A
Corpus integrity
SEC replay on filing dates, frozen fixture corpora for stress events, and live ingest with temporal gates at every layer.
Research
Narrative dynamics as a measurable state-space — between NLP research and market microstructure.
Financial markets do not move on prices alone. They move on the stories institutions tell themselves — earnings narratives, macro framing, regulatory language, and cross-source disagreement in real time.
Our work treats that language as geometry: documents embedded, windowed, and scored for coherence, semantic drift, instability composites, and regime archetypes. The output is observatory evidence — not sentiment polarity, not LLM summaries, and not directional forecasts.
Methodology
B
Metrics v2
Rolling windows produce primitives (coherence, drift, persistence) and composites (instability, velocity) with documented definitions.
C
Observatory memory
Findings append to longitudinal stores — analog chains link current narrative shape to historical precedents.
D
Validation discipline
Forward stress alignment is reported as percentile context against an entity's own history — restrained, descriptive, reproducible.
Systems
Research outputs ship as operational systems — CLI terminals, observatory monitors, and annotated backtest replay. The flagship implementation is the Narrative Volatility Engine.