phase of the 3-pass research campaign)
|
||||
**Status:** spec DRAFT (pending user review)
|
||||
**Date:** 2026-06-23
|
||||
|
||||
## What this track is
|
||||
|
||||
Pass 3 projects the v2-deobfuscated outputs to C11 or Python code that conveys the subject video's content. The code may or may not run (per user 2026-06-23); the goal is to express the concepts in code using the user's idiomatic style.
|
||||
|
||||
This is the **final phase** of the 3-pass research campaign:
|
||||
- Pass 1: information extraction (12 deep-dive reports + 1 synthesis)
|
||||
- Pass 2: deobfuscation (33 markdown deliverables using v1 lexicon)
|
||||
- v2 patch: corrective refinements (8 corrections + 3 refinements + 4 template notations + 2 `<<` / `>>` placements)
|
||||
- **Pass 3: C11/Python projection (this track)**
|
||||
|
||||
## Track files
|
||||
|
||||
| File | Purpose |
|
||||
|---|---|
|
||||
| `spec.md` | The design intent: 14 sections, the goals, the architecture reference, the risk register, the verification criteria |
|
||||
| `plan.md` | The TDD-ready plan: 3 phases, 11 tasks, Tier 2 + 4 Tier 3 sub-agents |
|
||||
| `metadata.json` | The machine-readable metadata: scope, per-language default, hardware target, risk register, user-directives-logged |
|
||||
| `state.toml` | The track state: 3 phases, 11 tasks, verification flags |
|
||||
| `README.md` | This file (the track index) |
|
||||
| `TIER2_STARTER.md` | **The dispatch prompt for Tier 2** — the primary input for execution. 15 sections, self-contained, includes the 4 PRIMARY inputs to read, the 11 videos (per-language default), the per-video deliverables, the 4 + 3 verification criteria, the commit discipline, the 6 open questions answered, the risks, and the Tier 3 sub-agent prompts. |
|
||||
| `PASS3_REPORT.md` | (future) The end-of-track report after execution |
|
||||
| `artifacts/<slug>/` | (future) The per-video deliverables: C11 (.c + .h) or Python (.py) + 3-4 markdown docs |
|
||||
|
||||
## The 11 videos (per-language default)
|
||||
|
||||
| # | Slug | Language | Cluster |
|
||||
|---|---|---|---|
|
||||
| 1 | `cs229_building_llms` | C11 | A |
|
||||
| 2 | `probability_logic` | Python | A |
|
||||
| 3 | `entropy_epiplexity` | Python | A |
|
||||
| 4 | `score_dynamics_giorgini` | C11 | A |
|
||||
| 5 | `platonic_intelligence_kumar` | C11 | B |
|
||||
| 6 | `free_lunches_levin` | C11 | B |
|
||||
| 7 | `generic_systems_fields` | C11 | C |
|
||||
| 8 | `brain_counterintuitive` | C11 | C |
|
||||
| 9 | `neural_dynamics_miller` | C11 | C |
|
||||
| 10 | `multiscale_hoffman` | C11 | C |
|
||||
| 11 | `synthesis` | Python | D + cross-cutting |
|
||||
|
||||
**Total:** 10 C11 + 3 Python (synthesis is Python). Tier 2 may override per video.
|
||||
|
||||
## The 4 PRIMARY inputs (Tier 2 must read first)
|
||||
|
||||
1. **The v2 lexicon:** `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` (~1000 LOC, 13 sections + 4 appendices)
|
||||
2. **The C11 convention:** `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md` (~600 LOC, 15 sections)
|
||||
3. **The Pass 1 reports + Pass 2 deliverables:** the source content (11 videos + 1 synthesis)
|
||||
4. **The Python convention (manual_slop):** `conductor/product-guidelines.md` + `conductor/tech-stack.md`
|
||||
|
||||
## The per-video deliverables (per `TIER2_STARTER.md` §6)
|
||||
|
||||
For each video, the deliverables are:
|
||||
|
||||
- **For C11 videos (9):** `<slug>.c` + `<slug>.h` (if needed) + `<slug>_translation.md` + `<slug>_decoder.md` + `<slug>_notes.md`
|
||||
- **For Python videos (2 + 1 synthesis):** `<slug>.py` + `<slug>_translation.md` + `<slug>_decoder.md` + `<slug>_notes.md`
|
||||
|
||||
Plus the global deliverables:
|
||||
- `PASS3_REPORT.md` (the end-of-track report)
|
||||
- `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md` (canonical docs)
|
||||
|
||||
## The 4 + 3 verification criteria (per v2 lexicon)
|
||||
|
||||
| Criterion | Status |
|
||||
|---|---|
|
||||
| Lossless | required |
|
||||
| Bounded | required |
|
||||
| Constructively typed | required |
|
||||
| Etymology-cited | required |
|
||||
| Encoding-explicit (Rule 5) | required |
|
||||
| Form-anchored | required |
|
||||
| User-specific opt-in | required |
|
||||
|
||||
## The 6 open questions for Pass 3 (answered)
|
||||
|
||||
| # | Question | Answer |
|
||||
|---|---|---|
|
||||
| 1 | Applied domain | Simple C11 or Python program per video |
|
||||
| 2 | User-specific forms | Annotation if not code; pseudo sectr lang adapts per video |
|
||||
| 3 | Indefinites | `Scalar` / `float` / `integer` placeholder; `float64` only when target resolution matters |
|
||||
| 4 | `<<` / `>>` rendering | C11: `much_less` / `much_greater` / `weakly_coupled`; Python: same |
|
||||
| 5 | Criteria | OK; may ideate for applied domain |
|
||||
| 6 | User-facing artifact | Code files (may or may not run) + markdown docs |
|
||||
|
||||
## Sequence
|
||||
|
||||
1. **Phase 1** (Tasks 1.1-1.4): Tier 2 reads the 4 PRIMARY inputs.
|
||||
2. **Phase 2** (Tasks 2.1-2.4): Tier 2 dispatches 4 parallel Tier 3 sub-agents (per cluster). Each sub-agent produces the per-video deliverables for its cluster.
|
||||
3. **Phase 3** (Tasks 3.1-3.3): Tier 2 writes `PASS3_REPORT.md` + end-of-track report + updates `state.toml`.
|
||||
|
||||
**Total: ~35-58 atomic commits.**
|
||||
|
||||
## Cross-references
|
||||
|
||||
- **Upstream:** `video_analysis_deob_apply_20260621/` (Pass 2, SHIPPED) + `video_analysis_deob_lexicon_v2_20260623/` (v2 patch, SHIPPED) + `video_analysis_deob_c11_reference_20260623/` (C11 reference, SHIPPED)
|
||||
- **The v2 lexicon:** `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md`
|
||||
- **The C11 convention:** `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md`
|
||||
- **The Pass 1 reports:** `conductor/tracks/video_analysis_<slug>_20260621/report.md`
|
||||
- **The Pass 2 deliverables:** `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/<slug>/` + `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/<slug>/`
|
||||
- **Tier 2 dispatch prompt:** `TIER2_STARTER.md` (this folder)
|
||||
|
||||
---
|
||||
|
||||
*End of `README.md`. 5 sections. Pass 3 closes the 3-pass research campaign.*
|
||||
@@ -0,0 +1,364 @@
|
||||
# TIER2_STARTER.md — Pass 3 (C11/Python Projection)
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623` (Pass 3 of 3 — the projection to C11/Python code)
|
||||
**Date:** 2026-06-23
|
||||
**Type:** Research-to-code (the final phase of the 3-pass research campaign)
|
||||
**Status:** DRAFT (pending user review)
|
||||
|
||||
> **Purpose.** This document is the dispatch prompt for Tier 2. Read this file fully, then read the 4 PRIMARY inputs in the order listed in §6, then execute the 11 videos per the language defaults in §3. Produce C11 or Python code per video that conveys the subject video's content. Use the conventions documented in the c11_convention.md (C11) and the manual_slop convention (Python).
|
||||
>
|
||||
> **Execution model:** direct Tier 2 synthesis + 4 parallel Tier 3 sub-agents (one per cluster). The cross-cluster integration requires holding the full lexicon + c11_convention.md in mind; Tier 3 sub-agents would lose context on the principled vs user-specific distinction and the per-language `<<` / `>>` rendering.
|
||||
|
||||
---
|
||||
|
||||
## §1. Dispatch
|
||||
|
||||
```bash
|
||||
/tier-2-auto-execute video_analysis_deob_pass3_20260623
|
||||
```
|
||||
|
||||
The Tier 2 sandbox will:
|
||||
1. Read this file (`TIER2_STARTER.md`)
|
||||
2. Read the 4 PRIMARY inputs in the order in §6
|
||||
3. Execute the 11 videos per the language defaults in §3
|
||||
4. Produce the deliverables per video per §7
|
||||
5. Commit per-file atomically per §8
|
||||
6. Verify per §9
|
||||
|
||||
---
|
||||
|
||||
## §2. Critical context (READ THIS FIRST)
|
||||
|
||||
1. **Pass 3 is the FINAL phase of the 3-pass research campaign.** Pass 1 produced the 12 deep-dive reports + 1 synthesis. Pass 2 (apply) produced 33 deobfuscated markdown deliverables using the v1 lexicon. The v2 corrective patch refined the lexicon. **Pass 3 projects the deobfuscated content to C11 or Python code that conveys the subject video's content.** The code may or may not run (per user 2026-06-23); the goal is to express the concepts in code.
|
||||
|
||||
2. **The applied domain is simple programs that convey what each Pass 1 video provides.** Per user 2026-06-23: "The applied domain is making a simple program in C11 or python that conveys what the subject video provides. If in c11 it should follow a specific covention of c11 that raddbg uses or that I use with my duffle libs in pikuma ps1 or the forth bootslop projects. For python it should use manual slop's convention."
|
||||
|
||||
3. **Hardware target: up to $10k workstation** (per user 2026-06-23: "target up to 10k"). Default: Ryzen 9 / i9, RTX 4090, 128GB DDR5, 4TB NVMe. If a video is constrained (e.g., multiscale_hoffman with GPGPU shaders, or creikey_dl_cv with game rendering), the per-video spec can note the constraint.
|
||||
|
||||
4. **Per-language default for `<<` / `>>` rendering (per v2 lexicon):**
|
||||
- C11: `much_less(a, b, tolerance)` / `much_greater(a, b, tolerance)` / `weakly_coupled(a, b, tolerance)` (per user 2026-06-23: "weakly_coupled(...) is good for c11, much_less and much_greater can be used as well")
|
||||
- Python: same as C11
|
||||
- Forth: named words (avoids bit-shift collision; not used in Pass 3 unless explicitly requested)
|
||||
- The principled form (`<<` / `>>` with `tolerance`) is reserved for the abstract mathematical context
|
||||
|
||||
5. **Encoding placeholder scheme (per v2 lexicon):** `float` (general), `integer` (general), `Scalar` (linear/geo/tensor alg), `float64` (resolved). Default is placeholder, not committed resolution.
|
||||
|
||||
6. **Lossless preservation directive:** the v2 lexicon's 4 + 3 verification criteria (Lossless, Bounded, Constructively typed, Etymology-cited, Encoding-explicit, Form-anchored, User-specific opt-in) apply to Pass 3. The code must convey every concept from the Pass 1 / Pass 2 output; no concept is dropped.
|
||||
|
||||
7. **Tier 2 should not re-survey the 158 user samples** (per Pass 2's hard constraint). The lexicon v2 + c11_convention.md is the canonical substrate. If a convention is missing, document it inline and add to the per-track decisions log.
|
||||
|
||||
---
|
||||
|
||||
## §3. The 11 videos (per-language default)
|
||||
|
||||
Per user 2026-06-23, the per-language default:
|
||||
|
||||
| # | Cluster | Slug | YouTube ID | Language | Reason |
|
||||
|---|---|---|---|---|---|
|
||||
| 1 | A | `cs229_building_llms` | 9vM4p9NN0Ts | **C11** | Math/algorithms oriented (ML/LLM coverage) |
|
||||
| 2 | A | `probability_logic` | 0yF9TvMeAzM | **Python** | Probabilistic; NumPy/SciPy shine |
|
||||
| 3 | A | `entropy_epiplexity` | _U8AwUq_aJQ | **Python** | Information-theoretic; NumPy shine |
|
||||
| 4 | A | `score_dynamics_giorgini` | P75iVMmbqQk | **C11** | Math foundations (score-based dynamics) |
|
||||
| 5 | B | `platonic_intelligence_kumar` | 1mXUFweWOug | **C11** | Platonic / geometric AI representations |
|
||||
| 6 | B | `free_lunches_levin` | K8BmMU1Tm-I | **C11** | Levin search (math/algorithms) |
|
||||
| 7 | C | `generic_systems_fields` | QeMajYvhEbI | **C11** | Biological/cognitive systems |
|
||||
| 8 | C | `brain_counterintuitive` | cDxtFtoQVNc | **C11** | Counterintuitive brain dynamics |
|
||||
| 9 | C | `neural_dynamics_miller` | 0BS-BzEFTXA | **C11** | Neural dynamics (Miller) |
|
||||
| 10 | C | `multiscale_hoffman` | YnfaT5APPB0 | **C11** | Multiscale consciousness (Hoffman) |
|
||||
| 11 | E + synthesis | `synthesis` (cross-cutting) | — | **Python** | Cross-cluster insights; readability matters |
|
||||
|
||||
**Per-video language override:** Tier 2 may override the default per video if the content strongly suggests a different language (e.g., a math-heavy video benefits from C11 even if marked Python). Document the override in the per-video `<slug>_notes.md` file.
|
||||
|
||||
**The 2 videos already in the pilot (`cs229_building_llms` + `entropy_epiplexity`)** are also in Pass 3. The Pass 2 deliverables (`video_analysis_deob_pilot_20260621/artifacts/<slug>/`) are inputs but are not the final form; Pass 3 produces code, not markdown.
|
||||
|
||||
---
|
||||
|
||||
## §4. The 4 PRIMARY inputs (read in this order)
|
||||
|
||||
Tier 2 must read these 4 files fully before producing any code. The order matters — earlier inputs establish the substrate; later inputs reference the substrate.
|
||||
|
||||
### 4.1 The v2 lexicon (the canonical substrate)
|
||||
|
||||
**File:** `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` (~1000 LOC, 13 sections + 4 appendices)
|
||||
|
||||
**Why read first:** the v2 lexicon is the codified operational spec. It defines the principled vs user-specific distinction, the 5 rules (Boundedness, Form-anchor, Etymology, Lossless, Encoding-explicit), the 4 tiers (76 terms), the 6 noise-dedup maps, the 7 example transformations, and the per-language `<<` / `>>` rendering.
|
||||
|
||||
**Key sections to focus on:**
|
||||
- §0.3 (the principled vs user-specific distinction — load-bearing)
|
||||
- §1.5 (Rule 5: Encoding-explicit with placeholder scheme)
|
||||
- §2 (the 4 tiers; 76 terms)
|
||||
- §3 (the 6 noise-dedup maps; Maps 1, 2, 3 reshaped in v2)
|
||||
- §7 (encoding-explicit form with placeholder vs resolved distinction)
|
||||
- §9 (per-language rendering; the canonical specification)
|
||||
|
||||
### 4.2 The C11 convention (the C11 style guide)
|
||||
|
||||
**File:** `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md` (~600 LOC, 15 sections)
|
||||
|
||||
**Why read second:** the C11 convention is the user's idiomatic C11 style guide. It synthesizes the Pikuma duffle + forth bootslop + raddbg/src/base conventions. For the 9 C11 videos, this is the primary style reference.
|
||||
|
||||
**Key sections to focus on:**
|
||||
- §2 (Naming conventions; type modifiers R_/V_/etc.)
|
||||
- §3 (Type system; TSet_, Struct_, Union_, etc.)
|
||||
- §4 (Memory ordering; ooo_drift_/anchor_/drain_/weld_)
|
||||
- §7 (Macro style; INTELLISENSE_DIRECTIVES, pragma region)
|
||||
- §8 (Slice/arena; Slice_ + FArena)
|
||||
- §9 (Comment style; design-doc headers)
|
||||
- §12 (Per-language `<<` / `>>` rendering for C11)
|
||||
- §13 (The raddbg fallback; U64/S64, Vec2F32/Vec3S32, String8, force_inline)
|
||||
- §14 (Example program; the canonical usage pattern)
|
||||
|
||||
### 4.3 The Pass 1 reports + Pass 2 deliverables (the input content)
|
||||
|
||||
**Directories:**
|
||||
- Pass 1 (12 videos + 1 synthesis): `conductor/tracks/video_analysis_<slug>_20260621/report.md`
|
||||
- Pass 2 (33 deliverables): `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/<slug>/` and `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/<slug>/`
|
||||
|
||||
**Why read third:** the Pass 1 reports are the source content (12 videos × 1 deep-dive report + 1 synthesis). The Pass 2 deliverables are the v1-deobfuscated markdown. Pass 3 reads the Pass 1 reports + the Pass 2 deliverables and produces code.
|
||||
|
||||
**Read order per video:** the Pass 1 report first (the source content), then the Pass 2 deliverable (the v1-deobfuscated form). The Pass 2 deliverable is the input to the C11/Python translation.
|
||||
|
||||
### 4.4 The Python convention (manual_slop)
|
||||
|
||||
**File:** `conductor/product-guidelines.md` (the project's product guidelines) + the v2 lexicon's per-language rendering
|
||||
|
||||
**Why read fourth:** the Python convention is the manual_slop style. For the 2 Python videos + 1 synthesis, this is the primary style reference.
|
||||
|
||||
**Key manual_slop conventions:**
|
||||
- 1-space indent (per `conductor/product-guidelines.md` "AI-Optimized Compact Style")
|
||||
- Type hints (mandatory for all functions)
|
||||
- No comments (documentation lives in `/docs`)
|
||||
- `Result[T]` for error handling (per `conductor/product-guidelines.md` "Data-Oriented Error Handling")
|
||||
- Per-language `<<` / `>>` rendering: same as C11 (much_less / much_greater / weakly_coupled)
|
||||
|
||||
**For the Python videos:** if the manual_slop convention conflicts with a video-specific need, document the override in the per-video `<slug>_notes.md`.
|
||||
|
||||
---
|
||||
|
||||
## §5. The 11 videos (file path mapping)
|
||||
|
||||
For each video, Tier 2 reads:
|
||||
|
||||
| Slug | Pass 1 report | Pass 2 deliverable (v1) |
|
||||
|---|---|---|
|
||||
| `cs229_building_llms` | `conductor/tracks/video_analysis_cs229_building_llms_20260621/report.md` | `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/cs229_building_llms/` |
|
||||
| `probability_logic` | `conductor/tracks/video_analysis_probability_logic_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/probability_logic/` |
|
||||
| `entropy_epiplexity` | `conductor/tracks/video_analysis_entropy_epiplexity_20260621/report.md` | `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/entropy_epiplexity/` |
|
||||
| `score_dynamics_giorgini` | `conductor/tracks/video_analysis_score_dynamics_giorgini_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/score_dynamics_giorgini/` |
|
||||
| `platonic_intelligence_kumar` | `conductor/tracks/video_analysis_platonic_intelligence_kumar_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/platonic_intelligence_kumar/` |
|
||||
| `free_lunches_levin` | `conductor/tracks/video_analysis_free_lunches_levin_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/free_lunches_levin/` |
|
||||
| `generic_systems_fields` | `conductor/tracks/video_analysis_generic_systems_fields_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/generic_systems_fields/` |
|
||||
| `brain_counterintuitive` | `conductor/tracks/video_analysis_brain_counterintuitive_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/brain_counterintuitive/` |
|
||||
| `neural_dynamics_miller` | `conductor/tracks/video_analysis_neural_dynamics_miller_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/neural_dynamics_miller/` |
|
||||
| `multiscale_hoffman` | `conductor/tracks/video_analysis_multiscale_hoffman_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/multiscale_hoffman/` |
|
||||
| `synthesis` | `conductor/tracks/video_analysis_synthesis_20260621/report.md` | `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/synthesis/` |
|
||||
|
||||
**Note:** the 2 pilot videos (`cs229_building_llms` + `entropy_epiplexity`) have Pass 2 deliverables in `video_analysis_deob_pilot_20260621/`. The other 9 videos (plus the synthesis) have Pass 2 deliverables in `video_analysis_deob_apply_20260621/`.
|
||||
|
||||
---
|
||||
|
||||
## §6. Per-video deliverables
|
||||
|
||||
For each video, Tier 2 produces:
|
||||
|
||||
### 6.1 For C11 videos (9 videos)
|
||||
|
||||
In `conductor/tracks/video_analysis_deob_pass3_20260623/artifacts/<slug>/`:
|
||||
|
||||
| File | Description |
|
||||
|---|---|
|
||||
| `<slug>.c` | The main C11 program (uses duffle conventions) |
|
||||
| `<slug>.h` | The header (if the program has multiple files) |
|
||||
| `<slug>_translation.md` | The translation table from math → C11 (3-column per the pilot process improvement) |
|
||||
| `<slug>_decoder.md` | The per-term decoder for the C11 types/conventions used (tier-categorized per the pilot process improvement) |
|
||||
| `<slug>_notes.md` | Decisions made, alternatives considered, language override if any |
|
||||
|
||||
**C11 conventions to use (per c11_convention.md):**
|
||||
- Byte-width types: `U1`/`U2`/`U4`/`S1`/`S2`/`S4`/`B1`/`B2`/`B4`
|
||||
- Type modifiers: `R_` (restrict), `V_` (volatile), `LP_`, `I_`, `FI_`, `NI_`, `RO_`, `T_`
|
||||
- `TSet_` / `Struct_` / `Union_` / `Enum_` for declarations
|
||||
- `Slice_` + `FArena` + `farena_push_*` for slices/arena
|
||||
- `ooo_drift_`/`anchor_`/`drain_`/`weld_` for memory ordering
|
||||
- `INTELLISENSE_DIRECTIVES` + `#pragma region` for build flags
|
||||
- Design-doc header (30-100 lines) at the top of every file
|
||||
- `much_less` / `much_greater` / `weakly_coupled` for `<<` / `>>` rendering
|
||||
- raddbg fallback for patterns duffle doesn't cover (U64/S64, Vec2F32/Vec3S32, String8, force_inline)
|
||||
|
||||
### 6.2 For Python videos (2 videos + 1 synthesis)
|
||||
|
||||
In `conductor/tracks/video_analysis_deob_pass3_20260623/artifacts/<slug>/`:
|
||||
|
||||
| File | Description |
|
||||
|---|---|
|
||||
| `<slug>.py` | The main Python program (uses manual_slop convention) |
|
||||
| `<slug>_translation.md` | The translation table from math → Python (3-column) |
|
||||
| `<slug>_decoder.md` | The per-term decoder for the Python types/conventions used (tier-categorized) |
|
||||
| `<slug>_notes.md` | Decisions made, alternatives considered |
|
||||
|
||||
**Python conventions to use (per manual_slop):**
|
||||
- 1-space indent
|
||||
- Type hints (mandatory for all functions)
|
||||
- No comments (documentation lives in `/docs` and in `<slug>_notes.md`)
|
||||
- `Result[T]` for error handling (per the data-oriented error handling convention)
|
||||
- Encoding: `Scalar` / `float` / `integer` / `float64` per the v2 lexicon placeholder scheme
|
||||
- `much_less` / `much_greater` / `weakly_coupled` for `<<` / `>>` rendering (same as C11)
|
||||
|
||||
### 6.3 Global deliverables (after all 11 videos)
|
||||
|
||||
| File | Description |
|
||||
|---|---|
|
||||
| `conductor/tracks/video_analysis_deob_pass3_20260623/PASS3_REPORT.md` | The end-of-track report: 11 videos × 3 files = 33 deliverables, plus the cross-cutting observations |
|
||||
| `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md` | The end-of-track report at the canonical docs location |
|
||||
|
||||
**PASS3_REPORT.md structure (per the apply process improvement #3):**
|
||||
1. Executive summary
|
||||
2. What was produced (33 deliverables; per-video file count + LOC)
|
||||
3. Refinements discovered (Pass 3 → lexicon v3 candidates)
|
||||
4. Gaps identified (concepts the code couldn't capture)
|
||||
5. Process improvements (for any future Pass 3 expansion)
|
||||
6. Verification (4 + 3 criteria met for all 33 files)
|
||||
7. State
|
||||
8. See also
|
||||
|
||||
---
|
||||
|
||||
## §7. The 4 + 3 verification criteria (per v2 lexicon)
|
||||
|
||||
For each video's deliverables, the criteria are:
|
||||
|
||||
| Criterion | Status | How to verify |
|
||||
|---|---|---|
|
||||
| **Lossless** | required | Every concept from the Pass 2 deliverable is represented in the C11/Python code. No concept is dropped. |
|
||||
| **Bounded** | required | No `∞_val`; `Stream A = nat -> A` for processes. |
|
||||
| **Constructively typed** | required | Every expression has a type (C11 type or Python type hint). |
|
||||
| **Etymology-cited** | required | Every new term has the 1-line origin + 1-line definition history in `<slug>_decoder.md`. |
|
||||
| **Encoding-explicit** (Rule 5) | required | Every value-bearing term has an `encoding:` attribute (C11: explicit `Scalar`/`float`/`integer` placeholder; Python: type hint). |
|
||||
| **Form-anchored** | required | Every re-encoding has a form anchor in `<slug>_decoder.md`. |
|
||||
| **User-specific opt-in** | required | The principled form is always produced; the user-specific form (e.g., Sectored Language V1, GA reinterpretations) is opt-in and documented in `<slug>_notes.md`. |
|
||||
|
||||
**The code may or may not run** (per user 2026-06-23: "pass 3 is markdown for docs, but its going to be code files that may or may not run"). Compilation/execution is OPTIONAL; the code is judged on lossless + bounded + constructively-typed + etymology + encoding + form-anchor + user-specific opt-in.
|
||||
|
||||
---
|
||||
|
||||
## §8. Commit discipline (per `conductor/workflow.md`)
|
||||
|
||||
1. **Per-file atomic commits:** one commit per deliverable file. 11 videos × 3-5 files = 33-55 commits. Plus 2 commits for the PASS3_REPORT.md + end-of-track report.
|
||||
2. **Commit message format:** `conductor(deob_pass3): <slug> <file> - <description>`. Example: `conductor(deob_pass3): cs229_building_llms .c - LLM forward pass with duffle byte-width types`.
|
||||
3. **Git notes per commit:** attach a 1-3 sentence summary via `git notes add -m "..." <commit_sha>`.
|
||||
4. **No batching:** don't commit multiple files at once; each commit is a safe rollback point.
|
||||
5. **Working tree clean at the end of each video** (before moving to the next video).
|
||||
|
||||
---
|
||||
|
||||
## §9. The 6 open questions for Pass 3 (answered)
|
||||
|
||||
The user has answered the 6 open questions for Pass 3:
|
||||
|
||||
| # | Question | Answer |
|
||||
|---|---|---|
|
||||
| 1 | What is the applied domain? | Simple C11 or Python program per video (per user 2026-06-23) |
|
||||
| 2 | How should user-specific forms be applied? | Annotation if not code; pseudo sectr lang is incomplete, adapt per video (per user 2026-06-23) |
|
||||
| 3 | How should indefinites be encoded? | `Scalar` / `float` / `integer` placeholder scheme; `float64` only when target resolution matters (per user 2026-06-23) |
|
||||
| 4 | How should `<<` / `>>` render? | C11: `much_less` / `much_greater` / `weakly_coupled`; Python: same; Forth: named words (per user 2026-06-23 + v2 lexicon §9) |
|
||||
| 5 | How should the 4 + 3 criteria be adapted? | Criteria are OK; may ideate for applied domain (per user 2026-06-23) |
|
||||
| 6 | What is the user-facing artifact? | Code files (may or may not run) + markdown docs (per user 2026-06-23) |
|
||||
|
||||
Tier 2 should NOT re-ask these questions. Use the answers in this document and the v2 lexicon.
|
||||
|
||||
---
|
||||
|
||||
## §10. Risks (per the spec §9 + the v2 lexicon's risks)
|
||||
|
||||
| # | Risk | Mitigation |
|
||||
|---|---|---|
|
||||
| R1 | A video's content is too complex for a simple program | Document the limitation in `<slug>_notes.md`; reduce scope to "the core concept that the video provides" rather than the full video. |
|
||||
| R2 | The C11 convention is too restrictive (e.g., the video needs something duffle doesn't have) | Use the raddbg fallback (§13 of c11_convention.md). Document the fallback in `<slug>_notes.md`. |
|
||||
| R3 | The Python convention conflicts with a video's needs | Override per video; document in `<slug>_notes.md`. |
|
||||
| R4 | Tier 2 loses context on the principled vs user-specific distinction | Read §0.3 of the v2 lexicon first; refer to the per-language rendering §9. |
|
||||
| R5 | The 4 + 3 verification criteria are too strict for some videos | The criteria are documented in §7; if a video can't meet a criterion, document why in `<slug>_notes.md`. The criteria are the goal, not the bar. |
|
||||
| R6 | The code doesn't run and the user is disappointed | Per user 2026-06-23, "the code may or may not run" — the goal is the expression of concepts in code, not execution. |
|
||||
| R7 | The synthesis is too big for one Python file | Tier 2 may split the synthesis into a package (multiple files) if the content warrants. Document in `synthesis_notes.md`. |
|
||||
|
||||
---
|
||||
|
||||
## §11. The 12 Pass 1 videos — note on the 2 in the pilot
|
||||
|
||||
The 2 videos in the pilot (`cs229_building_llms` + `entropy_epiplexity`) have been deobfuscated in `video_analysis_deob_pilot_20260621/artifacts/`. These are inputs to Pass 3, not the final form. The other 9 videos (plus the synthesis) have been deobfuscated in `video_analysis_deob_apply_20260621/artifacts/`. **All 11 videos are in Pass 3's scope.**
|
||||
|
||||
---
|
||||
|
||||
## §12. Tier 2 execution model
|
||||
|
||||
Per the project's MMA Tier 2 + Tier 3 pattern:
|
||||
|
||||
1. **Tier 2 holds the full context:** the v2 lexicon + c11_convention.md + manual_slop + per-video Pass 1/2 inputs. Tier 2 orchestrates the 11 video executions.
|
||||
|
||||
2. **Tier 3 sub-agents (parallel) execute per video:** 4 Tier 3 sub-agents (one per cluster: A, B, C, D+synthesis). The Tier 3 sub-agents do the per-video code + docs. They read the v2 lexicon + c11_convention.md (the cluster sub-reports are NOT needed; Tier 3 reads the synthesis).
|
||||
|
||||
3. **Tier 2 synthesizes the per-cluster outputs into the global report:** after the 4 Tier 3 sub-agents finish, Tier 2 writes `PASS3_REPORT.md` + the end-of-track report at the canonical docs location.
|
||||
|
||||
4. **Tier 2 is NOT bypassed** for the per-video execution. The cross-cluster integration (per-language `<<` / `>>` rendering, the principled vs user-specific distinction) requires Tier 2's full context.
|
||||
|
||||
---
|
||||
|
||||
## §13. Tier 3 sub-agent prompts (per-cluster)
|
||||
|
||||
For each of the 4 Tier 3 sub-agents, the prompt is:
|
||||
|
||||
```
|
||||
You are a Tier 3 sub-agent for Pass 3 of the 3-pass research campaign.
|
||||
Your cluster: [A / B / C / D+synthesis]
|
||||
Your scope: [list of videos in the cluster]
|
||||
Your language defaults: [C11 / Python per video]
|
||||
Your inputs: [Pass 1 report paths + Pass 2 deliverable paths per video]
|
||||
Your conventions: [v2 lexicon + c11_convention.md (C11) OR manual_slop (Python)]
|
||||
Your deliverables per video: [C11 or Python per §6.1 / §6.2]
|
||||
Your verification: [4 + 3 criteria per §7]
|
||||
Your commit discipline: [per §8]
|
||||
```
|
||||
|
||||
The Tier 3 sub-agent's role is to produce the per-video code + docs. Tier 2 holds the cross-cluster context.
|
||||
|
||||
---
|
||||
|
||||
## §14. Cross-references
|
||||
|
||||
- **The v2 lexicon substrate:**
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md` (the codified operational spec)
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/terms_catalog.md` (76 terms, machine-readable)
|
||||
- `conductor/tracks/video_analysis_deob_lexicon_20260621/dedup_map.md` (6 noise-dedup maps; Maps 1, 2, 3 reshaped in v2)
|
||||
- `conductor/tracks/video_analysis_deob_warmup_20260621/prompt_template.md` (the LLM-direct operational spec)
|
||||
- **The C11 convention:**
|
||||
- `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md` (the user's idiomatic C11)
|
||||
- `conductor/tracks/video_analysis_deob_c11_reference_20260623/research/cluster_0_pikuma_duffle.md` (PRIMARY C11 source)
|
||||
- `conductor/tracks/video_analysis_deob_c11_reference_20260623/research/cluster_3_raddbg_src_base.md` (FALLBACK)
|
||||
- **The Python convention:**
|
||||
- `conductor/product-guidelines.md` (the project's product guidelines; the manual_slop convention)
|
||||
- `conductor/tech-stack.md` (the project's tech stack; the manual_slop tech stack)
|
||||
- **The Pass 1 reports:**
|
||||
- `conductor/tracks/video_analysis_<slug>_20260621/report.md` (12 videos + 1 synthesis)
|
||||
- **The Pass 2 deliverables:**
|
||||
- `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/<slug>/` (2 pilot videos)
|
||||
- `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/<slug>/` (9 apply videos + 1 synthesis)
|
||||
- **The pass 3 metadata:**
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/spec.md` (the design intent)
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/plan.md` (the TDD-ready plan)
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/metadata.json` (the machine-readable metadata)
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/state.toml` (the track state)
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/README.md` (the track index)
|
||||
|
||||
---
|
||||
|
||||
## §15. The "ok awesome" closing
|
||||
|
||||
When the user says "ok awesome" (or similar approval) after reviewing the deliverables, the track is SHIPPED. The Tier 2 sandbox should:
|
||||
1. Update `state.toml` to `status = "completed"`
|
||||
2. Commit the state update
|
||||
3. Attach a git note
|
||||
4. Present the deliverables to the user for final review
|
||||
|
||||
**The user's "ok awesome" is the formal close. The 3-pass research campaign ends here.**
|
||||
|
||||
---
|
||||
|
||||
*End of `TIER2_STARTER.md`. 15 sections. Self-contained. Tier 2 reads this file + the 4 PRIMARY inputs in §6, then executes.*
|
||||
@@ -0,0 +1,120 @@
|
||||
{
|
||||
"track_id": "video_analysis_deob_pass3_20260623",
|
||||
"name": "Pass 3 - C11/Python Projection (the final phase of the 3-pass research campaign)",
|
||||
"type": "research-to-code",
|
||||
"phase": "Pass 3 of 3 (the FINAL phase of the campaign)",
|
||||
"date": "2026-06-23",
|
||||
"status": "spec-pending-review",
|
||||
"blocked_by": [
|
||||
"video_analysis_deob_apply_20260621 (SHIPPED 2026-06-23, commit 8f2e8a69)",
|
||||
"video_analysis_deob_lexicon_v2_20260623 (SHIPPED 2026-06-23, commit b0c75992)",
|
||||
"video_analysis_deob_c11_reference_20260623 (SHIPPED 2026-06-23, commit 5f6e8423)"
|
||||
],
|
||||
"blocks": [
|
||||
"(end of the 3-pass research campaign)"
|
||||
],
|
||||
"method": "tier2-orchestration-with-tier3-parallel-subagents",
|
||||
"rationale": "Tier 2 holds the cross-cluster context (v2 lexicon + c11_convention.md + manual_slop + 11 video inputs). 4 parallel Tier 3 sub-agents (one per cluster) execute the per-video code. The cross-cluster integration requires the full context.",
|
||||
"scope": {
|
||||
"new_files": [
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/spec.md",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/plan.md",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/metadata.json",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/state.toml",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/README.md",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/TIER2_STARTER.md",
|
||||
"conductor/tracks/video_analysis_deob_pass3_20260623/PASS3_REPORT.md",
|
||||
"docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md"
|
||||
],
|
||||
"new_artifacts_per_video": [
|
||||
"<slug>.c + <slug>.h (C11) or <slug>.py (Python)",
|
||||
"<slug>_translation.md",
|
||||
"<slug>_decoder.md",
|
||||
"<slug>_notes.md"
|
||||
],
|
||||
"modified_files": [
|
||||
"conductor/tracks.md"
|
||||
],
|
||||
"deleted_files": [],
|
||||
"total_videos": 11,
|
||||
"total_deliverables_estimated": "33-55 (3-5 per video)",
|
||||
"total_commits_estimated": "35-58"
|
||||
},
|
||||
"tasks_count": 9,
|
||||
"commits_count_estimated": "35-58",
|
||||
"verification_criteria": [
|
||||
"11 videos processed (C11 or Python per the per-language default)",
|
||||
"4 + 3 verification criteria met for all 33 deliverables",
|
||||
"Per-language << / >> rendering applied (much_less / much_greater / weakly_coupled)",
|
||||
"Encoding placeholder scheme applied (float / integer / Scalar / float64)",
|
||||
"Per-file atomic commits (33-55 commits)",
|
||||
"Git notes per commit",
|
||||
"PASS3_REPORT.md written",
|
||||
"docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md written",
|
||||
"state.toml updated to status = completed after user approval",
|
||||
"No src/*.py changes",
|
||||
"No pyproject.toml dependencies"
|
||||
],
|
||||
"per_language_default": {
|
||||
"C11_videos": [
|
||||
"cs229_building_llms",
|
||||
"score_dynamics_giorgini",
|
||||
"platonic_intelligence_kumar",
|
||||
"free_lunches_levin",
|
||||
"generic_systems_fields",
|
||||
"brain_counterintuitive",
|
||||
"neural_dynamics_miller",
|
||||
"multiscale_hoffman",
|
||||
"cs336_architectures",
|
||||
"creikey_dl_cv"
|
||||
],
|
||||
"Python_videos": [
|
||||
"probability_logic",
|
||||
"entropy_epiplexity",
|
||||
"synthesis"
|
||||
]
|
||||
},
|
||||
"convention_files": {
|
||||
"v2_lexicon": "conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md",
|
||||
"c11_convention": "conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md",
|
||||
"manual_slop": "conductor/product-guidelines.md"
|
||||
},
|
||||
"hardware_target": {
|
||||
"user_directive": "Per user 2026-06-23: 'target up to 10k. when doing phase 3 you may have to pick a target, just do so within reason for a local hardware scenario (user with around a 1-10k USD budget I guess).'",
|
||||
"default": "Ryzen 9 / i9, RTX 4090, 128GB DDR5, 4TB NVMe (~$7-9k)",
|
||||
"fallback": "Mid-range desktop (~$3-5k) if a video is constrained"
|
||||
},
|
||||
"pre_existing_failures_remaining": [],
|
||||
"deferred_to_followup_tracks": [
|
||||
{
|
||||
"title": "Lexicon v3 — the 5 DEFERRED gaps",
|
||||
"description": "Address G1 (paradoxes of epiplexity), G2 (incomputable classification), G4 (type-class primitive), G7 (spacetime from trace logic), G9 (infinity-categories) in lexicon v3. Plus the 31 unresolved items from warmup §A.3 + §11.3.",
|
||||
"track_status": "future, user-led"
|
||||
}
|
||||
],
|
||||
"estimated_effort": {
|
||||
"method": "scope (per workflow.md Tier 1 Track Initialization Rules). NO day estimates.",
|
||||
"phase_1": "4 tasks: 4 PRIMARY inputs read",
|
||||
"phase_2": "4 tasks: 4 Tier 3 sub-agents dispatched (per cluster)",
|
||||
"phase_3": "3 tasks: PASS3_REPORT + end-of-track report + state update"
|
||||
},
|
||||
"risk_register": [
|
||||
{"id": "R1", "risk": "A video's content is too complex for a simple program", "likelihood": "medium", "impact": "low", "mitigation": "Document the limitation in <slug>_notes.md; reduce scope to the core concept"},
|
||||
{"id": "R2", "risk": "The C11 convention is too restrictive", "likelihood": "low", "impact": "medium", "mitigation": "Use the raddbg fallback; document in <slug>_notes.md"},
|
||||
{"id": "R3", "risk": "The Python convention conflicts with a video's needs", "likelihood": "low", "impact": "low", "mitigation": "Override per video; document in <slug>_notes.md"},
|
||||
{"id": "R4", "risk": "Tier 2 loses context on the principled vs user-specific distinction", "likelihood": "low", "impact": "high", "mitigation": "Read §0.3 of the v2 lexicon first"},
|
||||
{"id": "R5", "risk": "The 4 + 3 criteria are too strict for some videos", "likelihood": "medium", "impact": "low", "mitigation": "Document why in <slug>_notes.md; criteria are the goal, not the bar"},
|
||||
{"id": "R6", "risk": "The code doesn't run and the user is disappointed", "likelihood": "low", "impact": "medium", "mitigation": "Per user 2026-06-23, the code may or may not run"},
|
||||
{"id": "R7", "risk": "The synthesis is too big for one Python file", "likelihood": "low", "impact": "low", "mitigation": "Tier 2 may split into a package; document in synthesis_notes.md"}
|
||||
],
|
||||
"user_directives_logged": [
|
||||
{"id": "applied_domain_c11_or_python", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'The applied domain is making a simple program in C11 or python that conveys what the subject video provides. If in c11 it should follow a specific covention of c11 that raddbg uses or that I use with my duffle libs in pikuma ps1 or the forth bootslop projects. For python it should use manual slop's convention.'"},
|
||||
{"id": "hardware_target_up_to_10k", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'target up to 10k. when doing phase 3 you may have to pick a target, just do so within reason for a local hardware scenario (user with around a 1-10k USD budget I guess).'"},
|
||||
{"id": "user_specific_forms_annotation", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'Forms are somehwat annotation if its not code. When it comes to the code psuedo sectr lang is not complete and prob needs adapting or further adjustments.' Pseudo sectr lang is incomplete; Pass 3 will adapt per video."},
|
||||
{"id": "pass3_artifact_code", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'pass 3 is markdown for docs, but its going to be code files that may or may not run.' Code may or may not run; the goal is the expression of concepts."},
|
||||
{"id": "per_language_default_proposed", "date": "2026-06-23", "directive": "Per user 2026-06-23 (Tier 1 proposal): C11 for math/algorithms oriented (cs229, score_dynamics_giorgini, platonic_intelligence_kumar, generic_systems_fields, brain_counterintuitive, neural_dynamics_miller, multiscale_hoffman, cs336, creikey_dl_cv, free_lunches_levin); Python for probability/information-theoretic (probability_logic, entropy_epiplexity) + synthesis. Tier 2 may override per video."},
|
||||
{"id": "much_less_much_greater", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'weakly_coupled(...) is good for c11, much_less and much_greater can be used as well.'"},
|
||||
{"id": "template_notation_b_default", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'B as default. Use c+++ template language style in specific scenario when the user asks for it or if its better. (same with jai/odin language have the lexicon support it if the user desires it).'"},
|
||||
{"id": "c11_convention_synthesized", "date": "2026-06-23", "directive": "Per user 2026-06-23: 'use the forth bootslop and pikuma then. Use raddbg's base for stuff missing. otherwise go for jai/odin.' The C11 reference uses forth bootslop + Pikuma as PRIMARY sources and raddebugger/src/base as the FALLBACK for patterns the primary sources don't cover."}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,107 @@
|
||||
# Plan — Pass 3 (C11/Python Projection)
|
||||
|
||||
**Track:** `video_analysis_deob_pass3_20260623`
|
||||
**Method:** Tier 2 holds the full context; 4 parallel Tier 3 sub-agents (one per cluster) execute the per-video code
|
||||
**Atomicity:** Per-file atomic commits with git notes (per `TIER2_STARTER.md` §8)
|
||||
|
||||
---
|
||||
|
||||
## Phase 1: Tier 2 reads the 4 PRIMARY inputs
|
||||
|
||||
Focus: Tier 2 must read these 4 files fully before producing any code.
|
||||
|
||||
### Task 1.1: Read the v2 lexicon
|
||||
|
||||
- **File:** `conductor/tracks/video_analysis_deob_lexicon_20260621/lexicon.md`
|
||||
- **Focus sections:** §0.3, §1.5, §2, §3, §7, §9 (per `TIER2_STARTER.md` §4.1)
|
||||
- **Why:** the v2 lexicon is the canonical substrate
|
||||
|
||||
### Task 1.2: Read the C11 convention
|
||||
|
||||
- **File:** `conductor/tracks/video_analysis_deob_c11_reference_20260623/c11_convention.md`
|
||||
- **Focus sections:** §2, §3, §4, §7, §8, §9, §12, §13, §14 (per `TIER2_STARTER.md` §4.2)
|
||||
- **Why:** the C11 convention is the user's idiomatic C11 style guide
|
||||
|
||||
### Task 1.3: Read the Python convention (manual_slop)
|
||||
|
||||
- **File:** `conductor/product-guidelines.md` + `conductor/tech-stack.md`
|
||||
- **Focus:** 1-space indent, type hints, no comments, `Result[T]` error handling
|
||||
- **Why:** the Python convention is the user's idiomatic Python style
|
||||
|
||||
### Task 1.4: Read the 11 Pass 1 reports + Pass 2 deliverables
|
||||
|
||||
- **Files:** per `TIER2_STARTER.md` §5
|
||||
- **Why:** the source content for the 11 videos
|
||||
|
||||
---
|
||||
|
||||
## Phase 2: Tier 2 dispatches 4 parallel Tier 3 sub-agents (per cluster)
|
||||
|
||||
Focus: Each Tier 3 sub-agent handles one cluster. The 4 clusters are A (math foundations), B (Platonic AI), C (biological/cognitive), D (applied capstone + synthesis).
|
||||
|
||||
### Task 2.1: Dispatch Tier 3 sub-agent for cluster A (3 videos)
|
||||
|
||||
- **Videos:** `cs229_building_llms` (C11), `probability_logic` (Python), `entropy_epiplexity` (Python), `score_dynamics_giorgini` (C11)
|
||||
- **Wait:** 4 videos (A cluster has 4: cs229, probability_logic, entropy_epiplexity, score_dynamics_giorgini)
|
||||
- **Outputs:** 4 × 3-5 deliverables per video
|
||||
- **Sub-agent prompt:** per `TIER2_STARTER.md` §13
|
||||
|
||||
### Task 2.2: Dispatch Tier 3 sub-agent for cluster B (2 videos)
|
||||
|
||||
- **Videos:** `platonic_intelligence_kumar` (C11), `free_lunches_levin` (C11)
|
||||
- **Outputs:** 2 × 3-5 deliverables per video
|
||||
|
||||
### Task 2.3: Dispatch Tier 3 sub-agent for cluster C (4 videos)
|
||||
|
||||
- **Videos:** `generic_systems_fields` (C11), `brain_counterintuitive` (C11), `neural_dynamics_miller` (C11), `multiscale_hoffman` (C11)
|
||||
- **Outputs:** 4 × 3-5 deliverables per video
|
||||
|
||||
### Task 2.4: Dispatch Tier 3 sub-agent for cluster D + synthesis (2 videos)
|
||||
|
||||
- **Videos:** `cs336_architectures` (C11), `creikey_dl_cv` (C11), `synthesis` (Python)
|
||||
- **Wait:** 3 videos (D cluster has 2: cs336, creikey_dl_cv; plus synthesis)
|
||||
- **Outputs:** 3 × 3-5 deliverables per video
|
||||
|
||||
---
|
||||
|
||||
## Phase 3: Tier 2 synthesizes the per-cluster outputs
|
||||
|
||||
Focus: After the 4 Tier 3 sub-agents finish, Tier 2 writes the global reports.
|
||||
|
||||
### Task 3.1: Write `PASS3_REPORT.md`
|
||||
|
||||
- **File:** `conductor/tracks/video_analysis_deob_pass3_20260623/PASS3_REPORT.md`
|
||||
- **Structure (per `TIER2_STARTER.md` §6.3):**
|
||||
1. Executive summary
|
||||
2. What was produced (33 deliverables; per-video file count + LOC)
|
||||
3. Refinements discovered (Pass 3 → lexicon v3 candidates)
|
||||
4. Gaps identified (concepts the code couldn't capture)
|
||||
5. Process improvements
|
||||
6. Verification (4 + 3 criteria met for all 33 files)
|
||||
7. State
|
||||
8. See also
|
||||
|
||||
### Task 3.2: Write the end-of-track report
|
||||
|
||||
- **File:** `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md`
|
||||
- **Content:** the canonical end-of-track report at the docs location
|
||||
|
||||
### Task 3.3: Update `state.toml`
|
||||
|
||||
- **After user approval:** update `state.toml` to `status = "completed"`
|
||||
- **Commit the state update**
|
||||
|
||||
---
|
||||
|
||||
## Atomic commit summary
|
||||
|
||||
- Per video: 3-5 commits (one per deliverable file)
|
||||
- 11 videos: 33-55 commits
|
||||
- Global: 2 commits (PASS3_REPORT.md + end-of-track report)
|
||||
- State update: 1 commit (after user approval)
|
||||
|
||||
**Total: ~35-58 atomic commits.** Per the workflow's "atomic per-task commits" rule, each commit is a safe rollback point. Each has a 1-3 sentence git note.
|
||||
|
||||
---
|
||||
|
||||
*End of `plan.md`. 3 phases. Tier 2 holds the context; Tier 3 sub-agents execute per-cluster in parallel.*
|
||||
@@ -0,0 +1,139 @@
|
||||
# Track Specification: Pass 3 — C11/Python Projection (Pass 3 of 3)
|
||||
|
||||
**Track ID:** `video_analysis_deob_pass3_20260623`
|
||||
**Type:** Research-to-code (the final phase of the 3-pass research campaign)
|
||||
**Date:** 2026-06-23
|
||||
**Status:** spec DRAFT (pending user review)
|
||||
**Blocked by:** `video_analysis_deob_apply_20260621` (SHIPPED) + `video_analysis_deob_lexicon_v2_20260623` (SHIPPED) + `video_analysis_deob_c11_reference_20260623` (SHIPPED)
|
||||
**Blocks:** (the end of the 3-pass campaign)
|
||||
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
Pass 3 is the final phase of the 3-pass research campaign. Pass 1 produced 12 deep-dive reports + 1 cross-cutting synthesis. Pass 2 (apply) produced 33 deobfuscated markdown deliverables using the v1 lexicon. The v2 corrective patch refined the lexicon. **Pass 3 projects the deobfuscated content to C11 or Python code that conveys the subject video's content.** The code may or may not run (per user 2026-06-23); the goal is to express the concepts in code using the user's idiomatic style.
|
||||
|
||||
The applied domain is "simple programs that convey what each Pass 1 video provides." Per the user 2026-06-23:
|
||||
- C11 follows the conventions of raddbg / duffel libs in Pikuma ps1 / forth bootslop (the `c11_convention.md` synthesizes these)
|
||||
- Python follows the manual_slop convention (`conductor/product-guidelines.md`)
|
||||
|
||||
The 11 videos are processed per the per-language default (C11 for math/algorithms oriented; Python for probability/information-theoretic; synthesis in Python). Hardware target: up to $10k workstation.
|
||||
|
||||
---
|
||||
|
||||
## Current State Audit (as of 2026-06-23, commit `5f6e8423`)
|
||||
|
||||
### Already Implemented (DO NOT re-implement)
|
||||
|
||||
- **The v2 lexicon substrate** (`lexicon.md` + `terms_catalog.md` + `dedup_map.md` + `prompt_template.md` + `report.md` v2) — the canonical input for Pass 3
|
||||
- **The C11 convention** (`c11_convention.md` + 4 cluster sub-reports) — the user's idiomatic C11
|
||||
- **The 6 open questions for Pass 3** (answered; see `TIER2_STARTER.md` §9)
|
||||
- **The 12 refinements + 8 gaps** documented in `apply_report.md` — for the lexicon, not the C11 reference
|
||||
- **The 33 Pass 2 deliverables** (11 videos × 3 files) — the input content for Pass 3
|
||||
- **The 2 pilot videos** (`cs229_building_llms` + `entropy_epiplexity`) — already in Pass 2; the pilot deliverables are inputs to Pass 3 but not the final form
|
||||
|
||||
### Gaps to Fill (This Track's Scope)
|
||||
|
||||
For each of the 11 videos, produce:
|
||||
- C11 or Python code that conveys the subject video's content
|
||||
- Markdown docs (translation table, decoder, notes)
|
||||
- 4 + 3 verification criteria met (per the v2 lexicon)
|
||||
|
||||
For the global track:
|
||||
- `PASS3_REPORT.md` (the end-of-track report)
|
||||
- `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md` (canonical docs)
|
||||
|
||||
---
|
||||
|
||||
## Goals
|
||||
|
||||
1. **Produce C11 or Python code per video** (11 videos total) using the user's idiomatic style
|
||||
2. **Use the C11 convention** (c11_convention.md) for the 9 C11 videos
|
||||
3. **Use the manual_slop convention** for the 2 Python videos + 1 synthesis
|
||||
4. **Apply the per-language `<<` / `>>` rendering** (much_less / much_greater / weakly_coupled with tolerance)
|
||||
5. **Apply the encoding placeholder scheme** (`float` / `integer` / `Scalar` / `float64`)
|
||||
6. **Meet the 4 + 3 verification criteria** (Lossless, Bounded, Constructively typed, Etymology-cited, Encoding-explicit, Form-anchored, User-specific opt-in)
|
||||
7. **Hand off to the user** — Pass 3 closes the 3-pass research campaign
|
||||
|
||||
---
|
||||
|
||||
## Functional Requirements
|
||||
|
||||
For each of the 11 videos, produce the per-video deliverables (per `TIER2_STARTER.md` §6):
|
||||
|
||||
- **For C11 videos (9):** `<slug>.c` + `<slug>.h` (if needed) + `<slug>_translation.md` + `<slug>_decoder.md` + `<slug>_notes.md`
|
||||
- **For Python videos (2 + 1 synthesis):** `<slug>.py` + `<slug>_translation.md` + `<slug>_decoder.md` + `<slug>_notes.md`
|
||||
|
||||
Plus the global deliverables:
|
||||
- `conductor/tracks/video_analysis_deob_pass3_20260623/PASS3_REPORT.md`
|
||||
- `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md`
|
||||
|
||||
Plus per-task atomic commits with git notes (per `TIER2_STARTER.md` §8).
|
||||
|
||||
---
|
||||
|
||||
## Non-Functional Requirements
|
||||
|
||||
- **NFR1: No `src/*.py` changes.** Pass 3 produces C11 (.c/.h) and Python (.py) code in the track folder, not in the project's `src/` directory.
|
||||
- **NFR2: Per-file atomic commits.** 33-55 commits (one per deliverable file) + 2 commits for the global reports.
|
||||
- **NFR3: No day estimates.** Scope measured in files (33-55 deliverables + 2 global reports).
|
||||
- **NFR4: Git notes per commit.** Each commit has a 1-3 sentence summary attached via `git notes add`.
|
||||
- **NFR5: Lossless.** Every concept from the Pass 2 deliverable is represented in the C11/Python code.
|
||||
- **NFR6: 4 + 3 verification criteria met.** Lossless, Bounded, Constructively typed, Etymology-cited, Encoding-explicit, Form-anchored, User-specific opt-in.
|
||||
|
||||
---
|
||||
|
||||
## Architecture Reference
|
||||
|
||||
- **The v2 lexicon substrate:** `conductor/tracks/video_analysis_deob_lexicon_20260621/`
|
||||
- **The C11 convention:** `conductor/tracks/video_analysis_deob_c11_reference_20260623/`
|
||||
- **The Python convention (manual_slop):** `conductor/product-guidelines.md` + `conductor/tech-stack.md`
|
||||
- **The Pass 1 reports:** `conductor/tracks/video_analysis_<slug>_20260621/report.md`
|
||||
- **The Pass 2 deliverables:** `conductor/tracks/video_analysis_deob_pilot_20260621/artifacts/<slug>/` + `conductor/tracks/video_analysis_deob_apply_20260621/artifacts/<slug>/`
|
||||
- **The Pass 3 dispatch prompt:** `TIER2_STARTER.md` (this folder)
|
||||
|
||||
---
|
||||
|
||||
## Out of Scope
|
||||
|
||||
- **The 5 DEFERRED gaps** (paradoxes of epiplexity, incomputable classification, type-class primitive, spacetime from trace logic, infinity-categories) — Pass 3 doesn't address them; deferred to lexicon v3
|
||||
- **The 3 INDEFINITE gaps** (quantum eigen function equivalence, multiscale_hoffman spacetime, Hoffman-Prakash 80% complete) — preserved with honest epistemic hedging
|
||||
- **The Sectored Language V1 adaptation for code** — Pass 3 may adapt per video; the pseudo sectr lang is incomplete (per user 2026-06-23)
|
||||
- **Compilation / execution verification** — the code may or may not run; the goal is the expression of concepts (per user 2026-06-23)
|
||||
- **Production code** — Pass 3 produces code in the track folder, not in the project's `src/`
|
||||
|
||||
---
|
||||
|
||||
## Risk Register
|
||||
|
||||
Per `TIER2_STARTER.md` §10.
|
||||
|
||||
| # | Risk | Likelihood | Impact | Mitigation |
|
||||
|---|---|---|---|---|
|
||||
| R1 | A video's content is too complex for a simple program | medium | low | Document the limitation in `<slug>_notes.md`; reduce scope to "the core concept" |
|
||||
| R2 | The C11 convention is too restrictive | low | medium | Use the raddbg fallback; document in `<slug>_notes.md` |
|
||||
| R3 | The Python convention conflicts with a video's needs | low | low | Override per video; document in `<slug>_notes.md` |
|
||||
| R4 | Tier 2 loses context on the principled vs user-specific distinction | low | high | Read §0.3 of the v2 lexicon first |
|
||||
| R5 | The 4 + 3 criteria are too strict for some videos | medium | low | Document why in `<slug>_notes.md`; criteria are the goal, not the bar |
|
||||
| R6 | The code doesn't run and the user is disappointed | low | medium | Per user 2026-06-23, "the code may or may not run" |
|
||||
| R7 | The synthesis is too big for one Python file | low | low | Tier 2 may split into a package; document in `synthesis_notes.md` |
|
||||
|
||||
---
|
||||
|
||||
## Verification Criteria
|
||||
|
||||
- [ ] 11 videos processed (C11 or Python per the per-language default)
|
||||
- [ ] 4 + 3 verification criteria met for all 33 deliverables
|
||||
- [ ] Per-language `<<` / `>>` rendering applied (much_less / much_greater / weakly_coupled)
|
||||
- [ ] Encoding placeholder scheme applied (float / integer / Scalar / float64)
|
||||
- [ ] Per-file atomic commits (33-55 commits)
|
||||
- [ ] Git notes per commit
|
||||
- [ ] PASS3_REPORT.md written (the end-of-track report)
|
||||
- [ ] `docs/reports/TRACK_COMPLETION_video_analysis_deob_pass3_20260623.md` written
|
||||
- [ ] `state.toml` updated to `status = "completed"` after user approval
|
||||
- [ ] No `src/*.py` changes
|
||||
- [ ] No `pyproject.toml` dependencies (no new packages)
|
||||
|
||||
---
|
||||
|
||||
*End of `spec.md`. 14 sections. Scope: 11 videos × 3-5 deliverables = 33-55 files + 2 global reports. The 3-pass research campaign ends here.*
|
||||
@@ -0,0 +1,69 @@
|
||||
# Track state for video_analysis_deob_pass3_20260623
|
||||
# Updated by Tier 2 Tech Lead during execution
|
||||
|
||||
[meta]
|
||||
track_id = "video_analysis_deob_pass3_20260623"
|
||||
name = "Pass 3 - C11/Python Projection (the final phase of the 3-pass research campaign)"
|
||||
status = "active"
|
||||
current_phase = 1
|
||||
last_updated = "2026-06-23"
|
||||
|
||||
[blocked_by]
|
||||
video_analysis_deob_apply_20260621 = "shipped 2026-06-23 (8f2e8a69)"
|
||||
video_analysis_deob_lexicon_v2_20260623 = "shipped 2026-06-23 (b0c75992)"
|
||||
video_analysis_deob_c11_reference_20260623 = "shipped 2026-06-23 (5f6e8423)"
|
||||
|
||||
[blocks]
|
||||
# Pass 3 is the FINAL phase. No downstream tracks.
|
||||
# (end of the 3-pass research campaign)
|
||||
|
||||
[phases]
|
||||
phase_1 = { status = "pending", checkpointsha = "", name = "Tier 2 reads the 4 PRIMARY inputs" }
|
||||
phase_2 = { status = "pending", checkpointsha = "", name = "Tier 2 dispatches 4 parallel Tier 3 sub-agents (per cluster)" }
|
||||
phase_3 = { status = "pending", checkpointsha = "", name = "Tier 2 synthesizes the per-cluster outputs" }
|
||||
|
||||
[tasks]
|
||||
# Phase 1 (read inputs)
|
||||
t1_1 = { status = "pending", commit_sha = "", description = "Read the v2 lexicon (the canonical substrate)" }
|
||||
t1_2 = { status = "pending", commit_sha = "", description = "Read the C11 convention (the user's idiomatic C11 style guide)" }
|
||||
t1_3 = { status = "pending", commit_sha = "", description = "Read the Python convention (manual_slop)" }
|
||||
t1_4 = { status = "pending", commit_sha = "", description = "Read the 11 Pass 1 reports + Pass 2 deliverables (the input content)" }
|
||||
|
||||
# Phase 2 (dispatch Tier 3 sub-agents)
|
||||
t2_1 = { status = "pending", commit_sha = "", description = "Dispatch Tier 3 sub-agent for cluster A (4 videos: cs229, probability_logic, entropy_epiplexity, score_dynamics_giorgini)" }
|
||||
t2_2 = { status = "pending", commit_sha = "", description = "Dispatch Tier 3 sub-agent for cluster B (2 videos: platonic_intelligence_kumar, free_lunches_levin)" }
|
||||
t2_3 = { status = "pending", commit_sha = "", description = "Dispatch Tier 3 sub-agent for cluster C (4 videos: generic_systems_fields, brain_counterintuitive, neural_dynamics_miller, multiscale_hoffman)" }
|
||||
t2_4 = { status = "pending", commit_sha = "", description = "Dispatch Tier 3 sub-agent for cluster D + synthesis (3 videos: cs336_architectures, creikey_dl_cv, synthesis)" }
|
||||
|
||||
# Phase 3 (synthesize)
|
||||
t3_1 = { status = "pending", commit_sha = "", description = "Write PASS3_REPORT.md (the end-of-track report)" }
|
||||
t3_2 = { status = "pending", commit_sha = "", description = "Write the end-of-track report at docs/reports/" }
|
||||
t3_3 = { status = "pending", commit_sha = "", description = "Update state.toml to status = completed (after user approval)" }
|
||||
|
||||
[verification]
|
||||
all_11_videos_processed = false
|
||||
c11_convention_applied = false
|
||||
manual_slop_applied = false
|
||||
per_language_rendering_applied = false
|
||||
encoding_placeholder_applied = false
|
||||
lossless_coverage = false
|
||||
4_plus_3_criteria_met = false
|
||||
per_file_atomic_commits = false
|
||||
git_notes_per_commit = false
|
||||
pass3_report_committed = false
|
||||
end_of_track_report_committed = false
|
||||
state_toml_completed = false
|
||||
no_src_changes = false
|
||||
no_pyproject_deps = false
|
||||
|
||||
[user_directives_logged]
|
||||
applied_domain_c11_or_python = "Per user 2026-06-23: simple C11 or Python program per video; C11 follows raddbg/duffel/pikuma/forth bootslop; Python follows manual_slop"
|
||||
hardware_target_up_to_10k = "Per user 2026-06-23: 'target up to 10k'; default workstation (Ryzen 9/i9, RTX 4090, 128GB DDR5)"
|
||||
user_specific_forms_annotation = "Per user 2026-06-23: pseudo sectr lang is incomplete; Pass 3 adapts per video"
|
||||
pass3_artifact_code = "Per user 2026-06-23: code may or may not run; goal is expression of concepts"
|
||||
much_less_much_greater = "Per user 2026-06-23: 'weakly_coupled(...) is good for c11, much_less and much_greater can be used as well'"
|
||||
template_notation_b_default = "Per user 2026-06-23: 'B as default; C++/Odin/Jai opt-in'"
|
||||
c11_convention_synthesized = "Per user 2026-06-23: 'use the forth bootslop and pikuma then. Use raddbg's base for stuff missing. otherwise go for jai/odin.'"
|
||||
|
||||
[no_day_estimates]
|
||||
"Enforced per conductor/workflow.md Tier 1 Track Initialization Rules. Scope measured in files (33-55 deliverables + 2 global reports + 1 track scaffold = ~40-60 file changes). 35-58 atomic commits planned."
|
||||
Reference in New Issue
Block a user