Analysis Overview

Cadence performs three types of analysis to detect suspicious or AI-generated content:

Repository Analysis

Analyzes Git repositories for suspicious commit patterns that may indicate AI-generated or low-effort code. This includes examining:

• Commit metadata - authors, timestamps, commit frequency
• Code changes - file counts, line additions/deletions, change ratios
• Patterns - velocity anomalies, timing irregularities, structural consistency
• Content - commit messages, variable names, code structure

See Repository Analysis for detailed strategy information.

Git Analysis

Deep dive into Git commit history using 18 detection strategies. Cadence examines commit patterns across time to identify statistical anomalies and behavioral indicators of AI-generated code.

Detection strategies analyze:

• Velocity - additions/deletions per minute
• Timing - intervals between commits, burst patterns
• Size - individual commit sizes and consistency
• Structure - file dispersion, naming patterns, code organization
• Content - commit messages, error handling, code quality indicators

See Git Analysis for complete strategy documentation.

Web Content Analysis

Analyzes website content and text for patterns common in AI-generated or low-quality content ("slop"). Examines:

• Language patterns - overused phrases, generic terminology, missing specificity
• Grammar - suspiciously perfect grammar, uniform sentence structure
• Content - placeholder patterns, boilerplate text, lack of nuance
• Structure - excessive formatting, templated organization, repetitive sections

See Web Analysis for detection pattern details.

Start with Repository Analysis to understand how Cadence evaluates code quality, or jump to Git Analysis for detailed commit-level detection strategies.