Define the topic
Specify entities, claims, event boundaries, and exclusions so the timeline is about one research proposition rather than every use of a broad term.
Workflow · Topic chronology
Build a chronological, evidence-linked account of how a topic appeared and changed across selected Telegram sources and a defined period.
Research fit
Researchers reconstructing an event or narrative across public channels who need readable chronology, source-level citations, and explicit gaps instead of a relevance-ranked list with dates mixed together.
Specify entities, claims, event boundaries, and exclusions so the timeline is about one research proposition rather than every use of a broad term.
Choose the relevant channel set and an explicit period; record sources with no messages inside it.
Use literal matches for known wording and semantic search for paraphrases, translations, or conceptually related updates.
Arrange verified posts chronologically, preserve disagreements, and label source claims separately from independently confirmed facts.
State missing history, unavailable sources, retrieval limits, and whether the earliest row is merely earliest in the selected corpus.
Practical module
A strong timeline uses evidence rows first and narrative synthesis second, so a reviewer can challenge every turning point.
Analysis contract
The workflow combines hybrid semantic retrieval with deterministic exact matches and stored publication timestamps, then asks the model to synthesize only the cited context in chronological order.
Selected indexed Telegram sources and an explicit or interpreted publication period. Full-period summaries fail closed when the message count exceeds the configured safe limit.
A structured method for turning retrieved posts into a source-checkable chronology with explicit event types and coverage gaps.
Across [selected channels] from [start] to [end], build a chronological timeline for [defined topic]. Include exact and semantic evidence, source, timestamp, permalink, change type, and verification status. Preserve disagreements, name zero-message sources, and describe the earliest item only as earliest found in this corpus.
Hybrid retrieval plus chronological ordering of cited stored publication timestamps; synthesis is narrative and never replaces the underlying evidence rows.
# Telegram topic-timeline recipe A structured method for turning retrieved posts into a source-checkable chronology with explicit event types and coverage gaps. ## Calculation - Mode: plain_semantic - Formula: Hybrid retrieval plus chronological ordering of cited stored publication timestamps; synthesis is narrative and never replaces the underlying evidence rows. ## Query Across [selected channels] from [start] to [end], build a chronological timeline for [defined topic]. Include exact and semantic evidence, source, timestamp, permalink, change type, and verification status. Preserve disagreements, name zero-message sources, and describe the earliest item only as earliest found in this corpus. ## Inputs - **Topic:** Entities, claims, aliases, languages, and exclusions that bound the investigation. - **Sources:** The selected ready channels whose indexed history is in scope. - **Period:** Explicit start and end dates with timezone. - **Timeline taxonomy:** First mention, update, repetition, contradiction, correction, or shift. ## Expected output - Chronological evidence rows - Cited turning points - Source disagreements - Verification status - Coverage and earliest-evidence caveat ## Coverage checks - [ ] Do not infer propagation from timestamp order - [ ] Do not omit zero-message selected sources - [ ] Do not summarize an oversized period - [ ] Keep every substantive event linked to evidence - [ ] Review translations and paraphrases in the original post
Boundaries
Founding research pilot
Bring one recurring research job and a small set of public sources. We will assess fit personally before offering a pilot.
30 founding researcher places · rolling admission · personal reply within 2 business days
Analytics cluster
Continue researching