no black box

How to read the signals

Most sentiment tools hand your agent a number with no way to judge it. signaldaemon ships two derived signals plus an honesty rule. Here is exactly what each one means, what moves it, and how to read it — so your agent (or you) can trust it or discount it deliberately.

SIGNAL 01 · NARRATIVE STRENGTH = cross-source convergence

How many independent sources are converging on this story — right now

A single headline is noise. A narrative is many independent, trust-weighted sources landing on the same story at the same time. strength scores that convergence. It is the one number a single web search can't give you, and it's hard to fake — you can't manufacture independent corroboration.

What moves it (the inputs, not the recipe):

Independent source count — how many distinct sources carry the story (clustered across outlets, not one outlet repeated).
Source trust-weighting — sources are not equal; a trusted source counts for more than a content farm.
Recency — a story corroborated in the last hours weighs more than a stale one.
Cross-domain breadth — a story spanning several of our ~14 domains (e.g. protocol + macro + regulation) is structurally stronger than a single-niche one.

How to read it: higher = more corroborated by independent, trusted sources now. Rising momentum.members_24h = the story is accelerating. It is not a count of social mentions, and it is not a price prediction.

We publish the factors, not the exact weights or the source list — those are the proprietary part. You get enough to interpret the number; you don't get a recipe to reproduce or game it.

SIGNAL 02 · CAPITAL DIVERGENCE = narrative vs. price, two signed axes

Is the story matched by capital — and is it leading or lagging the market?

We cross narrative strength against real price/flow and return two independent signed axes — never inferred from each other:

direction ∈ up / down — the asset's absolute 7-day move.
vs_market ∈ outperform / underperform — the move relative to the market baseline (market_7d, shipped in every payload).

How to read it: in a crash an asset can be direction=down AND vs_market=outperform — it fell slower than the market, which is relative strength, not a bullish price move. Report both. narrative_no_flow = strong story, capital not following (yet). narrative_price_aligned = story matched by capital.

Honest status: we treat divergence as a structured observation, not a proven trading edge. We're backtesting whether divergence shifts precede forward returns and controlling for price momentum before we claim any predictive power. Use it as context and risk-framing, not as a buy/sell signal — until the evidence is in, which we'll publish either way.

THE RULE · FAIL SAFE

It says "I don't know" instead of inventing

When a narrative has no single tradeable asset (politics, enforcement, an IPO spillover) the payload says no_asset — we report the story and refuse to invent a ticker. When coverage of a topic is thin, the feed says so (coverage: "thin"). A signal you can trust is one that tells you what it doesn't know; a tool that always emits a number is hiding its gaps.

Pull all of this as structured data over REST or MCP. See it live across 14 domains on the narrative tracker.