@verbose_explorer usage observations
context: @verbose_explorer runs 875 threads with 83% resolution rate. 231 spawned subagents complete at 97.8%. this document summarizes patterns observed in the data.
patterns observed
spawn orchestration
231 subagents at 97.8% success. @verbose_explorer effectively parallelizes work across agents.
long thread commitment
resolution rates by thread length:
- 1-15 turns: ~10-15% resolution (many are likely spawn completions)
- 16-60 turns: 40-62% resolution
- 100+ turns: 78% resolution
when @verbose_explorer stays in a thread, resolution is high.
context provision
932-char average opener. file references in openers correlate with +25% success (66.7% vs 41.8%).
nix domain
70% success rate in nix-related work.
meta-work focus
skills, tooling, infrastructure — successful threads cluster in these areas.
data points
approval frequency
0.55 approvals/thread vs @concise_commander’s 1.54. @verbose_explorer’s resolution rate (83%) is higher than @concise_commander’s (60.5%), so the impact of approval frequency is unclear.
evening patterns (uncertain)
19:00-22:00 shows lower resolution rates. possible explanations:
- exploratory work in evening (different task type)
- fatigue effects
- sample size or confounding factors
insufficient evidence to make a causal claim.
successful thread examples
T-048b5e03— debugging migration script (988 turns, 3 steers, 14 approvals) → RESOLVEDT-5ac8bb63— coordinate sub-agents for roadmap (466 turns, 4 steers, 13 approvals) → RESOLVEDT-c7c1489c— refactor list component (433 turns, 1 steer, 3 approvals) → RESOLVED
pattern: complex work, sustained engagement, periodic approvals, minimal steering.
metrics
| metric | value |
|---|---|
| resolution rate | 83% |
| avg approvals/thread | 0.55 |
| spawn success rate | 97.8% |
| total threads | 875 |
| spawned subagents | 231 |
compiled from thread analysis | corrected 2026-01-09