Every entity in tech news has a coverage trajectory. Some are accelerating — more articles each week than the last. Some are decelerating — still covered, but less than before. And some are at inflection points — the rate of change itself is changing. Momentum detection separates these states using the same physics that describes moving objects: velocity, acceleration, and jerk.
Three Metrics
Velocity is the rate of coverage. If OpenAI appears in 50 articles this week, its velocity is 50. Velocity tells you who's in the news right now. But high velocity can mean an entity is consistently covered (Nvidia) or experiencing a one-time spike (a company in a scandal). Velocity alone doesn't distinguish between the two.
Acceleration is the change in velocity. If OpenAI appeared in 30 articles last week and 50 this week, its acceleration is positive — coverage is increasing. If it appeared in 70 last week and 50 this week, acceleration is negative — still heavily covered, but declining. Acceleration tells you the direction of the trend.
Jerk is the change in acceleration. If an entity's coverage was steady, then suddenly started accelerating, jerk is high. This is the earliest signal of a trajectory change — it detects the moment something shifts before the shift becomes obvious in velocity.
What Momentum Detects
Emerging entities. A company with low velocity but high acceleration is emerging — few articles so far, but each week brings more. TEXXR's emerging entity detection surfaces these before they hit mainstream coverage saturation.
Fading entities. High velocity but negative acceleration means an entity is still prominent but losing attention. This is the pattern of a story winding down — coverage continues by inertia, but the trajectory is downward.
Inflection points. A sudden change in jerk — from steady to accelerating, or from accelerating to decelerating — marks an inflection. These often correspond to product launches, scandals, leadership changes, or structural shifts.
Momentum vs. Signal Detection
TEXXR's Pulse signal detection uses z-scores to identify daily anomalies — entities getting far more or less coverage than their 30-day average. Momentum operates on a longer timescale. A z-score spike is a one-day event. Momentum tracks whether that spike is part of a trend or an outlier.
The strongest signals occur when both systems fire: a z-score spike (today is unusual) combined with positive acceleration (coverage has been building). That combination typically indicates a structural story, not just a news cycle.
Try It
Entity profiles on TEXXR (e.g., /entity/OpenAI) include momentum charts showing velocity and acceleration over time. The Pulse API provides momentum data via /api/momentum?entity=OpenAI.
For comparative momentum — tracking how two entities' trajectories relate — see the How TEXXR Works guide or use the CLI: python3 cli/texxr.py momentum OpenAI --compare Anthropic.