Structural metrics remain the strongest explanatory layer.
Findings
Structure remains the strongest explanatory layer.
The study found widespread AI-related patterns, but the clearest distinctions still came from concentration, redundancy, contributor breadth, and historical participation.
Higher contribution inequality generally corresponds to lower effective contributor redundancy.
AI-related patterns are widespread but are not the primary structural differentiator.
Historical long-tail participation remains important for resilience.
The traffic light measures confidence in the reading, not repository safety.
Research Visuals
Contribution Inequality and Redundancy
Higher redundancy
Higher concentration
Distributed participation
Concentrated dependency
The paper reports that higher contribution inequality is generally associated with lower contributor redundancy.
Caption: This visual summarizes the reported relationship. It does not redraw repository-level scatter points because original chart assets or underlying per-repository data were not present in this site repository.
Traffic-Light Distribution
Most fast-mode readings landed in the green confidence class, with a smaller yellow and orange share and no red class in the reported table.
Caption: The traffic light describes confidence in the structural reading. It does not say that green repositories are secure or that orange repositories are unsafe.
Inferred AI Influence
≈0.453
median inferred AI influence
AI influence signals appear widespread in the analyzed set, but the study does not treat them as direct authorship proof.
Caption: The score is an inferred, reproducible analytical layer. It should not be read as evidence that specific commits were written by AI.
AI behaves more like an ambient capability layer than a primary structural driver.