Batch-backfill 104 unfilled Seeker stub research entries

[rjwalters]

Problem

104 entries in `src/data/research/problems/*.json` are stubs — either `title === "[Problem Title]"` or `problemStatement.formal` contains the literal `[LaTeX formulation of the theorem/conjecture]` placeholder. PR #20251 hides these from `research-listings.json` so the public gallery stops surfacing them, but the data is still on disk and would render placeholder content if anyone deep-links to the slug.

Inventory (from PR #20251 build summary)

`Listings: 1626 (skipped 104 unfilled Seeker stubs)`

Sample slugs filtered today: erdos-1023-oq-01, bezout-identity-oq-02-oq-02, bounded-prime-gaps-oq-01, brouwer-fixed-point-oq-01, buffons-needle-oq-01-oq-01, cayley-hamilton-minpoly-oq-01, cevas-theorem-oq-02-oq-01, derangements-oq-02, greens-theorem-oq-01, buffons-needle-oq-02. Full set printed by running `npx tsx scripts/research/build.ts`.

Three tiers of stubs based on what's available to backfill from:

1. Has substantial knowledge.md (>1KB) and selection-report.md — e.g. `brouwer-fixed-point-oq-04-oq-04`, `buffons-needle-oq-01-oq-01`, `divisibility-rules-oq-02`, `lovasz-local-lemma-oq-02`. Backfill from those sources, mirroring the manual approach in PR fix(research): backfill law-of-cosines-oq-06 stub + remove misnamed duplicate #19979.
2. Has selection-report.md only — synthesize `problem.md` from the selection report's rationale and target.
3. Has only the minimal-template knowledge.md (~340 bytes) — Seeker created the entry but no Researcher ever worked on it. These are honestly empty; the right move may be to delete the directory and `src/data/research/problems/.json` rather than fabricate content.

Acceptance criteria

• All entries currently listed by the `Listings: ... (skipped N unfilled Seeker stubs)` summary either:
  • have a proper `title`, `problemStatement.formal`, `problemStatement.plain`, and reappear in listings; OR
  • are removed entirely (research dir + site JSON + registry entry).
• `isUnfilledStub()` reports 0 skipped stubs in the final build.

Suggested approach

• Categorize all 104 via a one-shot script (use the `isUnfilledStub` heuristic and check the sizes of `knowledge.md` / `selection-report.md`).
• Tier 1 → Researcher/Curator agent batch task, one PR per ~10 entries.
• Tier 2 → smaller manual or scripted backfill.
• Tier 3 → delete in a single PR.

Labels

curator-please-prioritize, area:research-data

[Sean-Kenneth-Doherty]

Codex progress update:

• Refreshed current upstream inventory at 105 unfilled Seeker research stubs by the issue heuristic.
• Opened [codex] Remove empty Seeker research stubs #20259 to remove 7 empty/no-workspace stubs that can merge cleanly from Sean's fork.
• Opened [codex] Backfill Seeker research stub metadata #20265 to backfill 3 existing Seeker stub JSON entries from local research notes: burnside-counting-oq-01, cauchy-schwarz-oq-02-oq-01, and partition-theorem-oq-04.
• Also prepared a 5-file upstream-local backfill commit (66a8eaa3a8) covering amgm-inequality-oq-04-oq-05 and binomial-theorem-oq-03-oq-01; those two files are absent from Sean's stale fork base, so I left them out of [codex] Backfill Seeker research stub metadata #20265 to avoid add/add merge noise.

If #20259 and #20265 both merge, the current upstream heuristic count should drop from 105 to 95.

[Sean-Kenneth-Doherty]

Codex progress update #2:

Opened #20268 with four additional Seeker stub backfills:

• brouwer-fixed-point-oq-04-oq-04
• cayley-hamilton-oq-01-oq-01
• lovasz-local-lemma-oq-02
• taylor-sincos-convergence-oq-01

Current open clean PRs against this issue:

• [codex] Remove empty Seeker research stubs #20259 removes 7 empty/no-workspace stubs.
• [codex] Backfill Seeker research stub metadata #20265 backfills 3 existing stub JSON entries.
• [codex] Backfill additional Seeker stub metadata #20268 backfills 4 more existing stub JSON entries.

If all three merge, the refreshed upstream heuristic count should drop from 105 to 91. Labels are still empty because this token cannot add upstream labels.

[Sean-Kenneth-Doherty]

Codex progress update #3:

Opened #20269 with two completed Seeker stub backfills:

• ballot-problem-oq-02-oq-02
• buffons-needle-oq-01-oq-01

Current open clean PRs against this issue:

• [codex] Remove empty Seeker research stubs #20259 removes 7 empty/no-workspace stubs.
• [codex] Backfill Seeker research stub metadata #20265 backfills 3 existing stub JSON entries.
• [codex] Backfill additional Seeker stub metadata #20268 backfills 4 existing stub JSON entries.
• [codex] Backfill completed Seeker stub metadata #20269 backfills 2 completed stub JSON entries.

If all four merge, the refreshed upstream heuristic count should drop from 105 to 89.

Also prepared but not published from Sean's stale fork base: upstream-local commit be635dc7fc includes divisibility-rules-oq-02 and sperner-ndim-oq-03; those two conflict when published from the old fork base and should be landed after a fork refresh or from an upstream-based branch.

[Sean-Kenneth-Doherty]

Batch 4 landed as PR #20270: #20270

This backfills three more Seeker stub metadata files:

• erdos-1107-oq-02
• cayley-hamilton-reduction-oq-02-oq-01-incomplete-01
• konigsberg-oq-03-oq-02

Validation passed on the publish branch: JSON parse/placeholder scan, git diff --check, research enrich dry-run, research build, annotations build in existing non-strict mode, TypeScript build, and git merge-tree --write-tree origin/main HEAD.

Projected heuristic count: if #20259, #20265, #20268, #20269, and #20270 all merge, the refreshed count should drop from 105 to 86. I also prepared basel-problem-oq-04 locally, but left it out of this publish branch because Sean-fork merge safety showed an upstream conflict on that file.

[Sean-Kenneth-Doherty]

Batch 5 landed as PR #20271: #20271

This backfills five more Seeker stub metadata files:

• bounded-prime-gaps-oq-03-oq-01-oq-04
• bezout-identity-oq-02-oq-01-oq-02-oq-02
• bezout-identity-oq-03-oq-04-oq-01
• shannon-entropy-oq-03-incomplete-01
• sum-of-kth-powers-oq-04-incomplete-01

Validation passed on the publish branch: JSON parse/placeholder scan, git diff --check, research enrich dry-run, research build, annotations build in existing non-strict mode, TypeScript build, and git merge-tree --write-tree origin/main HEAD.

Projected heuristic count: if #20259, #20265, #20268, #20269, #20270, and #20271 all merge, the refreshed count should drop from 105 to 81.

[Sean-Kenneth-Doherty]

Batch six is up: #20273 backfills five more non-empty Seeker stubs (cube dissection, FTC, Roth k=3, Cantor/Rice, and Konigsberg WIP). Validation passed with the standard research dry-run/build, annotations build, TypeScript build, targeted placeholder scan, diff check, and merge-tree check. If #20259, #20265, #20268, #20269, #20270, #20271, and #20273 all merge, the projected stub count drops from 105 to 76.

[Sean-Kenneth-Doherty]

Batch seven is up: #20274 backfills four more non-empty Seeker stubs (Erdős #109, Erdős #1108 OQ-01, Erdős #157 incomplete, and Erdős #871). Validation passed with the standard research dry-run/build, annotations build, TypeScript build, targeted placeholder scan, diff check, and merge-tree check. Two additional files were prepared locally in the origin-based batch branch (Abel-Ruffini Galois OQ-02 and multinomial covariance) but parked because the fork-based merge-tree check conflicted on those files. If #20259, #20265, #20268, #20269, #20270, #20271, #20273, and #20274 all merge, the projected stub count drops from 105 to 72.