Skip to content

Commit 02eba26

Browse files
check our claims
1 parent 8ce113e commit 02eba26

204 files changed

Lines changed: 3435 additions & 211 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

.gitignore

Lines changed: 25 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,28 @@
1+
# === HUF Project Additions ===
2+
3+
# Jupyter artifacts
4+
.ipynb_checkpoints/
5+
*.ipynb_checkpoints
6+
7+
# Python bytecode
8+
__pycache__/
9+
*.pyc
10+
*.pyo
11+
12+
# OS files
13+
.DS_Store
14+
Thumbs.db
15+
desktop.ini
16+
17+
# LibreOffice lock files
18+
.~lock.*
19+
20+
# Temporary files
21+
*.tmp
22+
*~
23+
24+
# === Node.js (original template) ===
25+
126
# Logs
227
logs
328
*.log

.ipynb_checkpoints/Untitled-checkpoint.ipynb

Lines changed: 0 additions & 3 deletions
This file was deleted.

HUFv4/.~lock.PROTOTYPE_dual_column.pdf#

Lines changed: 0 additions & 1 deletion
This file was deleted.

HUFv4/lu108120gvgi.tmp

-71.1 KB
Binary file not shown.

README.md

Lines changed: 41 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -8,6 +8,36 @@ HUF is presented here as an **open-loop framework**. The instrument reads. The h
88

99
---
1010

11+
## New Discovery: Entropy-Invariant Time Transformer (EITT)
12+
13+
**April 2026.** Shannon entropy appears empirically near-invariant under geometric-mean temporal decimation of compositional time series in the energy datasets we tested.
14+
15+
Measured: **0.18% variation** across a 341:1 compression ratio (daily → annual European electricity compositions, 8 carriers, 4089 trading days). Confirmed independently on 6 countries at **1.02% mean variation**, then driven to **0.029%** on a third independent test (EMBER monthly data, Copilot calibrator suite).
16+
17+
For comparison, Aitchison variance drops 55% and total variation drops 99.7% across the same ladder. Shannon entropy changed very little in the tested ladders.
18+
19+
This suggests that the geometric mean — a central operation in CoDa — may act as an entropy-stabilizing temporal filter for compositional time series with temporal structure.
20+
21+
**Cross-domain confirmation (April 9, 2026).** EITT has now been tested on five domains beyond energy: hardware degradation (Backblaze drive stats, K=4, 24 months — 0.03%), a 120-stock price-level portfolio (K=9, 74 months — 0.08%; price-level weighted, not market-cap), cosmological observation (Planck 353 GHz half-mission temporal split — 0.3%), and commodities (gold/silver ratio, K=2, 338 years — fails at 6.7%, holds at 0.38% when reconstructed to K=4, validated out-of-sample). Bootstrap confidence intervals computed for all domains. The EITT inversion principle: when EITT fails on K parts, increasing K until it holds reveals the system's true dimensionality. **Honesty notes:** The Planck spatial EITT claim has been retracted (HEALPix NESTED pixel ordering does not preserve spatial locality). The financial composition is from a personal stock selection, not a standard index. At 2:1 compression, arithmetic-mean decimation also preserves entropy; the geometric mean's advantage appears at higher compression ratios. See code/analysis/honesty_tests_results_2026april9.txt for full adversarial test results.
22+
23+
We are presenting this as an empirical finding with a mapped boundary, not as a theorem.
24+
25+
**The d(CoDa)/dt chain.** EITT describes what is conserved under temporal compression. The complementary question — how fast and in which direction is the composition moving — is answered by three temporal layers: perturbation velocity (scalar speed on the simplex), ILR balance trajectory (structural path), and balance derivative dB/dt (directed rate of structural change along each partition). In raw proportions, the zero-sum constraint (Σ dx_i/dt = 0) forces a relay — every gain financed by losses elsewhere. In ILR coordinates, the balance derivatives are independent, decomposing the relay into measurable structural rates. The derivative describes the dynamics; EITT describes the invariant. Together they form a temporal analysis framework for compositional time series.
26+
27+
| What to read | File | Time |
28+
|--------------|------|------|
29+
| The finding, the numbers, the mechanism | [`EITT_FINDING.md`](drafts/codawork-2026/EITT_FINDING.md) | 10 min |
30+
| Plain-language explanation of everything | [`EITT_COMPLETE_EXPLANATION.md`](drafts/codawork-2026/EITT_COMPLETE_EXPLANATION.md) | 15 min |
31+
| What this offers CoDa specifically | [`EITT_CODA_POSITION.md`](drafts/codawork-2026/EITT_CODA_POSITION.md) | 10 min |
32+
| How EITT relates to 10 entropy territories | [`EITT_ENTROPY_LANDSCAPE.md`](drafts/codawork-2026/EITT_ENTROPY_LANDSCAPE.md) | 20 min |
33+
| Adversarial testing (10 pass, 7 fail, boundary mapped) | [`EITT_Adversarial_001.json`](data/codawork-samples/S016/EITT_Adversarial_001.json) | 10 min |
34+
| Copilot's calibrator lab and boundary condition | [`data/eitt_lab/`](data/eitt_lab/) | 30 min |
35+
| The residual explained (Hessian footprint) | [`EITT_Residual_Analysis_001.json`](data/codawork-samples/S016/EITT_Residual_Analysis_001.json) | 15 min |
36+
37+
**Posture:** We found this empirically. We can't prove it. Can you?
38+
39+
---
40+
1141
## What HUF claims
1242

1343
- Composition can be monitored directly, rather than treated only as a statistical correction or secondary view.
@@ -97,7 +127,7 @@ If you find a failure mode not on this list, please open an issue.
97127
| Directory | Function | Description |
98128
|-----------|----------|-------------|
99129
| [`huf-gov/`](huf-gov/) | **Observation** | The open-loop instrument. Science, evidence, governance, tools. |
100-
| [`huf-cls/`](huf-cls/) | **Control** | The closed-loop actuator. Published for transparency, not for deployment. |
130+
| [`huf-cls/`](huf-cls/) | **Control** | The closed-loop actuator. Included for transparency. Not part of the Coimbra / CoDaWork ask. |
101131
| [`reference/`](reference/) | **Foundation** | 68 machine-readable JSON specifications, technical notes, pillar documents. |
102132

103133
### Supporting material
@@ -107,7 +137,12 @@ If you find a failure mode not on this list, please open an issue.
107137
| [`context-books/`](context-books/) | Four audience-specific editions of the core theory (general, engineering, physics, sciences). |
108138
| [`math-books/`](math-books/) | Formal mathematical foundations. 42 numbered items with full proofs. |
109139
| [`process/`](process/) | How this was built — collective review traces, session ledgers, governance documents. |
110-
| [`drafts/`](drafts/) | Submission papers (pre-validation), the Human-AI Accord. Work in progress. |
140+
| [`drafts/`](drafts/) | Submission papers (pre-validation), the Human-AI Accord, EITT materials. Work in progress. |
141+
| [`data/`](data/) | Reproducible datasets: EMBER energy, CoDaWork samples, EITT lab (calibration, adversarial, residual). |
142+
| [`code/`](code/) | Analysis scripts (Backblaze, energy, HellTest stress tests), diagnostic tools. |
143+
| [`src/`](src/) | Document generation build system (JS builders + shared modules + Python validation). |
144+
| [`notebooks/`](notebooks/) | Onboarding Jupyter series (5 interactive tutorials). |
145+
| [`docs/`](docs/) | Generated canonical documents: explorations, governance, 7 academic paper drafts, reference. |
111146

112147
### Governance documents
113148

@@ -132,9 +167,11 @@ Pre-CoDa tools (v2 analyzer, Universal analyzer) have been moved to [`archive/to
132167

133168
### CoDaWork 2026 Conference Materials
134169

135-
Abstract accepted by Prof. Juan Jose Egozcue, Chair of the Scientific Committee. Conference: June 1-6, 2026, Coimbra, Portugal. See [`drafts/codawork-2026/`](drafts/codawork-2026/) for the full preparation suite (18 files) including formulas, vocabulary, researcher profiles, battle card, executive summary, core explanation, and the HUF-CoDa translator.
170+
Abstract accepted by Prof. Juan Jose Egozcue, Chair of the Scientific Committee. Conference: June 1-6, 2026, Coimbra, Portugal. See [`drafts/codawork-2026/`](drafts/codawork-2026/) for the full preparation suite including formulas, vocabulary, researcher profiles, battle card, executive summary, core explanation, the HUF-CoDa translator, and all EITT materials.
171+
172+
**New in April 2026 — EITT:** The Entropy-Invariant Time Transformer discovery. Shannon entropy near-invariant under geometric-mean temporal decimation. Full finding, adversarial testing, residual analysis, calibration lab, boundary conditions, and an entropy landscape mapping EITT against 10 existing entropy territories. See [`EITT_FINDING.md`](drafts/codawork-2026/EITT_FINDING.md) and [`data/eitt_lab/`](data/eitt_lab/).
136173

137-
**New entry point for CoDa community:** [`START_HERE.md`](START_HERE.md) — includes a "For CoDa / Compositional Data Analysts" section with direct links to all relevant materials.
174+
**Entry point for CoDa community:** [`START_HERE.md`](START_HERE.md) — includes a "For CoDa / Compositional Data Analysts" section with direct links to all relevant materials including EITT.
138175

139176
---
140177

REPO_MAP.md

Lines changed: 96 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -1,58 +1,123 @@
11
# HUF Repository Map
22

3+
*Updated: 2026-04-09 (S016 — EITT discovery, repo reorganization)*
4+
35
## Folder Structure
46

57
```
68
HUF/
7-
├── docs/ # Canonical documents (latest versions only)
8-
│ ├── index.md # Reading order and document status
9-
│ ├── pillars/ # The four core HUF papers
10-
│ │ ├── HUF_Sufficiency_Frontier_v3.6.docx
11-
│ │ ├── HUF_Fourth_Category_v2.6.docx
12-
│ │ ├── HUF_Triad_Synthesis_v1.6.docx
13-
│ │ └── HUF_Collective_Trace_v5.10.docx
14-
│ ├── governance/ # Review tracking and category classification
15-
│ │ ├── review_catalog.md
16-
│ │ ├── HUF_Category_Class_Structure_Tree_v1.4.docx
17-
│ │ └── HUF_Collective_Review_March2026.docx
18-
│ ├── explorations/ # Case studies, proofs, formal explorations
19-
│ ├── reviews/ # Individual AI system review documents
20-
│ └── appendices/ # Code appendix, methodology appendix
21-
├── src/ # Source code
9+
├── README.md # What HUF claims, EITT discovery, how to test it
10+
├── START_HERE.md # Tiered entry paths (60s, 5m, 30m) + CoDa welcome
11+
├── CITATION.cff # Academic citation metadata
12+
├── CONTRIBUTING.md # 5-AI collective development model
13+
├── VERSION_MANIFEST.json # Machine-readable version tracking
14+
├── package.json # Node.js dependency (docx-js) for build system
15+
16+
├── huf-gov/ # The open-loop instrument
17+
│ ├── science/ # MONITOR-001 taxonomy, papers
18+
│ ├── evidence/case-studies/ # Energy, GDP, Backblaze, Ramsar, Planck, Toronto
19+
│ ├── governance/ # LOOP-001, KILL-001, SAFE-001, HANDOFF-001
20+
│ └── tools/spectrum-analyzer/ # v3 HTML analyzer + JSX source
21+
22+
├── huf-cls/ # The closed-loop actuator (published for transparency)
23+
│ ├── architecture/ # CL-01 through CL-05
24+
│ ├── formulas/ # Sigmoid actuator, K_eff, amplifier
25+
│ └── calibration/ # CL-05 sensitivity analysis
26+
27+
├── reference/ # Foundation
28+
│ ├── machine-readable/ # 68+ JSON specifications
29+
│ ├── pillars/ # Theory, methodology, architecture (.docx)
30+
│ ├── technical-notes/ # PLL, calibration, architecture, interaction studies
31+
│ └── wiki/ # Quick-reference articles
32+
33+
├── data/ # Reproducible datasets
34+
│ ├── manifests/ # Data, benchmark, and release manifests
35+
│ ├── checksums/ # SHA256 document integrity hashes
36+
│ ├── codawork-samples/S016/ # EITT adversarial, midrange confirmation, residual
37+
│ └── eitt_lab/ # ★ EITT calibration lab
38+
│ ├── copilot_outputs/ # Calibrator suite, boundary condition, PCA variants
39+
│ └── claude_analysis/ # Residual analysis (Hessian footprint)
40+
41+
├── drafts/ # Submission papers, conference materials, work in progress
42+
│ ├── codawork-2026/ # CoDaWork 2026 full preparation suite
43+
│ │ ├── EITT_FINDING.md # ★ The EITT discovery
44+
│ │ ├── EITT_CODA_POSITION.md # ★ What EITT offers CoDa
45+
│ │ ├── EITT_ENTROPY_LANDSCAPE.md # ★ EITT vs 10 entropy territories
46+
│ │ ├── EITT_COMPLETE_EXPLANATION.md # ★ Plain-language walkthrough
47+
│ │ ├── DANCE_CARD.md # Conference strategy + 8 union territories
48+
│ │ ├── THE_CORE.md # Coherence chain for CoDa community
49+
│ │ ├── FORMULA_REFERENCE.md # HUF↔CoDa formula reference
50+
│ │ ├── BATTLE_CARD.md # 10 hardest questions
51+
│ │ └── ... # Abstract, translator, explorer, ternary, etc.
52+
│ ├── accord/ # Human-AI Accord
53+
│ └── papers/ # Academic submissions
54+
55+
├── docs/ # Generated canonical documents
56+
│ ├── explorations/ # Case studies (Backblaze, energy, CDN, Planck, TTC)
57+
│ ├── governance/ # Review catalog, governance explainer, provenance
58+
│ ├── papers/ # 7 academic submissions (CBD, CoDa, GEOBON, IEEE, etc.)
59+
│ ├── reference/ # Handbooks, math foundations, Ramsar package
60+
│ ├── appendices/ # Code appendix, methodology appendix
61+
│ └── wiki/ # Quick-reference markdown articles
62+
63+
├── src/ # Document build system
2264
│ ├── builders/ # JavaScript docx-js builder scripts
2365
│ │ ├── pillars/ # Pillar document builders
2466
│ │ ├── volumes/ # Volume builders (Triad Synthesis)
2567
│ │ ├── governance/ # Trace and tree builders
2668
│ │ ├── explorations/ # Case study and proof builders
2769
│ │ └── reviews/ # AI review document builders
2870
│ ├── shared/ # Common modules (styles, citations, glossary)
29-
│ └── validation/ # Document validation scripts
30-
├── code/ # Executable code and experiments
31-
│ └── helltest/ # MC-4 detection stress test
32-
│ ├── scripts/ # Python scripts
33-
│ ├── notebooks/ # Jupyter notebooks
34-
│ └── results/ # Generated figures and reports
71+
│ └── validation/ # validate.py (docx integrity checker)
72+
73+
├── code/ # Executable analysis tools
74+
│ ├── analysis/ # Backblaze, energy analysis scripts
75+
│ ├── helltest/ # MC-4 detection stress test (scripts, notebooks, results)
76+
│ └── tools/ # JSX diagnostic utilities
77+
3578
├── notebooks/ # Educational content
36-
│ └── onboarding/ # HUF onboarding notebook series (00-04)
37-
├── data/ # Manifests and integrity data
38-
│ ├── manifests/ # Data, benchmark, and release manifests
39-
│ └── checksums/ # SHA256 document integrity hashes
40-
├── archive/ # Non-canonical historical artifacts
79+
│ └── onboarding/ # HUF onboarding series (00-04)
80+
81+
├── context-books/ # 4 audience editions (general, engineering, physics, sciences)
82+
├── math-books/ # Formal proofs (42 numbered items)
83+
84+
├── process/ # Development record
85+
│ ├── collective-reports/ # AI collective reviews and master packages
86+
│ ├── governance-docs/ # Voting register, decision log, deployment strategy
87+
│ ├── review-traces/ # Version-traced review documents (v1.0-v5.10)
88+
│ └── session-ledgers/ # Per-session JSON records
89+
90+
├── archive/ # Preserved history (nothing deleted, just organized)
91+
│ ├── HUFv4/ # Pre-Phase 3 working directory (version histories)
92+
│ ├── concepts/ # Experimental research (Kardashev, unity-vcore)
93+
│ ├── legacy-builders/ # Superseded builder scripts
94+
│ ├── legacy-generated-docs/ # Superseded .docx outputs
95+
│ ├── legacy-traces/ # Early review traces
96+
│ ├── papers-v1/ # First-draft academic papers
4197
│ ├── prototypes/ # Layout prototypes
42-
│ ├── legacy-generated-docs/ # Superseded .docx (prefer git history)
43-
│ └── legacy-builders/ # Superseded builder scripts
98+
│ ├── ramsar-pre-standardized/ # Early Ramsar materials
99+
│ ├── superseded/ # Explicitly superseded items
100+
│ └── tools-pre-coda/ # Pre-CoDa analyzer versions
101+
44102
└── .github/workflows/ # CI: validate and build
45103
```
46104

105+
★ = New in S016 (April 2026) — EITT discovery materials
106+
47107
## Document Status Legend
48108

49109
| Status | Meaning |
50110
|--------|---------|
51111
| **Canonical** | Current latest version, lives in `docs/` |
52-
| **Superseded** | Replaced by a newer version; preserved in git history |
53-
| **Archival** | Historical artifact; lives in `archive/` if needed |
54-
| **Experimental** | Work in progress; lives in `code/` or working branch |
112+
| **Active Draft** | Work in progress, lives in `drafts/` |
113+
| **Superseded** | Replaced by newer version; preserved in git history or `archive/` |
114+
| **Archival** | Historical artifact; lives in `archive/` with provenance note |
115+
| **Experimental** | Exploratory work; lives in `code/` or `archive/concepts/` |
55116

56117
## Release Policy
57118

58119
Only canonical documents live on `main`. Superseded versions are accessible through git tags, commit history, or GitHub Releases (binary attachments). The `VERSION_MANIFEST.json` tracks what is current.
120+
121+
## For CoDa Researchers
122+
123+
If you're arriving from CoDaWork 2026 or the compositional data analysis community, start with [`START_HERE.md`](START_HERE.md) — the "For CoDa" section points directly to the EITT finding and all supporting materials. The short version: we found that Shannon entropy is near-invariant under geometric-mean temporal decimation, and we think the CoDa community is best positioned to prove (or disprove) it.

0 commit comments

Comments
 (0)