Skip to content

Commit f05c748

Browse files
coda refinements
1 parent 498dece commit f05c748

602 files changed

Lines changed: 57212 additions & 4783 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.

ai-refresh/AI_COLD_START_TEST.json

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -52,7 +52,7 @@
5252
"Q06_formula": "Write the perturbation difference formula and the drift flagging threshold.",
5353
"Q07_eitt_proof": "Name all four EITT proof domains and give one key number from each.",
5454
"Q08_quantum": "What is the Bell test result? How much does it exceed the classical bound?",
55-
"Q09_pll": "How many rules are in the PLL discipline? What does PLL stand for?",
55+
"Q09_cip": "How many rules are in the CIP? What does CIP stand for? What does PLL stand for?",
5656
"Q10_architecture": "What is Keff_fill? What aggregator does it use and at what value of p?"
5757
}
5858
},
@@ -109,7 +109,7 @@
109109
"test_result_2026-04-13": "10/10 questions. Integrity FAIL — caught 2 real bugs (stale hash, wrong char_count). Structure 4/5. 2 attempts (first attempt reviewed architecture instead of taking test)."
110110
},
111111
"grok": {
112-
"history": "Adversarial reviewer. Performed PLL compliance checks, Keff_fill verification, SOPDT robustness derivation.",
112+
"history": "Adversarial reviewer. Performed CIP compliance checks, Keff_fill verification, SOPDT robustness derivation.",
113113
"role": "Adversarial testing. Try to break the refresh. Find what's missing or wrong.",
114114
"access": "Can browse GitHub directly.",
115115
"test_result_2026-04-13": "10/10 questions. Integrity PASS. Structure 5/5. 6 adversarial findings (caught stale manifest paths). 1 attempt. Highest overall score."

ai-refresh/CHATGPT_COLD_START_BRIEFING.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -26,7 +26,7 @@ If you can't access the URL, tell me and I'll paste it.
2626
6. Perturbation difference formula and drift threshold.
2727
7. Four EITT proof domains with one key number each.
2828
8. Bell test result — how much above classical bound?
29-
9. PLL discipline — how many rules, what does PLL stand for?
29+
9. CIP — how many rules in the Compositional Integrity Protocol?
3030
10. Keff_fill — what aggregator, what value of p?
3131

3232
For each: answer, confidence (HIGH/MEDIUM/LOW), pass/fail.

ai-refresh/CHATGPT_VERIFICATION_CHECKLIST.json

Lines changed: 440 additions & 0 deletions
Large diffs are not rendered by default.
Lines changed: 157 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,157 @@
1+
{
2+
"_meta": {
3+
"type": "CODAWORK_2026_PRESENTATION_OUTLINE",
4+
"version": "2.0",
5+
"created": "2026-04-19",
6+
"author": "Peter Higgins / Claude (Anthropic)",
7+
"purpose": "Revised CoDaWork 2026 presentation outline. CoDa-first approach: geometric mean proof as the lead, CoDa bridges as the body, system of systems as the depth. Strategy only — final format TBD after ChatGPT verification review.",
8+
"conference": "11th International Workshop on Compositional Data Analysis (CoDaWork 2026)",
9+
"dates": "1–5 June 2026",
10+
"location": "Coimbra, Portugal",
11+
"submitted_abstract_page": 25,
12+
"submitted_title": "Compositional monitoring of energy-mix drift on the simplex",
13+
"revision_note": "v2.0 pivots from 'energy monitoring with a CoDa seed' to 'CoDa contribution with energy as the flagship domain'. The geometric mean proof is now the headline. All terminology updated to CIP (not PLL), HCDT, correct EITT expansion."
14+
},
15+
16+
"core_message": {
17+
"one_sentence": "The EITT provides the first empirical measurement of the information cost of arithmetic averaging on the simplex — proving the CoDa community's 40-year insistence on geometric mean is not theoretical preference but empirical necessity.",
18+
"elevator_pitch": "We built a temporal diagnostic using standard CoDa tools. It shows that geometric-mean aggregation preserves 99.82% of compositional signal across 341:1 compression, while arithmetic mean destroys 20–40%. We tested it on 75 systems across 44 orders of magnitude. The simplex is the same everywhere.",
19+
"falsifiable_claim": "No existing monitoring framework performs compositional change detection at the carrier level with formal perturbation-based drift measurement."
20+
},
21+
22+
"presentation_structure": {
23+
"_note": "7-beat narrative arc. Target: 15-minute talk. Each beat maps to ~2 minutes.",
24+
25+
"beat_1_the_problem": {
26+
"duration": "2 minutes",
27+
"title": "The Arithmetic Problem",
28+
"content": "Compositional time series are everywhere. Energy agencies, geochemistry labs, epidemiologists — they all average compositions arithmetically. The CoDa community has argued for 40 years that this is wrong. But nobody has measured the cost.",
29+
"visual": "Side-by-side: geometric vs arithmetic average of energy mix data. Show the compositions diverging.",
30+
"coda_connection": "Ground this in Aitchison (1982): 'the geometric mean is the natural average on the simplex'. Cite Pawlowsky-Glahn et al. (2015). Start by crediting the community."
31+
},
32+
33+
"beat_2_the_tool": {
34+
"duration": "2 minutes",
35+
"title": "The Entropy-Invariant Time Transformer",
36+
"content": "Define EITT: Shannon entropy of compositional data measured at multiple temporal resolutions under geometric-mean decimation (the Aitchison barycenter). The invariance — or its absence — quantifies whether temporal aggregation preserves compositional structure.",
37+
"formula": "x̄_G = C(∏ₜ x(t)^(1/k)) — the Aitchison barycenter. H(x̄_G) ≈ H(x) when the simplex geometry is respected.",
38+
"visual": "The Higgins Decomposition 10-step pipeline diagram. Emphasise: every step uses standard CoDa tools.",
39+
"coda_connection": "CLR transform (Step 3), Aitchison variance (Step 4), geometric-mean decimation (Step 6) — all standard CoDa. The novelty is assembling them into a temporal diagnostic."
40+
},
41+
42+
"beat_3_the_proof": {
43+
"duration": "3 minutes (longest beat — the headline)",
44+
"title": "The Geometric Mean Proof",
45+
"content": "Present the numbers. European daily electricity prices: 0.18% entropy variation at 341:1 compression (geometric). NGFS Phase 4 scenarios: geometric 5yr 1.8% vs arithmetic 5yr 14.2%. Geometric 10yr 2.3% vs arithmetic 10yr 21.7%. Up to 41% destruction at arithmetic 10yr.",
46+
"key_numbers": {
47+
"geometric_341_to_1": "0.18%",
48+
"geometric_5yr": "1.8%",
49+
"geometric_10yr": "2.3%",
50+
"arithmetic_5yr": "14.2%",
51+
"arithmetic_10yr": "21.7%",
52+
"ratio": "Arithmetic destroys ~10× more signal than geometric at every scale"
53+
},
54+
"visual": "Bar chart: geometric vs arithmetic entropy variation side by side, across compression levels. The geometric bars are barely visible; the arithmetic bars are massive.",
55+
"coda_connection": "This is the empirical ammunition. Aitchison was right. Pawlowsky-Glahn was right. Now there's a number."
56+
},
57+
58+
"beat_4_drift_detection": {
59+
"duration": "2 minutes",
60+
"title": "Perturbation-Based Drift Detection",
61+
"content": "Show the EMBER energy monitoring results. Germany: nuclear shutdown flagged at dₐ=9.07 (2023–2024). Japan: Fukushima aftermath flagged at dₐ=9.05 (2013–2014, NOT 2011–2012 — the annual resolution matters). UK: three distinct drift events over 25 years. Each flag has a known physical cause.",
62+
"formula": "Drift flag when dₐ(t→t+1) > μ + 2σ — self-calibrated per country.",
63+
"visual": "Aitchison distance time series for Germany, Japan, UK with drift flags marked and annotated with causes.",
64+
"coda_connection": "Perturbation difference is Aitchison (1986). Drift detection via Aitchison distance is standard CoDa. The contribution is applying it to real-world policy data with interpretable results."
65+
},
66+
67+
"beat_5_geochemistry_validation": {
68+
"duration": "2 minutes",
69+
"title": "Validation in CoDa's Home Territory",
70+
"content": "40,666 real igneous rock samples. 8 major oxides. 37 of 39 TAS rock types pass EITT. Foidite fails at PR=32% — legitimate boundary condition (deep mantle phase chaos destroys temporal autocorrelation). The variation matrix identifies SiO₂/MgO as the dominant driver, matching known petrology.",
71+
"visual": "Pass rate by TAS type. Highlight Foidite as the informative failure.",
72+
"coda_connection": "Geochemistry IS CoDa. Aitchison himself used rock compositions. This validates EITT in the community's own backyard. The Foidite failure defines a boundary condition the community should find scientifically interesting."
73+
},
74+
75+
"beat_6_the_vertex_theorem": {
76+
"duration": "2 minutes",
77+
"title": "A New Dynamic on the Simplex",
78+
"content": "The Vertex Theorem: d(σ²_A)/dt = (2/D)Σ clr_i · clr_i' = 0 at dynamic equilibrium, where clr(t*) ⊥ clr'(t*). This is an exact identity from the chain rule applied to Aitchison variance. It gives a dynamic equilibrium condition — the point where composition restructures but stress is stationary. Tested on 30 domains: 28 show parabolic structure with R² > 0.6.",
79+
"formula": "d(σ²_A)/dt = (2/D)(clr · clr') = 0 ⟺ clr(t*) ⊥ clr'(t*)",
80+
"visual": "σ²_A trajectory with parabola fit for 3–4 domains (room acoustics R²=0.998, energy, geochemistry, stellar nucleosynthesis). Show the vertex as the lock point.",
81+
"coda_connection": "This is a CoDa dynamics result. CoDa has been mostly static (cross-sectional). The Vertex Theorem extends it to temporal analysis using existing tools (CLR, σ²_A). CoDa scientists will immediately understand the CLR orthogonality condition."
82+
},
83+
84+
"beat_7_what_this_means": {
85+
"duration": "2 minutes",
86+
"title": "What This Means for CoDa",
87+
"content": "Summary: (1) Geometric mean proof — empirical ammunition for the community. (2) Temporal extension — EITT + Vertex Theorem take CoDa from static to dynamic. (3) New domains — nuclear physics, particle physics, gravitational waves. 75 systems, 44 orders of magnitude. (4) Operational chain — the Higgins Decomposition, 10 steps, all standard CoDa tools. The Compositional Integrity Protocol (CIP) — 6 rules — keeps it clean.",
88+
"visual": "System-of-systems map showing the 12 experiments across the scale range. Simple version — just domain names and scale, no internal jargon.",
89+
"coda_connection": "End where you started: CoDa tools work. They work in time. They work across domains. The simplex is the same everywhere. And now we can measure the cost of ignoring it.",
90+
"open_question": "For the community — what other temporal compositional datasets would benefit from this analysis? (Invite collaboration.)"
91+
}
92+
},
93+
94+
"strategic_neighbours": {
95+
"_note": "Other CoDaWork 2026 presenters whose work connects to ours. Potential allies.",
96+
"Narayana_p29": {
97+
"title": "COPD microbiome perturbation",
98+
"connection": "Closest methodological ally — perturbation-based analysis of compositional time series in biology",
99+
"bridge": "Our drift detection maps directly to their disease progression monitoring"
100+
},
101+
"Ascari_p9": {
102+
"title": "Energy-mix clustering with zeros",
103+
"connection": "Same domain (energy mix), complementary approach (clustering vs temporal monitoring)",
104+
"bridge": "Their zero-handling technique could address our multiplicative delta zero replacement"
105+
},
106+
"Vega_Baquero_p57": {
107+
"title": "Portfolio allocation as composition",
108+
"connection": "Financial compositions — portfolio rebalancing is the same temporal problem as energy transition",
109+
"bridge": "Best EITT bridge — financial data would be a compelling fourth flagship domain"
110+
},
111+
"Kanjiradan_Veetil_p26": {
112+
"title": "Cancer mortality time series — VAR forecasting",
113+
"connection": "Compositional time series in health data — needs drift detection",
114+
"bridge": "Our temporal tools could extend their VAR approach with CoDa-native change detection"
115+
}
116+
},
117+
118+
"what_to_avoid": {
119+
"terminology_traps": [
120+
"Do NOT say 'PLL' to a CoDa audience — they will hear 'Phase-Locked Loop' and be confused about the engineering context. Say 'σ²_A parabola' or 'compositional stress trajectory'.",
121+
"Do NOT use 'HCDT' — CoDa people won't know it. Describe the toolkit by its components.",
122+
"Do NOT use 'DADC/DADI/ADAC' — these are acoustic engineering heritage terms. Say 'contamination analysis' or 'residual decomposition'.",
123+
"Do NOT use 'BTL' or 'Rogue Wave Audio' — engineering origin story is not relevant to CoDa.",
124+
"Do NOT use 'HUF-GOV' or 'open-loop doctrine' — governance concepts are internal to HUF.",
125+
"DO use: CLR, σ²_A, Aitchison distance, perturbation, simplex, closure, log-ratio, geometric mean, Shannon entropy — all CoDa/information theory native terms.",
126+
"DO use: CIP (Compositional Integrity Protocol) — it's new but the concept (analysis rules) is clear."
127+
],
128+
"content_traps": [
129+
"Do NOT lead with quantum analogies — the Bell test will make CoDa scientists suspicious of overclaiming",
130+
"Do NOT present the nuclear staircase — it's exploratory and CoDa-irrelevant",
131+
"Do NOT present dark matter or gravitational wave results as the lead — they are depth, not headline",
132+
"Do NOT claim universality — say 'tested across 75 systems' not 'proven universal'",
133+
"Do NOT present the 93% bound as a theorem — it's an observed pattern (L1)",
134+
"DO acknowledge failures honestly — Foidite PR=32%, Estonia 8.43%, adversarial boundary at temporal autocorrelation",
135+
"DO present the Contamination Doctrine — 'prefer a weaker but cleaner claim over a stronger contaminated one'. CoDa scientists will respect this."
136+
]
137+
},
138+
139+
"depth_available_on_request": {
140+
"_note": "If someone asks 'have you tested this beyond energy and geochemistry?', this is the depth you can reveal.",
141+
"nuclear_physics": "SEMF binding energy as 4-part composition. 3,554 nuclides. Shell effects detected via σ²_A perturbations.",
142+
"particle_physics": "CKM and PMNS mixing matrices as simplex compositions. QGP freeze-out: EITT at 0.27% across the nuclear-quark phase boundary.",
143+
"gravitational_waves": "GW150914 mass budget as 4-part composition. Lowest entropy in all 75 systems (H/Hmax=0.175).",
144+
"full_inventory": "75 systems, 44 orders of magnitude, 12 experiments, 100% coverage. The Higgins Decomposition applied uniformly to every system."
145+
},
146+
147+
"maps_to_system_of_systems": {
148+
"_note": "How the CoDa presentation maps to the internal HUF system. For Peter's reference.",
149+
"beat_1": "Background — CoDa community context",
150+
"beat_2": "EITT definition — maps to FAST_REFRESH canonical_names.EITT",
151+
"beat_3": "Geometric mean proof — maps to FAST_REFRESH geometric_mean_empirical_proof",
152+
"beat_4": "EMBER results — maps to EXP-01, EXP-02, FAST_REFRESH ember_monitoring",
153+
"beat_5": "Geochemistry — maps to EXP-05, EXP-05b",
154+
"beat_6": "Vertex Theorem — maps to FAST_REFRESH pll_parabola_discovery.vertex_theorem",
155+
"beat_7": "System of systems — maps to FAST_REFRESH experiment_chain"
156+
}
157+
}

0 commit comments

Comments
 (0)