01
Parallel npm audits cut doctor wall time in half
10/10 benchmark runs; focused 3/3; full suite red/stalled at 83%
result11.99 s median52.6% faster
→
// complete phase 1 index
Experiments
Every result here comes from an isolated Bunsen profile or worktree. Nothing on this site means a change has been promoted to Herman. Candidate status means the experiment produced evidence worth review—not permission to ship it.
02
Twenty-five isolated lifecycle cycles leave no survivors
25/25 lifecycle safety; workload compliance 22/25
resultNo process growth · 0 survivor cyclesgateway stop max 1.24 s
→
03
The web tool works; exact-format obedience is the weaker link
10 isolated fetches; corrected secret scanner PASS
result10/10 factual · 9/10 exact formatMedian 35.49 s
→
04
Content fingerprints close a rapid-config identity gap
10/10 synthetic identity routes plus full Honcho plugin suite
resultSHA-256 content key detects same-size, same-mtime edits424 passed
→
05
Fallback works and the usage receipt names the served model
Primary, direct fallback, and forced fallback all passed
resultFallback to minimax-m3 in 25.95 sCanonical config hash unchanged
→
06
Seven local patches were sorted before promotion
TTS/footer candidate; cron toggle held for repair
result7 apply to fork and upstream; 0 already landed65 targeted tests passed
→