Discovery: Overview of traces (stats, timeline, transcript list)
Analysis: Line-level model attribution—the centerpiece (blame, blame –blocks, transcript view)
Output: Preview and apply annotations (report, annotate –dry-run)
Real example: 80 trace files showing 378 edits across the ai-blame.rs project, with Haiku and Sonnet models collaborating. Shows which AI model wrote each line of code, when it was edited, and how to embed that provenance back into files.
All commands shown in dry-run mode—safe to explore before applying changes.