← charleslikesdata.com afk cockpit
11 repos · 1,324 runs · 1,085 issues · 962 merged · data through Jun 7 – Sep 16, 2026

Autonomous-PR outcomes across the fleet. Cost is API-equivalent (notional on a Pro subscription). Quarantines are a safety signal — afk refused to ship — not failures.

Fleet state · not yet wired up
Live gate/promote/merge and dependency-frontier state isn't ingested yet — this panel is a placeholder for a future feature, not a fabricated number.
97%
landed rate
n = 989 gate-passed
97%
landed rate (executor repos)
n = 11 exec repos
$4.11
API-equiv $ / merged PR
n = 962 merged
fully loaded: includes failed runs, quarantines, and scout overhead
33%
quarantine (safety) rate
n = 1,085 issues
17%
escalation rate
n = 1,324 runs
0%
rejected rate
n = 2,790 proposals
Pipeline

Stage funnel

proposed
2,790
promoted
1,112
40% of proposed
attempted
1,085
98% of promoted
passed gate
989
91% of attempted
published
968
98% of passed gate
merged
962
99% of published
By repository

Outcomes per repo

landed quarantined other
Orbit-Wars
151 · 85%
afk-agent-system
550 · 90%
afk-cockpit
106 · 94%
breakroom
37 · 84%
graphmark
15 · 93%
household-bms
52 · 92%
household-command-center
2 · 50%
housing-commute-analysis
15 · 93%
kaggriculture
22 · 91%
ragmark
9 · 89%
the-workshop
126 · 80%
Producers

Proposal quality

dropped
257
harvest
4
mis-sized
198
pending
1217
rejected
1
shipped
1112
superseded
1
Safety axis

Quarantine by category

354 total
other
86
no-op
46
review
34
scope
32
capability
22
integration-conflict
20
circuit-breaker
19
build-crash
17
deny-surface
14
auth
10
environmental
10
setup-crash
10
test-conflict
10
timeout
8
coupling
4
idle-timeout
4
unrecoverable-signature
3
flaky-test
2
security
2
rate-limit
1
Retries & recovery

Runs per issue

1.22 avg runs/issue
12345
Retries & recovery

Quarantine → retry rescue

retry rescues 62% of first-run quarantines · 103% of those actually retried
74%
first-try pass rate
14%
final-state quarantine (issues)
27%
quarantine rate (runs)
Cost · API-equivalent
fleet telemetry · per attempt · Jun 7 – Sep 16, 2026

Spend by model tier

$3,215.37 total across the window
sonnet$2235.17 · avg $1.71/attempt · n=1307 attempts
unknown tier$670.16 · avg $1.69/attempt · n=397 attempts
opus$310.04 · avg $2.37/attempt · n=131 attempts
⚠ 21% of attempt spend ($670.16 · 397 attempts) has no recorded model_tier — telemetry gap, not a model
Overhead · orchestrator + scout ops
$734.02
849 cost events · additive to the tier spend above · not a model tier
Claude Code transcripts, afk worktrees only · per API call · Aug 17 – Sep 16, 2026
$0.03
median / call
$0.06
p90 / call
1,852
calls logged
whole machine, all projects: $6,860.79 across 36,607 calls
Cost dial

Per-API-call cost distribution

1568
198
54
22
4
5
1
<$.05.05–.10.10–.15.15–.20.20–.30.30–.50$.50+
Model mix

Opus share by day

Aug 17 range: 0%–47% Sep 16
Throughput

Slices per day

1,324 runs
06-0706-1706-2907-1407-2808-0708-1708-3009-12
Context cost

Tokens by type, per day

123.4M tokens · 95.7% cache reads — the context is overwhelmingly cached
cache read cache write input output