# First-mention research recipe

A bounded chronology method for finding earliest indexed evidence without overstating provenance or global completeness.

## Calculation

- Mode: earliest_matches
- Formula: Order deterministic matching SourceItems by stored external_created_at ascending within the selected corpus; use semantic hits only as separately reviewed candidates.

## Query

Within [selected channels] and [optional dates], find the earliest indexed post matching [exact phrase or carefully defined claim]. Separate exact lexical matches from semantic candidates, show timestamps, excerpts, source names, and permalinks, and state corpus gaps. Say earliest in this selected corpus, not original source or first on Telegram.

## Inputs

- **Claim definition:** A distinctive phrase, URL, identifier, or tightly bounded proposition.
- **Selected corpus:** Channels and indexed-history coverage included in the search.
- **Date bounds:** Optional start and exclusive end timestamps with timezone.
- **Match standard:** Exact wording required or semantic paraphrases allowed for review.

## Expected output

- Chronological matching-post list
- Earliest verified corpus timestamp
- Exact-versus-semantic distinction
- Direct evidence links
- Coverage and provenance disclaimer

## Coverage checks

- [ ] Do not claim global Telegram completeness
- [ ] Do not infer copying from order alone
- [ ] Verify semantic candidates manually
- [ ] Keep timezone and publication dates explicit
- [ ] Record missing source history before publishing