Journal — 2026-07-10

Promotion — hydraulic water computers (headless)

Promoted 10-inbox/raw/2026-07-09-hop-hydraulic-water-computers.md — the convergent-invention capture about water-as-computer, invented twice. Ran the full /promote workflow autonomously (no pause, no git; the auto-commit agent will pick this up).

What I found: three factual claims plus an implicit synthesis. I promoted three notes and skipped one claim.

Notes created:

• claim-phillips-moniac-hydraulic-economic-computer-1949 — Claim 1, the MONIAC. The object is institutionally anchored (Science Museum Group object no. 1995-210, Tier 1–2), so I kept the object claim, but the tidy numbers — ~£400 build cost, "within 2%" accuracy — rest only on Tier-4 Wikipedia, so I flagged them [unverified-quant] and kept the note seedling.
• claim-lukyanov-water-integrator-1936-hydraulic-differential-equation-solver — Claim 3, the Soviet Water Integrator. The genuinely novel entity of the capture (scored novel at 0.593). Entirely single-sourced to Wikipedia, so [unverified-source], seedling.
• claim-hydraulic-analog-computers-independently-invented-economics-engineering — the synthesis observation, which is the actual reason the capture exists. It plugs straight into the vault's existing convergent-invention thread (claim-reverse-mode-multiple-independent-discovery, claim-labview-max-independent-patch-cord-convergence) and its skeptical counterweight (claim-chandler-simultaneous-discoveries-are-incremental-repackagings).

Questions routed:

• question-verify-moniac-cost-and-accuracy-figures — pointing at Phillips's own 1950 Economica paper and the museum's costing for the £400/2% figures.
• question-verify-lukyanov-water-integrator-primary — pointing at the Polytechnic Museum (Moscow) holdings and Russian engineering histories.

What I skipped and why: Claim 2 — the POW color (radio hidden in a clog, a secret water heater wired into the camp lighting, the MBE). Striking, and it's the emotional core the capture leaned on, but it's biographical color, not a load-bearing vault claim — the capture author said as much ("color rather than a load-bearing new claim"). I left it in the inbox and recorded it in not_promoted: so it reads as judged, not missed. I preserved the "rehearsal" framing (water to survive a camp, then water to explain an economy) as commentary in the synthesis note rather than asserting the POW facts as fact in neutral voice. Also left behind: the further-leads (Phillips curve, the ENIAC-echo naming, the Soviet-bloc tech-transfer thread) — saved hooks, not distinct claims.

Retrieve-before-write: no collision — nothing on Phillips, MONIAC, or Lukyanov existed in 30-notes/. But the convergence thread is well-populated, so I wired all three new notes densely into it.

What felt off: the sourcing is thin across the board — all three claims trace to Wikipedia, and the object corroboration (Science Museum) only covers that the MONIAC exists, not the quant details resting on it. No tier inflation in the capture, to its credit: it honestly labeled everything Tier 3–4 and flagged its own post-worthiness as "maybe… color." The thing I'd flag to Cali is duplicated research shape: this is the second Wikipedia-sourced convergent-invention hop I've promoted from this batch worker in two days (the first was claim-labview-max-independent-patch-cord-convergence, LabVIEW/Max, 2026-07-09). Same structure, same weakness — a pleasing "two lone inventors, same idea" story carrying the whole hop on Tier-4 sourcing. The pattern is real and the vault clearly values it, but the batch worker keeps reaching for the tidy version and stopping before a primary. Worth watching whether these converge into an MOC on convergent invention — or whether they're a tell that the hop gate rewards narrative tidiness over verifiability.

— Seek

Promotion — physicist's harpsichord → ivory "passport" (headless)

Promoted 10-inbox/raw/2026-07-09-hop-instrument-ivory-passport.md, the five-hop chain that runs from a materials physicist's Baroque-music hobby out to international wildlife law. Ran the full /promote workflow autonomously — no pause, no git (the auto-commit agent will pick this up).

Winnowing — 3 claims found → 3 notes → 2 claims skipped:

Notes created:

• claim-martinez-miranda-parallel-physics-and-harpsichord-careers — Claim 1, the physicist-with-a-conservatory-degree hook. Tier-4 biography, uncontested, so it clears the floor but stays seedling. A genuine orphan — nothing in the vault is adjacent — which is the whole reason the hop was worth capturing.
• claim-cites-musical-instrument-passport-certificate — the international half of Claim 3, the "passport." The pre-February-26-1976 acquisition cutoff is Tier-1 fws.gov and load-bearing; but the certificate's mechanics (CoP16 Bangkok 13 Mar 2013, 3-year validity, non-transferable) rest only on Tier-3 trade press, so I flagged those [unverified-quant] and kept it seedling.
• claim-us-worked-ivory-de-minimis-200g-exception — the domestic half of Claim 3, split out as its own atomic note. The 200 g threshold + pre-July-6-2016 manufacture date are Tier-1 fws.gov. I deliberately separated it from the passport note because the capture bundled two genuinely distinct legal mechanisms (border travel vs. domestic sale) that even run on different dates — easy to blur, worth keeping apart.

Questions routed:

• question-verify-cites-musical-instrument-certificate-cop16-primary — pointing at CITES Resolution Conf. 16.8 and the CoP16 record to lift the passport note's Tier-3 mechanics to primary.
• question-source-harpsichord-hip-revival-craft-history — the home for the claim I didn't promote (see below), pointed at Hubbard's Three Centuries of Harpsichord Making (1965) and the Galpin Society.

What I skipped and why: Claim 2 — the harpsichord-revival craft history (early piano-frame instruments reversed toward Baroque light construction; Hubbard/Dowd/Skowroneck; Galpin Society 1946) — and the hop-3 jack/plectrum quill-vs-delrin detail. Both are interesting and genuinely distinct, but the capture recorded them from unattributed web searches with no source URL at all. I will not manufacture provenance to fill a source_url, so rather than write a note resting on nothing, I skipped both and routed them to a sourcing question. Recorded in not_promoted: so they read as judged, not missed. Also left behind: the further-leads (Bouchet Prize, CITES negotiation history, her actual liquid-crystal research) — saved hooks, not claims.

Retrieve-before-write: zero collisions. The semantic-index hint and the capture both predicted this, and searching 30-notes/ and 50-questions/ confirmed it — the vault holds nothing on harpsichords, ivory, CITES, or Martínez-Miranda. The three new notes form a small self-contained cluster; I wired them densely to each other (the two ivory-law notes are strong cross-links; the biography note bridges in via the instrument-materials theme). Three notes is below the MOC threshold, so none yet.

What felt off: two things. (1) Sourcing is lopsided. The capture's Tier-1 spine (fws.gov) is real and well-quoted, but two of its four "domains" (the harpsichord craft history, the plectrum mechanism) were carried entirely on sourceless web searches — the hop chain narrated findings hop-by-hop but only preserved URLs for the first and last hops. That's a capture-discipline gap: the middle of the chain looks as authoritative as the ends but isn't citable. (2) I could not independently re-verify the Tier-1 fws.gov quotes — WebFetch was permission-blocked in this headless run — so the two law notes are marked audit_status: capture-verified (bee read the primary; queen's re-check blocked), not verified-verbatim. No tier inflation in the capture's own labels, to its credit; it honestly marked the bio Tier 4 and the law Tier 1. The one structural nit: it filed a two-mechanism legal claim as a single "core claim," which I split on promotion.

— Seek