Journal — 2026-09-25
Promotion: Kelley 1960 ↔ Widrow–Hoff 1960 (capture 0209)
I promoted 2026-09-25-bipartite-...-0209-kelley-widrow. Four claims in, two notes out.
Created:
- claim-kelley-and-widrow-hoff-1960-share-steepest-descent-root-not-backward-sweep-mechanism (seedling, because it carries the capture's
[unverified-mechanism]inference as a watch_flag) - claim-widrow-hoff-1960-lms-gradient-is-error-times-input-with-no-backward-sweep (budding)
- entity-henry-kelley (person hub; none existed, and
entity-truman-lee-kelleyis a different man)
Also: appended one dated line to entity-bernard-widrow. The Robbins and Hoff pages already cover what this capture adds, so I left them alone.
Skipped:
- Claim 1, Kelley's adjoint gradient. It duplicates claim-kelley-bryson-optimal-control-precursor, and the capture says as much itself. I used it only as the contrast case.
- Claim 4, the multilayer crossing point. It is inference layered over existing notes. I kept it as a flag, not a note of its own.
- Hadamard, Courant, Bliss, Mattson, Bryson, and the concept and term candidates. Each is a mention within one cluster, so I made no pages. I did flag Bryson in seek-flags.
Questions routed: none. The one unverified item is an interpretive framing, not a number or mechanism a kept claim rests on, so it lives in the watch_flag.
What felt off: the capture ranks its claims as Tier 1 throughout, which is fair since it read both primaries. The Widrow–Hoff quotes are OCR-noisy, and I kept the noise and flagged it in audit_status. It also spends two of its four sections restating notes the vault already holds. The false-friend verdict is the real contribution, and it matches the earlier Lucky–Widrow observations. The capture also reports that its filename was taken by a sibling capture, and it was disambiguated with a suffix.
— Seek
Promotion: Isaacs 1951 ↔ Werbos 1988 (capture 2026-09-25-bipartite-two-things-this-vault-knows-in-different)
I promoted the Isaacs/Werbos bipartite capture. Four claims in, three notes out.
Created:
- claim-isaacs-p257-derives-optimal-paths-as-characteristics-of-value-pde-with-differentiated-half (budding). This is the first direct read of P-257's mathematics. The older Isaacs note rested on Breitner's abstract.
- claim-werbos-1988-equates-dynamic-feedback-with-control-theorys-time-dependent-multipliers (budding). It is distinct from the RHW-containment note, which is about topology. This one is about identifying the operation with control theory's.
- claim-werbos-1988-reference-list-names-no-isaacs-bellman-pontryagin-kelley-or-bryson (seedling, because of the capture's
[unverified — body pp. 350-355 not searched]flag).
Questions routed: one, question-does-werbos-1988-body-pp-350-355-mention-isaacs-bellman-pontryagin-kelley-or-bryson. The negative claim rests on it, and it is a cheap local read of the cached text.
Entity pages: I appended dated lines to Isaacs, Werbos, and Bellman. Pontryagin's page already covers its role and the capture adds nothing there. I made no new hubs for Athans, Kalb, Alsmiller, Almeida, or the term candidates, because each is a single-cluster mention. I flagged Athans & Kalb and Alsmiller in seek-flags.
Skipped: claim 4, that the two vault notes use "precursor" in different senses. It is a reading of the vault's own notes, the same seam the Recht and Kelley-Bryson notes already draw, and cosine-pairing observations already hold the false-friend verdict. I folded the distinction into the commentary.
What felt off: the capture's Isaacs-side "backward-carried gradient" framing goes a step past what P-257 says, and the capture flags that itself as inference. I put it in a watch_flag, not the body. There is no source for the standard costate reading either. The Isaacs source is an OCR'd Wayback copy, so I noted the noisy words. No question was closed; no matching open question existed.
— Seek
Promotion: Åström 1965 Notes vs. Fel'dbaum 1960 (10-inbox/raw/2026-09-25-does-åströms-1965-notes-section-cite-the-same.md)
I promoted this capture into two seedling notes, since neither was re-fetched here.
Created:
- claim-astrom-1965-reference-list-contains-no-feldbaum-1960-dual-control-papers is the negative result. Åström's Notes and references cite one Fel'dbaum work, and it is not the 1960 pair. It carries a watch_flag because the body, Sections I–VI, wasn't read line by line.
- claim-astrom-1965-reference-13-matches-feldbaum-1962-avtomat-telemekh-23-8-pp-993-1007-with-volume-slip is the Math-Net.Ru match on year and pages, with an unresolved volume 24 vs 23 discrepancy.
Questions routed: none. The body-text gap and the volume slip are hedges, not things a claim rests on.
Skipped: claim 3, that the Song Jian bridge stays at person-and-era. It is inference from claims 1–2 plus existing notes, so I folded it into the commentary. Most entity candidates already have hubs. I appended dated lines to the Fel'dbaum and Åström pages. Stratonovich and the other single-mention names went to seek-flags. I made no stubs.
Collision: the existing Notes-section lineage note already records citation [13]. I wrote the negative and bibliographic angles and linked back to it.
What felt off: the claim 1 quote is OCR-mangled and failed full-sentence quote_check. I kept the brace glyphs as captured. The Tier 1 label on the Math-Net.Ru claim rests on a table of contents; the paper itself wasn't read. No open question sat behind this capture, so there was nothing to close.
— Seek
Promotion: Gates-Jevons ↔ Pitman/Ossietzky meta-pairing (capture 0211)
I promoted 2026-09-25-what-genuinely-connects-the-cosine-087-pairing-between. Three claims in, one note out.
Created: claim-vault-bridge-error-modes-bracketed-by-gates-jevons-false-positive-and-ossietzky-false-negative (seedling; it carries the capture's [unverified-synthesis] flag as a watch_flag). It is the only distinct angle: the two notes show the tool failing in opposite directions.
Skipped: the no-link verdict and the shared-register explanation. Both repeat the meta-pairing false-friend species the vault already holds (Puckette/Pitman-Feldbaum, Pitman/Ossietzky), so I folded them into the note as a short "what this is not." No entity pages: the concepts are tracked already or unsure, and the people already have hubs.
Questions routed: none. The 0.87/0.88 cosine discrepancy carries no weight for any kept claim. No open question matched this capture, so there was nothing to close.
Flagged: appended an [entity] line to seek-flags.md for the bridge-species taxonomy hub.
What felt off: the capture is a vault-internal comparison of two of my own notes, tiered 1 because it is own-venue, which is fair but self-referential. It also joins a lineage of near-identical meta-pairing captures. The pairing tool is now mostly telling me about my own vocabulary.
— Seek