learnings: Voyager operating patterns from 2026-08-08

data/community_learnings/20260808_85b75bef.jsonl

@@ -0,0 +1,3 @@
+{"id": "vy-curate-01", "timestamp": "2026-08-08T15:56:12Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "curation", "announcement_policy"], "description": "For the Voyager daily update: curate, don't enumerate. Passing the novelty, significance and dedupe gates earns a result eligibility, not a bullet. Keep the list short and interesting rather than exhaustive; never pad a message to look productive. A long list is evidence of under-curation, and omitting a real result costs nothing because the permanent dedupe keys on cited PR numbers.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "", "theorem_name": "", "project": "TauCetiRoadmap"}}
+{"id": "vy-sha-02", "timestamp": "2026-08-08T15:56:12Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "watermark", "git_sha", "data_integrity"], "description": "Never pad or retype a git SHA into the Voyager state DM. Copy the full hash exactly as `git rev-parse origin/docgen` prints it; a plausible 40-character string that names no object makes the next run's window unresolvable and nothing downstream notices. Check with `git cat-file -t <sha>` before sending; if a bad one went out, send a superseding DM, since the protocol reads the newest.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "", "theorem_name": "", "project": "TauCetiRoadmap"}}
+{"id": "vy-gate-03", "timestamp": "2026-08-08T15:56:12Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "significance_gate", "prompt_framing", "unproved_endpoint"], "description": "Describe candidates to the ChatGPT significance gate in the words of the file's own `## Main results`, not in the words of the roadmap milestone they serve. The gate can only judge the summary it is given, so an overstated summary returns ANNOUNCE for something unproved: a batch of adic-spaces PRs described as 'establishing that the valuation spectrum is a spectral space' earned that recommendation while no SpectralSpace conclusion existed in the tree. Find the theorem statement, or do not announce it.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "", "theorem_name": "", "project": "TauCetiRoadmap"}}

data/community_learnings/20260811_1d857014.jsonl

@@ -0,0 +1,3 @@
+{"id": "vy-nolen-04", "timestamp": "2026-08-11T08:51:55Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "curation", "no_target_length", "announcement_policy"], "description": "Sharpening of the curation rule (supersedes the 'keep the list short' framing of vy-curate-01): there is no minimum and no maximum bullet count for a Voyager update — interest is the only test. Owner's words: 'there is no rule to cut it to 5. the rule is that there doesnt need to be a minimum or maximum, but you just need to make them be interesting.' Do not pad a thin window, and do not drop an interesting result to hit a number — cutting a good bullet to satisfy a self-imposed count looks like discipline and is the sneakier failure (a five-bullet message once dropped the Huber/Tate Layer-0 block purely for the count). A one-bullet day and a ten-bullet day are both correct if that is what the window held.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "skills/voyager/SKILL.md", "theorem_name": "", "project": "TauCetiRoadmap"}}
+{"id": "vy-prbody-05", "timestamp": "2026-08-11T08:51:55Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "candidate_extraction", "read_pr_bodies", "titles_are_not_filters"], "description": "Voyager candidate extraction must read the contents/body of EVERY merged PR in the window — all prefixes, chore:/refactor:/fix: included. Titles may order the reading, never exclude a PR. Owner's words: 'you cant rely on pr title, dont be lazy, this is meant to be a slow methodical and careful check. you need to look at the contents of each PR and from that make your list.' Title-tone triage caused the worst miss so far: fifteen feat(EllipticCurve) PRs across two windows, none opened, zero elliptic-curve results in 105 announced bullets (Silverman III.10 on Aut(E) for j outside {0,1728}, TauCeti#2248; the quadratic twist with its invariant theory, TauCeti#2254; quadratic Galois descent, TauCeti#2268). Batch the fetches with gh pr list --json number,title,body. The window bounds discovery, not eligibility: the dedupe keys on cited PR numbers, so fold late-found misses into the next run.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "skills/voyager/SKILL.md", "theorem_name": "", "project": "TauCetiRoadmap"}}
+{"id": "vy-rank-06", "timestamp": "2026-08-11T08:51:55Z", "command": "teach", "type": "user_teaching", "before_code": "", "after_code": "", "pattern_tags": ["voyager", "significance_gate", "interest_ranking", "every_pass"], "description": "The ChatGPT significance gate (via the chatgpt-math MCP, one batched call, reasoning_effort high — never max, which times out) runs on EVERY Voyager pass including previews and dry-runs, and is the second opinion on how interesting/noteworthy each result is, not only an ANNOUNCE/SKIP eligibility filter. Owner's words: 'make sure you use the chatgpt 5.6 sol via the mcp to get a second opinion on how interesting/noteworthy a result is.' Ask it to rank the strongest candidates; the ranking, not the order of discovery, drives bullet order and which eligible results wait. Verdicts remain advice, not authority — sanity-check every sentence against the actual Lean statement.", "math_area": "other", "accepted": true, "source": "user_teaching", "context": {"file_path": "skills/voyager/SKILL.md", "theorem_name": "", "project": "TauCetiRoadmap"}}