weekly indicator two axes immutable archive methodology 2026-07-10.v1 artifacts as of 2026-08-04

AI supply × demand, on one honest clock.

A slow fundamental read of where the AI build-out sits: demand intensity from usage, adoption, capex, and announced projects; supply tightness from component-cycle health, chokepoints, and capacity relief. Every print keeps its inputs, knowledge dates, coverage, and methodology hash.

no forward print · validation gate fired

The retrospective proxy pack cannot authorize publication.

This pack is explicitly marked preliminary_proxy. Its values were assembled after the episodes and cannot masquerade as point-in-time evidence. Numerically, committed_pipeline correlates 96.3% with the heat index—above the pre-declared 95% redundancy limit. The weekly recorder keeps collecting inputs, but a forward print requires a separately reviewed point-in-time evidence manifest and every numerical gate.

Backtest-first launch

Three retrospective proxy episodes

Select a point to inspect its full calculation below.

demand supply diffusion agreement

Calculation audit

Reproduce any point from raw input to index

Pick a plotted date. Every row exposes the raw observation, frozen normalization anchors, exact transform, effective weight, and contribution to the axis. Axis = Σ(score × frozen weight × source coverage) ÷ included coverage.

plotted date

Demand

59.1

Supply

37.8

Diffusion

48.5

Agreement

78.7

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.18 trailing-vs-prior fractional growth observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.18 − 0.0) ÷ (1.0 − 0.0) = 59.0	59.0	20%	11.8
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.08 median model-download growth observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.08 − 0.0) ÷ (0.5 − 0.0) = 58.0	58.0	20%	11.6
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.04 median blended price change observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.04) ÷ (0.0 − -0.5) = 54.0	54.0	20%	10.8
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.28 TTM capex YoY growth observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.28 − 0.1) ÷ (0.6 − 0.1) = 68.0	68.0	20%	13.6
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	8.5 average announced GW arriving per year over the next three years observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (8.5 − 7.5) ÷ (15.0 − 7.5) = 56.7	56.7	20%	11.333333

Axis equation

(59 × 0.2 × 1 + 58 × 0.2 × 1 + 54 × 0.2 × 1 + 68 × 0.2 × 1 + 56.666666666666664 × 0.2 × 1) ÷ 1 = 59.13

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	BALANCED categorical gate state observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 BALANCED maps directly to 50.0	50.0	33.333333%	16.666667
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.44 0-1 measured/curated tension observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 × (0.44 − 0.2) ÷ (0.5 − 0.2) = 40.0	40.0	33.333333%	13.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.34 share of committed demand landing inside the relief window observed 2017-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 × (0.34 − 0.2) ÷ (0.5 − 0.2) = 23.3	23.3	33.333333%	7.777778

Axis equation

(50 × 0.3333333333333333 × 1 + 40 × 0.3333333333333333 × 1 + 23.333333333333336 × 0.3333333333333333 × 1) ÷ 1 = 37.78

Diffusion index

(59.1 + 37.8) ÷ 2 = 48.5

Agreement score

100 − |59.1 − 37.8| = 78.7

Selected immutable print

object replays/preliminary-proxy-v1/prints/2017-12-31.json sha256 e715f20bdd0854f4c97eeb4410eae8593410f605f398a23357633d3a1699fdd3

Demand

66.5

Supply

78.6

Diffusion

72.5

Agreement

87.9

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.31 trailing-vs-prior fractional growth observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.31 − 0.0) ÷ (1.0 − 0.0) = 65.5	65.5	20%	13.1
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.2 median model-download growth observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.2 − 0.0) ÷ (0.5 − 0.0) = 70.0	70.0	20%	14
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.02 median blended price change observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.02) ÷ (0.0 − -0.5) = 52.0	52.0	20%	10.4
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.35 TTM capex YoY growth observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.35 − 0.1) ÷ (0.6 − 0.1) = 75.0	75.0	20%	15
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	10.5 average announced GW arriving per year over the next three years observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (10.5 − 7.5) ÷ (15.0 − 7.5) = 70.0	70.0	20%	14

Axis equation

(65.5 × 0.2 × 1 + 70 × 0.2 × 1 + 52 × 0.2 × 1 + 75 × 0.2 × 1 + 70 × 0.2 × 1) ÷ 1 = 66.50

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.68 0-1 measured/curated tension observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.68 − 0.5) ÷ (0.85 − 0.5) = 75.7	75.7	33.333333%	25.238095
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.66 share of committed demand landing inside the relief window observed 2018-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.66 − 0.5) ÷ (0.9 − 0.5) = 70.0	70.0	33.333333%	23.333333

Axis equation

(90 × 0.3333333333333333 × 1 + 75.71428571428572 × 0.3333333333333333 × 1 + 70 × 0.3333333333333333 × 1) ÷ 1 = 78.57

Diffusion index

(66.5 + 78.6) ÷ 2 = 72.5

Agreement score

100 − |66.5 − 78.6| = 87.9

Selected immutable print

object replays/preliminary-proxy-v1/prints/2018-03-31.json sha256 ffbd711439019cd0efd7088ae8ea0ea1314c97f5b7bdada331bde402d992feab

Demand

67.3

Supply

95.8

Diffusion

81.6

Agreement

71.5

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.37 trailing-vs-prior fractional growth observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.37 − 0.0) ÷ (1.0 − 0.0) = 68.5	68.5	20%	13.7
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.27 median model-download growth observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.27 − 0.0) ÷ (0.5 − 0.0) = 77.0	77.0	20%	15.4
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	0.04 median blended price change observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 × (0.1 − 0.04) ÷ (0.1 − 0.0) = 30.0	30.0	20%	6
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.41 TTM capex YoY growth observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.41 − 0.1) ÷ (0.6 − 0.1) = 81.0	81.0	20%	16.2
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	12.0 average announced GW arriving per year over the next three years observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (12.0 − 7.5) ÷ (15.0 − 7.5) = 80.0	80.0	20%	16

Axis equation

(68.5 × 0.2 × 1 + 77 × 0.2 × 1 + 30.000000000000004 × 0.2 × 1 + 81 × 0.2 × 1 + 80 × 0.2 × 1) ÷ 1 = 67.30

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.91 0-1 measured/curated tension observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.91 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.88 share of committed demand landing inside the relief window observed 2018-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.88 − 0.5) ÷ (0.9 − 0.5) = 97.5	97.5	33.333333%	32.5

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 97.5 × 0.3333333333333333 × 1) ÷ 1 = 95.83

Diffusion index

(67.3 + 95.8) ÷ 2 = 81.6

Agreement score

100 − |67.3 − 95.8| = 71.5

Selected immutable print

object replays/preliminary-proxy-v1/prints/2018-06-30.json sha256 d14e5f49bfe19655811909a35e534db7cb89d444e8aedff1bfdefb82c874f3b0

Demand

40.1

Supply

90.0

Diffusion

65.1

Agreement

50.1

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	-0.02 trailing-vs-prior fractional growth observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 × (-0.02 − -0.2) ÷ (0.0 − -0.2) = 45.0	45.0	20%	9
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	-0.06 median model-download growth observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 × (-0.06 − -0.2) ÷ (0.0 − -0.2) = 35.0	35.0	20%	7
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	0.08 median blended price change observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 × (0.1 − 0.08) ÷ (0.1 − 0.0) = 10.0	10.0	20%	2
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.24 TTM capex YoY growth observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.24 − 0.1) ÷ (0.6 − 0.1) = 64.0	64.0	20%	12.8
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	7.0 average announced GW arriving per year over the next three years observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 × (7.0 − 0.0) ÷ (7.5 − 0.0) = 46.7	46.7	20%	9.333333

Axis equation

(45.00000000000001 × 0.2 × 1 + 35 × 0.2 × 1 + 10.000000000000002 × 0.2 × 1 + 64 × 0.2 × 1 + 46.666666666666664 × 0.2 × 1) ÷ 1 = 40.13

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.88 0-1 measured/curated tension observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.88 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.74 share of committed demand landing inside the relief window observed 2018-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.74 − 0.5) ÷ (0.9 − 0.5) = 80.0	80.0	33.333333%	26.666667

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 80 × 0.3333333333333333 × 1) ÷ 1 = 90.00

Diffusion index

(40.1 + 90.0) ÷ 2 = 65.1

Agreement score

100 − |40.1 − 90.0| = 50.1

Selected immutable print

object replays/preliminary-proxy-v1/prints/2018-09-30.json sha256 602bbd2172b504e0cc4601eb05e9e27fa25de85055fff5dbaf9a7833af5f4d28

Demand

29.1

Supply

26.4

Diffusion

27.8

Agreement

97.3

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	-0.16 trailing-vs-prior fractional growth observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 × (-0.16 − -0.2) ÷ (0.0 − -0.2) = 10.0	10.0	20%	2
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	-0.14 median model-download growth observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 × (-0.14 − -0.2) ÷ (0.0 − -0.2) = 15.0	15.0	20%	3
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.18 median blended price change observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.18) ÷ (0.0 − -0.5) = 68.0	68.0	20%	13.6
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.03 TTM capex YoY growth observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 × (0.03 − -0.1) ÷ (0.1 − -0.1) = 32.5	32.5	20%	6.5
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	3.0 average announced GW arriving per year over the next three years observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 × (3.0 − 0.0) ÷ (7.5 − 0.0) = 20.0	20.0	20%	4

Axis equation

(10.000000000000002 × 0.2 × 1 + 15 × 0.2 × 1 + 68 × 0.2 × 1 + 32.5 × 0.2 × 1 + 20 × 0.2 × 1) ÷ 1 = 29.10

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	BREAKING categorical gate state observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 BREAKING maps directly to 10.0	10.0	33.333333%	3.333333
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.53 0-1 measured/curated tension observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.53 − 0.5) ÷ (0.85 − 0.5) = 54.3	54.3	33.333333%	18.095238
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.29 share of committed demand landing inside the relief window observed 2018-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 × (0.29 − 0.2) ÷ (0.5 − 0.2) = 15.0	15.0	33.333333%	5

Axis equation

(10 × 0.3333333333333333 × 1 + 54.28571428571429 × 0.3333333333333333 × 1 + 14.999999999999995 × 0.3333333333333333 × 1) ÷ 1 = 26.43

Diffusion index

(29.1 + 26.4) ÷ 2 = 27.8

Agreement score

100 − |29.1 − 26.4| = 97.3

Selected immutable print

object replays/preliminary-proxy-v1/prints/2018-12-31.json sha256 638ccf7d06d1363d859bb84ef3eb2ba2fc2d5d4efaeebd01925e988b37bbf0f9

Demand

82.9

Supply

51.1

Diffusion

67.0

Agreement

68.2

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.66 trailing-vs-prior fractional growth observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.66 − 0.0) ÷ (1.0 − 0.0) = 83.0	83.0	20%	16.6
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.42 median model-download growth observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.42 − 0.0) ÷ (0.5 − 0.0) = 92.0	92.0	20%	18.4
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.24 median blended price change observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.24) ÷ (0.0 − -0.5) = 74.0	74.0	20%	14.8
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.49 TTM capex YoY growth observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.49 − 0.1) ÷ (0.6 − 0.1) = 89.0	89.0	20%	17.8
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	11.5 average announced GW arriving per year over the next three years observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (11.5 − 7.5) ÷ (15.0 − 7.5) = 76.7	76.7	20%	15.333333

Axis equation

(83 × 0.2 × 1 + 92 × 0.2 × 1 + 74 × 0.2 × 1 + 89 × 0.2 × 1 + 76.66666666666667 × 0.2 × 1) ÷ 1 = 82.93

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	BALANCED categorical gate state observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 BALANCED maps directly to 50.0	50.0	33.333333%	16.666667
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.57 0-1 measured/curated tension observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.57 − 0.5) ÷ (0.85 − 0.5) = 60.0	60.0	33.333333%	20
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.46 share of committed demand landing inside the relief window observed 2020-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 × (0.46 − 0.2) ÷ (0.5 − 0.2) = 43.3	43.3	33.333333%	14.444444

Axis equation

(50 × 0.3333333333333333 × 1 + 59.99999999999999 × 0.3333333333333333 × 1 + 43.333333333333336 × 0.3333333333333333 × 1) ÷ 1 = 51.11

Diffusion index

(82.9 + 51.1) ÷ 2 = 67.0

Agreement score

100 − |82.9 − 51.1| = 68.2

Selected immutable print

object replays/preliminary-proxy-v1/prints/2020-09-30.json sha256 1418686da2b7ef1c7f214388a8ab1ce3be6f841680d0799574b9c9a7ecc807c6

Demand

90.9

Supply

81.7

Diffusion

86.3

Agreement

90.8

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.81 trailing-vs-prior fractional growth observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.81 − 0.0) ÷ (1.0 − 0.0) = 90.5	90.5	20%	18.1
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.46 median model-download growth observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.46 − 0.0) ÷ (0.5 − 0.0) = 96.0	96.0	20%	19.2
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.31 median blended price change observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.31) ÷ (0.0 − -0.5) = 81.0	81.0	20%	16.2
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.57 TTM capex YoY growth observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.57 − 0.1) ÷ (0.6 − 0.1) = 97.0	97.0	20%	19.4
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	13.5 average announced GW arriving per year over the next three years observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (13.5 − 7.5) ÷ (15.0 − 7.5) = 90.0	90.0	20%	18

Axis equation

(90.5 × 0.2 × 1 + 96 × 0.2 × 1 + 81 × 0.2 × 1 + 97 × 0.2 × 1 + 90 × 0.2 × 1) ÷ 1 = 90.90

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.72 0-1 measured/curated tension observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.72 − 0.5) ÷ (0.85 − 0.5) = 81.4	81.4	33.333333%	27.142857
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.69 share of committed demand landing inside the relief window observed 2020-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.69 − 0.5) ÷ (0.9 − 0.5) = 73.7	73.7	33.333333%	24.583333

Axis equation

(90 × 0.3333333333333333 × 1 + 81.42857142857143 × 0.3333333333333333 × 1 + 73.74999999999999 × 0.3333333333333333 × 1) ÷ 1 = 81.73

Diffusion index

(90.9 + 81.7) ÷ 2 = 86.3

Agreement score

100 − |90.9 − 81.7| = 90.8

Selected immutable print

object replays/preliminary-proxy-v1/prints/2020-12-31.json sha256 1aedc390a36feaf71ed59ff27e41f037bdded7829274f2294f9cfb70e210b96a

Demand

96.4

Supply

96.7

Diffusion

96.5

Agreement

99.7

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.92 trailing-vs-prior fractional growth observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.92 − 0.0) ÷ (1.0 − 0.0) = 96.0	96.0	20%	19.2
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.48 median model-download growth observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.48 − 0.0) ÷ (0.5 − 0.0) = 98.0	98.0	20%	19.6
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.38 median blended price change observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.38) ÷ (0.0 − -0.5) = 88.0	88.0	20%	17.6
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.63 TTM capex YoY growth observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 0.63 ≥ hot 0.6 → 100.0 (clipped)	100.0	20%	20
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	15.5 average announced GW arriving per year over the next three years observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 15.5 ≥ hot 15.0 → 100.0 (clipped)	100.0	20%	20

Axis equation

(96 × 0.2 × 1 + 98 × 0.2 × 1 + 88 × 0.2 × 1 + 100 × 0.2 × 1 + 100 × 0.2 × 1) ÷ 1 = 96.40

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.94 0-1 measured/curated tension observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.94 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.93 share of committed demand landing inside the relief window observed 2021-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 0.93 ≥ hot 0.9 → 100.0 (clipped)	100.0	33.333333%	33.333333

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1) ÷ 1 = 96.67

Diffusion index

(96.4 + 96.7) ÷ 2 = 96.5

Agreement score

100 − |96.4 − 96.7| = 99.7

Selected immutable print

object replays/preliminary-proxy-v1/prints/2021-03-31.json sha256 bc96f036420971ccc875d218f052e10167439d77fb7db2e402cc37602e184127

Demand

84.3

Supply

95.0

Diffusion

89.6

Agreement

89.3

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.54 trailing-vs-prior fractional growth observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.54 − 0.0) ÷ (1.0 − 0.0) = 77.0	77.0	20%	15.4
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.31 median model-download growth observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.31 − 0.0) ÷ (0.5 − 0.0) = 81.0	81.0	20%	16.2
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.29 median blended price change observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.29) ÷ (0.0 − -0.5) = 79.0	79.0	20%	15.8
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.51 TTM capex YoY growth observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.51 − 0.1) ÷ (0.6 − 0.1) = 91.0	91.0	20%	18.2
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	14.0 average announced GW arriving per year over the next three years observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (14.0 − 7.5) ÷ (15.0 − 7.5) = 93.3	93.3	20%	18.666667

Axis equation

(77 × 0.2 × 1 + 81 × 0.2 × 1 + 79 × 0.2 × 1 + 91 × 0.2 × 1 + 93.33333333333334 × 0.2 × 1) ÷ 1 = 84.27

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.9 0-1 measured/curated tension observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.9 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.86 share of committed demand landing inside the relief window observed 2021-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.86 − 0.5) ÷ (0.9 − 0.5) = 95.0	95.0	33.333333%	31.666667

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 95 × 0.3333333333333333 × 1) ÷ 1 = 95.00

Diffusion index

(84.3 + 95.0) ÷ 2 = 89.6

Agreement score

100 − |84.3 − 95.0| = 89.3

Selected immutable print

object replays/preliminary-proxy-v1/prints/2021-06-30.json sha256 0644a4b9ebfe531805795ef7b1dc67db9439adfdbbf0425320d87bbdef4643cd

Demand

65.4

Supply

78.6

Diffusion

72.0

Agreement

86.8

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.21 trailing-vs-prior fractional growth observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.21 − 0.0) ÷ (1.0 − 0.0) = 60.5	60.5	20%	12.1
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.12 median model-download growth observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.12 − 0.0) ÷ (0.5 − 0.0) = 62.0	62.0	20%	12.4
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.14 median blended price change observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.14) ÷ (0.0 − -0.5) = 64.0	64.0	20%	12.8
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.34 TTM capex YoY growth observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.34 − 0.1) ÷ (0.6 − 0.1) = 74.0	74.0	20%	14.8
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	10.0 average announced GW arriving per year over the next three years observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (10.0 − 7.5) ÷ (15.0 − 7.5) = 66.7	66.7	20%	13.333333

Axis equation

(60.5 × 0.2 × 1 + 62 × 0.2 × 1 + 64 × 0.2 × 1 + 74 × 0.2 × 1 + 66.66666666666667 × 0.2 × 1) ÷ 1 = 65.43

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	EARLY-CRACK categorical gate state observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 EARLY-CRACK maps directly to 55.0	55.0	33.333333%	18.333333
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.83 0-1 measured/curated tension observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.83 − 0.5) ÷ (0.85 − 0.5) = 97.1	97.1	33.333333%	32.380952
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.77 share of committed demand landing inside the relief window observed 2021-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.77 − 0.5) ÷ (0.9 − 0.5) = 83.8	83.8	33.333333%	27.916667

Axis equation

(55 × 0.3333333333333333 × 1 + 97.14285714285714 × 0.3333333333333333 × 1 + 83.75 × 0.3333333333333333 × 1) ÷ 1 = 78.63

Diffusion index

(65.4 + 78.6) ÷ 2 = 72.0

Agreement score

100 − |65.4 − 78.6| = 86.8

Selected immutable print

object replays/preliminary-proxy-v1/prints/2021-09-30.json sha256 cd278121f17646347ccead9211ed06bb7a842a5190143c3267b5f7b1c51f9d34

Demand

40.6

Supply

20.6

Diffusion

30.6

Agreement

79.9

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	-0.09 trailing-vs-prior fractional growth observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 × (-0.09 − -0.2) ÷ (0.0 − -0.2) = 27.5	27.5	20%	5.5
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.03 median model-download growth observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.03 − 0.0) ÷ (0.5 − 0.0) = 53.0	53.0	20%	10.6
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.16 median blended price change observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.16) ÷ (0.0 − -0.5) = 66.0	66.0	20%	13.2
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.02 TTM capex YoY growth observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 × (0.02 − -0.1) ÷ (0.1 − -0.1) = 30.0	30.0	20%	6
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	4.0 average announced GW arriving per year over the next three years observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 × (4.0 − 0.0) ÷ (7.5 − 0.0) = 26.7	26.7	20%	5.333333

Axis equation

(27.500000000000004 × 0.2 × 1 + 53 × 0.2 × 1 + 66 × 0.2 × 1 + 30.000000000000004 × 0.2 × 1 + 26.666666666666668 × 0.2 × 1) ÷ 1 = 40.63

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	BREAKING categorical gate state observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 BREAKING maps directly to 10.0	10.0	33.333333%	3.333333
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.38 0-1 measured/curated tension observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 × (0.38 − 0.2) ÷ (0.5 − 0.2) = 30.0	30.0	33.333333%	10
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.33 share of committed demand landing inside the relief window observed 2022-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 × (0.33 − 0.2) ÷ (0.5 − 0.2) = 21.7	21.7	33.333333%	7.222222

Axis equation

(10 × 0.3333333333333333 × 1 + 30 × 0.3333333333333333 × 1 + 21.666666666666668 × 0.3333333333333333 × 1) ÷ 1 = 20.56

Diffusion index

(40.6 + 20.6) ÷ 2 = 30.6

Agreement score

100 − |40.6 − 20.6| = 79.9

Selected immutable print

object replays/preliminary-proxy-v1/prints/2022-12-31.json sha256 8cccdb288fa50ca74d396f02d3bac0aa30405b47d6ac5bf43069ff07ef0e9de0

Demand

67.8

Supply

51.4

Diffusion

59.6

Agreement

83.5

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.33 trailing-vs-prior fractional growth observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.33 − 0.0) ÷ (1.0 − 0.0) = 66.5	66.5	20%	13.3
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.19 median model-download growth observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.19 − 0.0) ÷ (0.5 − 0.0) = 69.0	69.0	20%	13.8
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.28 median blended price change observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.28) ÷ (0.0 − -0.5) = 78.0	78.0	20%	15.6
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.29 TTM capex YoY growth observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.29 − 0.1) ÷ (0.6 − 0.1) = 69.0	69.0	20%	13.8
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	8.5 average announced GW arriving per year over the next three years observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (8.5 − 7.5) ÷ (15.0 − 7.5) = 56.7	56.7	20%	11.333333

Axis equation

(66.5 × 0.2 × 1 + 69 × 0.2 × 1 + 78 × 0.2 × 1 + 69 × 0.2 × 1 + 56.666666666666664 × 0.2 × 1) ÷ 1 = 67.83

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	BALANCED categorical gate state observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 BALANCED maps directly to 50.0	50.0	33.333333%	16.666667
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.54 0-1 measured/curated tension observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 50 + 50 × (0.54 − 0.5) ÷ (0.85 − 0.5) = 55.7	55.7	33.333333%	18.571429
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.49 share of committed demand landing inside the relief window observed 2023-03-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 × (0.49 − 0.2) ÷ (0.5 − 0.2) = 48.3	48.3	33.333333%	16.111111

Axis equation

(50 × 0.3333333333333333 × 1 + 55.71428571428572 × 0.3333333333333333 × 1 + 48.33333333333333 × 0.3333333333333333 × 1) ÷ 1 = 51.35

Diffusion index

(67.8 + 51.4) ÷ 2 = 59.6

Agreement score

100 − |67.8 − 51.4| = 83.5

Selected immutable print

object replays/preliminary-proxy-v1/prints/2023-03-31.json sha256 ba8ad1f7f6f4443402d2fa7053f1f421c8912d60be35ca269021d89817d4371c

Demand

94.0

Supply

92.9

Diffusion

93.5

Agreement

98.9

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.97 trailing-vs-prior fractional growth observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.97 − 0.0) ÷ (1.0 − 0.0) = 98.5	98.5	20%	19.7
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.44 median model-download growth observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.44 − 0.0) ÷ (0.5 − 0.0) = 94.0	94.0	20%	18.8
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.46 median blended price change observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.46) ÷ (0.0 − -0.5) = 96.0	96.0	20%	19.2
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.55 TTM capex YoY growth observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.55 − 0.1) ÷ (0.6 − 0.1) = 95.0	95.0	20%	19
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	13.0 average announced GW arriving per year over the next three years observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (13.0 − 7.5) ÷ (15.0 − 7.5) = 86.7	86.7	20%	17.333333

Axis equation

(98.5 × 0.2 × 1 + 94 × 0.2 × 1 + 96 × 0.2 × 1 + 95 × 0.2 × 1 + 86.66666666666666 × 0.2 × 1) ÷ 1 = 94.03

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.86 0-1 measured/curated tension observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.86 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.81 share of committed demand landing inside the relief window observed 2023-06-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 50 + 50 × (0.81 − 0.5) ÷ (0.9 − 0.5) = 88.8	88.8	33.333333%	29.583333

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 88.75 × 0.3333333333333333 × 1) ÷ 1 = 92.92

Diffusion index

(94.0 + 92.9) ÷ 2 = 93.5

Agreement score

100 − |94.0 − 92.9| = 98.9

Selected immutable print

object replays/preliminary-proxy-v1/prints/2023-06-30.json sha256 db32c8ce0882abfb371647c8c911a8241a1c3949c4e1d0d7e281180a8b249593

Demand

92.3

Supply

96.7

Diffusion

94.5

Agreement

95.7

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	1.18 trailing-vs-prior fractional growth observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 1.18 ≥ hot 1.0 → 100.0 (clipped)	100.0	20%	20
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.36 median model-download growth observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.36 − 0.0) ÷ (0.5 − 0.0) = 86.0	86.0	20%	17.2
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.43 median blended price change observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.43) ÷ (0.0 − -0.5) = 93.0	93.0	20%	18.6
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.46 TTM capex YoY growth observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.46 − 0.1) ÷ (0.6 − 0.1) = 86.0	86.0	20%	17.2
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	14.5 average announced GW arriving per year over the next three years observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 50 + 50 × (14.5 − 7.5) ÷ (15.0 − 7.5) = 96.7	96.7	20%	19.333333

Axis equation

(100 × 0.2 × 1 + 86 × 0.2 × 1 + 93 × 0.2 × 1 + 86 × 0.2 × 1 + 96.66666666666666 × 0.2 × 1) ÷ 1 = 92.33

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.96 0-1 measured/curated tension observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.96 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.9 share of committed demand landing inside the relief window observed 2023-09-30 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 0.9 ≥ hot 0.9 → 100.0 (clipped)	100.0	33.333333%	33.333333

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1) ÷ 1 = 96.67

Diffusion index

(92.3 + 96.7) ÷ 2 = 94.5

Agreement score

100 − |92.3 − 96.7| = 95.7

Selected immutable print

object replays/preliminary-proxy-v1/prints/2023-09-30.json sha256 8c459114d1d0fdeeb0be5b2c658e94365a2d830faece6b56cc45b0214ed7bfce

Demand

94.2

Supply

96.7

Diffusion

95.4

Agreement

97.5

Demand intensity

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Inference usage / spend growth `usage_growth` OpenRouter token-spend tracer; historical replays use the closest dated end-demand usage proxy.	0.88 trailing-vs-prior fractional growth observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 1.0 → 100 50 + 50 × (0.88 − 0.0) ÷ (1.0 − 0.0) = 94.0	94.0	20%	18.8
Open-model adoption breadth `adoption_breadth` Median download growth across a frozen Hugging Face model panel; replay uses a dated adoption-breadth proxy.	0.49 median model-download growth observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.2 → 0 · neutral 0.0 → 50 · hot 0.5 → 100 50 + 50 × (0.49 − 0.0) ÷ (0.5 − 0.0) = 99.0	99.0	20%	19.8
Provider unit-cost deflation `unit_cost_deflation` Falling published price per token is the Jevons elasticity input; cheaper is hotter only while usage also grows.	-0.49 median blended price change observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.1 → 0 · neutral 0.0 → 50 · hot -0.5 → 100 50 + 50 × (0.0 − -0.49) ÷ (0.0 − -0.5) = 99.0	99.0	20%	19.8
Hyperscaler capital commitment `capital_commitment` Point-in-time aggregate MSFT/GOOGL/AMZN/META capex growth from SEC filings.	0.39 TTM capex YoY growth observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold -0.1 → 0 · neutral 0.1 → 50 · hot 0.6 → 100 50 + 50 × (0.39 − 0.1) ÷ (0.6 − 0.1) = 79.0	79.0	20%	15.8
Committed data-center pipeline `committed_pipeline` Source-cited dc_projects ledger phased into a near-term annual power-arrival schedule.	15.2 average announced GW arriving per year over the next three years observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.0 → 0 · neutral 7.5 → 50 · hot 15.0 → 100 15.2 ≥ hot 15.0 → 100.0 (clipped)	100.0	20%	20

Axis equation

(94 × 0.2 × 1 + 99 × 0.2 × 1 + 99 × 0.2 × 1 + 79 × 0.2 × 1 + 100 × 0.2 × 1) ÷ 1 = 94.20

Supply tightness

coverage 100.0%

Input	Raw observation	Frozen transform	Score	Effective weight	Contribution
Fused component-cycle gate `supply_cycle_gate` The live fused memory gate; episode replay uses the equivalent point-in-time component-cycle state.	TIGHT categorical gate state observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	BREAKING → 10.0 · EARLY-CRACK → 55.0 · INTACT → 85.0 · OVERSUPPLY → 10.0 · BALANCED → 50.0 · TIGHT → 90.0 TIGHT maps directly to 90.0	90.0	33.333333%	30
Chokepoint tension `chokepoint_tension` Highest current supply-chain stranglehold/tightness score, with its source and relief clock.	0.92 0-1 measured/curated tension observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.85 → 100 0.92 ≥ hot 0.85 → 100.0 (clipped)	100.0	33.333333%	33.333333
Capacity relief-window tension `relief_window_tension` Power-vertical relief tension live; equivalent announced-capacity relief evidence in historical episodes.	0.94 share of committed demand landing inside the relief window observed 2023-12-31 known 2026-07-10T00:00:00 evidence retrospective_proxy internal record — not published	cold 0.2 → 0 · neutral 0.5 → 50 · hot 0.9 → 100 0.94 ≥ hot 0.9 → 100.0 (clipped)	100.0	33.333333%	33.333333

Axis equation

(90 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1 + 100 × 0.3333333333333333 × 1) ÷ 1 = 96.67

Diffusion index

(94.2 + 96.7) ÷ 2 = 95.4

Agreement score

100 − |94.2 − 96.7| = 97.5

Selected immutable print

object replays/preliminary-proxy-v1/prints/2023-12-31.json sha256 aca458ef3d602d61468a6803ea79d6222fed6514679f12f4eeba6c07eaa53fb0

Reproducibility contract

preliminary_proxy

Re-running the frozen methodology against the exact hash-bound input pack reproduces the numbers above. That establishes computational repeatability. It does not establish historical validity: this pack remains retrospective, is not independently point-in-time verified, and cannot authorize a forward print.

pass · component mapping preregistered pass · immutable objects verified fail · point-in-time evidence verified fail · independent sources verified

frozen methodology · versioned file

2026-07-10.v1 file sha256 8e11f875cb6c750d3389f9170481737784fbc4662aadfaa01ce0765aea7756da canonical model digest recorded in every print a2e714e7daa99a6771c9290b4a008795358032e887356e8f042c203bb08b6c1f

input pack · sha256

replays/preliminary-proxy-v1/inputs.json c3b1427d4845a69a399315674961739549532eb381bee0be3a17bbfc5b9bde69

evidence manifest · sha256

replays/preliminary-proxy-v1/evidence.json 6cef4b9dd74cc03627df466cd14f32c321f5f78410871727d553433059b71779

archive manifest · sha256

replays/preliminary-proxy-v1/manifest.json 3cced9b4cbafac49b6de813405a8a03eb519ff78f91d2e96f0306e7d47ffc0b8

Deterministic replay

uv run ak cycle thermometer \
  --spec <spec>.json \
  --replay <replay>.json \
  --replay-evidence <replay>.evidence.json \
  --replay-out /tmp/thermometer-replay.json

The command fails closed on evidence/hash mismatch. Compare the generated object digests with the immutable hashes listed here.

Replay archive

2018 MLCC

2017-12-31 heat 48.5 SUPPLY_AHEAD · agreement 78.7
2018-03-31 heat 72.5 DEMAND_LED_TIGHTNESS · agreement 87.9
2018-06-30 heat 81.6 DEMAND_LED_TIGHTNESS · agreement 71.5
2018-09-30 heat 65.1 SUPPLY_CONSTRAINED · agreement 50.1
2018-12-31 heat 27.8 SUPPLY_GLUT · agreement 97.3

2021 ABF

2020-09-30 heat 67.0 DEMAND_LED_EXPANSION · agreement 68.2
2020-12-31 heat 86.3 DEMAND_LED_TIGHTNESS · agreement 90.8
2021-03-31 heat 96.5 DEMAND_LED_TIGHTNESS · agreement 99.7
2021-06-30 heat 89.6 DEMAND_LED_TIGHTNESS · agreement 89.3
2021-09-30 heat 72.0 DEMAND_LED_TIGHTNESS · agreement 86.8

2023 HBM

2022-12-31 heat 30.6 SUPPLY_AHEAD · agreement 79.9
2023-03-31 heat 59.6 DEMAND_LED_EXPANSION · agreement 83.5
2023-06-30 heat 93.5 DEMAND_LED_TIGHTNESS · agreement 98.9
2023-09-30 heat 94.5 DEMAND_LED_TIGHTNESS · agreement 95.7
2023-12-31 heat 95.4 DEMAND_LED_TIGHTNESS · agreement 97.5

Methodology and archived JSON

Equal weights. Evidence families carry equal weight within each axis. Missing or stale inputs are excluded and weights re-normalize; coverage never disappears.

Heat plus disagreement. The diffusion index is the mean of the two axes. The visible band is their range; agreement is 100 minus that width.

No retrospective edits. One JSON and Markdown artifact per date. A changed rerun for an existing date fails closed.

replay.json

AI supply × demand, on one honest clock.

The retrospective proxy pack cannot authorize publication.

Three retrospective proxy episodes

Reproduce any point from raw input to index

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Demand intensity

Supply tightness

Episode replay prints

2018 MLCC

2021 ABF

2023 HBM

Methodology and archived JSON