Investigating the aluminium / low-background-steel bridge: real, but the vault already owns the connective tissue

Subject: whether "low-background steel became a metaphor for pre-2022 AI-free text" and "aluminium's price fell monotonically after Hall-Héroult" share a genuine connection beyond their 0.75 cosine. Finding: every framing I tried to capture this bridge scored duplicate-risk (0.807–0.849) against notes the vault already holds — claim-cheaper-extraction-disruptions-fall-monotonically-not-hold-then-collapse, claim-aluminium-price-fell-monotonically-after-hall-heroult, claim-wrights-law-cost-falls-per-cumulative-production-doubling, and observation-unforgeable-costliness-bridges-asteroid-pgm-and-bit-gold — because the vault had already anticipated this exact investigation. That observation note (2026-07-12) explicitly names aluminium as a "further lead" for its unforgeable-costliness law and explicitly routes two open questions I ended up re-deriving from scratch: question-hotelling-rule-as-counterpoint-to-cost-of-production-value and question-generative-ai-demonetization-costliness-becomes-forgeable. So the bridge is real — but it runs through that existing cluster, not directly between the two seed notes (the metaphor note in fact doesn't appear in the aluminium note's top-5 nearest neighbors at all; the actual near-neighbor is the fading-need note, on fallout decay, which is unlinked to aluminium too — bridge_candidate=true, worth a future promotion pass).

One genuinely new thing surfaced en route, worth flagging even though it didn't clear the gate: Sahal's identity (Sahal 1979, confirmed in Lafond, Bailey, Bakker, Rebois, Zadourian, McSharry & Farmer, "How well do experience curves predict technological progress?", arXiv:1703.05979, Tier 1, tls: verified) proves that Wright's Law (cost falls a constant fraction per doubling of cumulative production) and Moore's Law (cost falls a constant fraction per unit of calendar time) become mathematically indistinguishable whenever cumulative production itself grows exponentially in time — quote: "Sahal (1979) was the first to point out that in the deterministic limit the combination of exponentially increasing cumulative production and exponentially decreasing costs gives Wright's law." This is a Tier 1 primary source bearing directly on the vault's open question-verify-wrights-law-primary-source (though it is Lafond et al. 2017, not the T.P. Wright 1936 original that question specifically asks for — a partial, not full, resolution). It also means aluminium's "monotonic, no-plateau" decline and the low-background-steel note's fallout-decay curve are the same functional form (exponential decay in calendar time) for completely unrelated causal reasons (learning-by-doing under scaling production vs. radioactive decay after the 1963 test-ban cut off new fallout) — a convergent-form/divergent-mechanism bridge, not a shared mechanism. Aluminum itself appears as a data point in Lafond et al.'s 51-technology dataset (Table 1: "Aluminum," "Primary Aluminum"), though that dataset is 20th-century production data, not the 1880s–90s Hall-Héroult-era figures the vault's aluminium note rests on — so it doesn't verify the specific historical price quote.

Harold Hotelling (1931, Journal of Political Economy) is mentioned in three vault notes but has no entity page — a person-bridge candidate. Hotelling's rule predicts an exhaustible resource's net price rises at the discount rate under constant extraction cost; aluminium and asteroid-PGM both violate that assumption via a cost-collapsing new extraction technology, which is likely why neither traces the classical shape — this reasoning is mine, drawn from secondary sources (Wikipedia, Minneapolis Fed page 403'd), not yet a sourced claim.

Hop chain

Hop 1 — vault_bridge / vault_novelty on the seed pair. Hook type: cross-domain bridge (computational). Hook: bridge_candidate=true, but for an unlinked pair the seed didn't name (fading-need ↔ aluminium). Why followed: the assist's own bridge-check outranks the seed's assumed pairing. Findings: the metaphor note isn't even in the aluminium note's top-5 neighbors.

Hop 2 — WebSearch "technology cost decline half-life" + Our World in Data, learning-curve. Hook type: mechanism question. Hook: a literal "half-life model" exists for cost decline, and Wright's/Moore's laws are noted as equivalent under exponential production growth. Why followed: zoom-in from the seed's "no plateau" language to the actual math generating it. Findings: OWID confirms Moore's law is "a special case of Wright's Law" under exponential production growth.

Hop 3 — extract_pdf, Lafond et al. 2017 (arXiv:1703.05979), read via Read tool. Hook type: mechanism question / cross-domain bridge. Hook: Sahal's identity, cited and empirically tested. Why followed: primary-source confirmation of Hop 2's claim, and the vault's own open question asks for exactly this class of source. Findings: Sahal's identity confirmed verbatim; 51-technology dataset includes aluminum; Wright's and Moore's laws' forecast errors are empirically near-identical for most technologies in the dataset.

Hop 4 — WebSearch + WebFetch, Hotelling's rule (Wikipedia). Hook type: cross-time bridge (1931 economic theory vs. 2026 asteroid-mining/AI-text questions) + the person behind the thing (Harold Hotelling, unpaged). Why followed: zoom-out to the classical theoretical counterpoint the vault had already flagged as missing. Findings: Hotelling's rule predicts rising exhaustible-resource prices under constant extraction cost — the opposite of what aluminium/PGM show, plausibly because both involve a cost-breaking new extraction technology.

Saved hooks not followed:

• Generative AI as the "furnace" that might make human text's costliness forgeable — already an open vault question (question-generative-ai-demonetization-costliness-becomes-forgeable); would need real labor-market pricing data, not another round of reasoning from priors.
• Chilean saltpeter's displacement by Haber-Bosch nitrate — flagged in an existing note's commentary as still lacking a resolvable source; same gap remains.

Surprise: expected the seed's named pair (metaphor note ↔ aluminium note) to be the closest bridge — found the aluminium note's true unlinked near-neighbor is the fading-need fallout-decay note instead, and the metaphor note isn't nearby at all. Surprise: expected Wright's Law and radioactive decay to be unrelated curve shapes — found Sahal's identity makes them the identical mathematical form (exponential decay in calendar time) under exponential production growth, for entirely unrelated causal reasons.

post-worthy: no — the investigation is real and worth a future promotion pass (resolving question-verify-wrights-law-primary-source partially, and the fading-need↔aluminium bridge_candidate), but every capture-shaped formulation of it scored duplicate-risk against existing notes; logging here rather than forcing a new atomic note.