Beyond Mention Counting: How Semantic Physics Detects Narrative Shifts

Editor’s note, October 2026: this post predates the launch, and parts of it no longer describe the product. We no longer call this “Semantic Physics”; it is part of the methodology. Narrative drift is still computed every day for the entities we track, but no agent fires on it, and at launch there is no public REST API and the MCP server does not serve drift. The argument stands.

Every media monitoring tool on the market does the same thing: count mentions. Entity X was mentioned 5,000 times this week, up 40% from last week. Alert triggered.

But volume doesn’t tell you what changed. A politician gets 10,000 mentions every week. A 40% spike could be a routine press conference or a career-ending scandal. Mention counting can’t distinguish between the two.

We built something different.

The Problem With Volume

Consider two scenarios:

Scenario A: A CEO is mentioned 2,000 times in a week. The topics are the same as last week — quarterly earnings, product launches, industry events. Nothing has changed about how this person is discussed. The volume is noise.

Scenario B: The same CEO is mentioned 2,000 times. But this week, the discourse has shifted from “innovation leader” to “under investigation.” The volume is identical. The narrative has completely changed.

Every existing tool treats these the same. Quantiveo doesn’t.

Semantic Physics: Position, Not Volume

We treat each entity’s public narrative as an object with a position in semantic space. Not a single number — a rich, high-dimensional representation of everything being said about them: the topics, the tone, the associations, the context.

Every week, we compute a new position. When consecutive positions are close together, the narrative is stable. When a position jumps, something real happened.

We call that jump narrative drift. And it’s a fundamentally different signal from mention volume.

What Drift Detection Catches

Our engine validated against 65 weeks of real-world data across thousands of entities. The top drift events consistently map to verifiable narrative shifts:

Celebrity loses home in wildfire — Narrative shifted from “entertainer” to “disaster victim/advocate.” Mentions spiked too, but drift captured the direction of the shift.
Coach fired mid-season — Narrative shifted from “playcaller” to “fired coach.” Volume alone wouldn’t tell you the narrative permanently changed.
Judge arrested in courtroom incident — Entity went from near-zero public awareness to national controversy. Both volume and drift spiked, but drift showed the narrative never returned to baseline.
Athlete in contract negotiations — Mentions were steady (sports coverage is always high), but drift detected the discourse shifting from “player performance” to “trade speculation.”

In every case, drift detected something that mention counting missed or couldn’t distinguish from noise.

Three Signals, Not One

Drift alone isn’t the full picture. We combine it with two additional signals:

Entropy — How many different topics is an entity being discussed across? High entropy means the narrative is fragmented and unpredictable. Low entropy means the discourse is focused. An entity with rising drift and rising entropy is in the middle of a narrative explosion.

Emergence — Our composite score that predicts which entities are about to break out. It combines narrative momentum, topic diversity, and historical volatility into a single “watch this entity” signal. This is what powers our “About to Blow Up” detection in FameBunny.

Why This Matters for AI Agents

AI agents need more than “entity X is trending.” They need to know what changed and in what direction. Our APIs return structured drift data, entropy scores, and emergence predictions that agents can reason about and act on.

An agent monitoring a portfolio of entities doesn’t need 10,000 mention alerts. It needs to know: “Entity X’s narrative shifted significantly this week. Here’s the drift score, the direction, and the confidence level.” That’s actionable intelligence.

Try It

We’re opening early access to the Quantiveo intelligence APIs. The drift detection, entropy scoring, and emergence prediction are all available through our REST API and MCP server.

Join the waitlist to be first in line. Founding members lock in 40% off for life.