XCurve reads four surfaces, discards attention spikes, scores what survives and returns a dated recommendation. Here is each stage in detail.
Analog ADHD systems
Publish long-form video first. Short-form is saturated. Searchable 12 to 20 minute walkthroughs have almost no quality supply.
A concept is tracked as one object across community posts, short-form video, search intent and retail rank. Movement confined to a single surface is treated as noise. Movement that migrates between surfaces is the predictive event.
Curves are held against a decay test. A spike that loses its slope within six weeks of peak attention is discarded regardless of how large the peak was. What remains is behaviour that outlived the moment that caused it.
Four inputs produce the score: volume trajectory, supply density, confirmed cross-surface migration and decay risk. A high score means the evidence is consistent, not that the outcome is guaranteed.
Output is a specific move, the reason it is on the table, the window in weeks and the sources behind it. Sources are named so you can check the reasoning rather than take the score on trust.
Roughly a third of tracked niches return do not enter. A tool that never tells you to stay out is a marketing tool. Closed and unresolved verdicts appear in the explorer alongside the open ones.
Keyword tools cannot see a term change shape. Tracking the concept means the signal survives renaming, which is exactly when early categories are most valuable and least legible.
Every brief lists the surfaces it drew from. You can disagree with a conclusion and still use the underlying evidence. Scores without provenance are not auditable and should not be trusted.
A notification tells you something moved. A window tells you how long the move stays actionable, which is the only part that affects whether you commit inventory or production time.