Network analysis of catch's 14,042 commits — collaborator topology,
revert patterns, subsystem distribution, and the act-then-explain dynamic.
14,042 commits analyzed
636 commits with co-authors
404 reverts issued
8,608 cross-referenced issues
01Co-author Network
02Revert Analysis
404
Total reverts issued
2.88% of all commits
79.5%
Unspecified reason
321 of 404 — no explanatory note
63
Test failures
15.6% — CI-driven reverts
12
Re-reverts
Re-landing previously reverted work
Reverts by reason
Reverts by subsystem (top 10)
Notable reverts — 20 selected
Date
Subject
Reason
Subsystem
03Subsystem Comparison
Commits vs. comments by subsystem (% of total, excluding "other")
Codes more than he comments: deprecation (1.97% commits, 0% comments), security (1.54% commits, 0% comments), render_pipeline (0.93% commits, 0% comments).
These are execution-mode subsystems — catch lands the work without extensive prior discussion.
Comments more than he commits: theming (1.87% commits, 16.87% comments), routing (2.16% commits, 8.07% comments), entity_system (7.48% commits, 20.96% comments).
These are contested-design subsystems where catch shapes decisions through review before landing code.
04Act-then-Explain Pattern
61.3%
Commit-first
catch committed to an issue before leaving a comment in the queue.
He acts, then explains — or often does not explain at all. The commit is the communication.
5,279 of 8,608 shared issues
38.6%
Comment-first
catch commented before committing — typically on contested design decisions,
blocking issues requiring architectural alignment, or issues he did not author.
3,325 of 8,608 shared issues
Behavioral flow — two modes
catch
commit
61.3% of cases
then (maybe)
comment
explains or
outcome
patch lands
catch
comment
38.6% of cases
reviews / aligns
discussion
outcome
commit follows
Deepest engagement — top issues by comment + commit density