An analytical framework for deriving how individuals drive outcomes from raw noise.
Each file becomes a read on what that person is likely to do next. Not an event probability — a ranked read on a named person's next move given their record and constraints, with analyst judgment always labeled as judgment.
The graph builds into living files on the people who make the decisions: what they've said, what they've done, and how they decide. Every file shares the same structure and is rebuilt whenever the evidence changes, tracking what people say against what they actually do.
Claims are resolved to the people, institutions, and relationships they touch. Nobody decides alone — mapping who surrounds each person lets pressure on one be read across everyone it touches.
Every item is broken into single claims, each tied to exactly where it came from. Those citations travel with every read, so any claim can be checked at its source.
Wires, filings, transcripts, market data, newsletters, and feeds stream in from thousands of sources. Established outlets only — no social media as primary evidence — and every source is rated for how far to trust it on the question being asked.
Why not just an LLM or Wikipedia
| Conspectus | LLMs | Wikipedia | |
|---|---|---|---|
| Cites every claim to a checkable source | ✓ | ✗ | ✓ |
| Rates how far to trust each source on the question being asked | ✓ | ✗ | ✗ |
| Models the incentives and relationships that drive each decision-maker | ✓ | ✗ | ✗ |
| Tracks stated positions against revealed behavior over time | ✓ | ✗ | ✗ |
| Adds working hypotheses from analysts who study the room, labeled as judgment | ✓ | ✗ | ✗ |
Tradecraft
Forecasting what these individuals will do is not the point. Conspectus provides the context surrounding a decision: who has influence, what they can affect, how they have behaved under pressure, and what observable indicators might signal a shift.