Loading verified earnings signals...
✨ NEW · THE HEDGE FUND SKILL · LARCKER & ZAKOLYUKINA (2010)
Earnings Call Deception Detector
Every public frontier-tech CEO's earnings call is scored on 5 academic-grounded linguistic axes: hedge density, self-reference avoidance, topic avoidance, extreme-positive language, and QoQ confidence trend. The model is based on Larcker & Zakolyukina's 2010 paper showing later-restated companies exhibit distinct linguistic markers on their calls. Scores above 60 = suspicious, above 75 = high alert.
Methodology.
Signals are extracted from public earnings call transcripts and IR press releases, then tagged by signal type (R&D investment, partnership, competitive threat, supply chain, capital shift, forward guidance) and target vertical (one of 11 frontier-tech categories). Every quote is verbatim and linked to its source transcript — we never paraphrase or synthesize. Primary sources: SEC EDGAR, company IR pages, Motley Fool transcripts, and Benzinga. The pipeline runs monthly and is augmented by LLM-assisted extraction (Claude) when the API key is configured.