举牌

口径

13G 行来自解析过持股百分比的举牌面板；13D 行来自 EDGAR 文件流（百分比在文件内，如实不猜）。每行链出 EDGAR 原文。

SC 13G passive-stake filing stream — SEC EDGAR daily crawler index (public domain, 17 U.S.C. §105), trailing ~120-day window, filed-date point-in-time; a SC 13G/A amendment is a new filing, never a silent overwrite. EFTS full-text search froze on the Schedule 13 family after 2024-12-17 (live-verified: forms=SC 13G returns 0 hits for 2025+ windows), so the daily index is the source. It lists a filing under EVERY covered company, so rows are DEDUPED per accession: the one listed company is the SUBJECT (ticker backfilled OFFLINE from the cached SEC company_tickers snapshot — a current-snapshot display label, not as-of-filing) and the co-indexed names surface as the FILER; ambiguous or unlisted groups honestly keep every member (placeholder filer / null ticker, counted in data_health.source_health), so the row count can slightly exceed the distinct-accession count. Honest v1 limits, disclosed not papered over: the share count and event date live INSIDE the filing documents and are NOT parsed here. The percent-of-class second stage IS bounded-parsed for the VISIBLE top-150 rows only (EDGAR index.json + primary document, ≤2 polite requests per row; structured cover-page tag first, then the 'Percent of class' label, then the 'X.X% of the ... class' narrative; /A amendments carry pct_prev from 'previous X%'-style sentences only): rows beyond the visible window are NOT parsed, and every extraction miss stays an honest null (counted in data_health.source_health), never a guess or a default. pct_status is DERIVED from parsed values only: exited = an explicit parsed 0, below_5 = a parsed value < 5; a null pct never yields a status. Group filings show the FIRST reporting person's cover-page percentage. Each row links the filing's EDGAR index page, which lists every document and names the reporting person. Counts are FILING counts, not holder counts. Display-only, exploratory, not a research claim; NOT part of any OOS pipeline.