← charleslikesdata.com afk cockpit
11 repos · 1,455 runs · 1,168 issues · 1,012 merged · data through Jun 7 – Oct 29, 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.
94%
landed rate
n = 1,071 gate-passed
94%
landed rate (executor repos)
n = 11 exec repos
$4.46
API-equiv $ / merged PR
n = 1,012 merged
fully loaded: includes failed runs, quarantines, and scout overhead
35%
quarantine (safety) rate
n = 1,168 issues
16%
escalation rate
n = 1,455 runs
0%
rejected rate
n = 3,835 proposals
Pipeline

Stage funnel

proposed
3,835
promoted
1,217
32% of proposed
attempted
1,168
96% of promoted
passed gate
1,071
92% of attempted
published
1,014
95% of passed gate
merged
1,012
100% of published
By repository

Outcomes per repo

landed quarantined other
Orbit-Wars
188 · 70%
afk-agent-system
568 · 91%
afk-cockpit
106 · 94%
breakroom
37 · 92%
graphmark
15 · 93%
household-bms
52 · 94%
household-command-center
2 · 50%
housing-commute-analysis
15 · 93%
kaggriculture
22 · 91%
ragmark
28 · 96%
the-vault
0 · 0%
the-workshop
135 · 78%
Producers

Proposal quality

dropped
717
harvest
5
mis-sized
320
pending
1574
rejected
1
shipped
1217
superseded
1
Safety axis

Quarantine by category

408 total
other
86
environmental
48
no-op
48
review
36
scope
33
capability
23
integration-conflict
20
circuit-breaker
19
build-crash
18
deny-surface
14
auth
10
setup-crash
10
test-conflict
10
rate-limit
8
timeout
8
idle-timeout
5
coupling
4
flaky-test
3
unrecoverable-signature
3
security
2
Retries & recovery

Runs per issue

1.25 avg runs/issue
123457
Retries & recovery

Quarantine → retry rescue

retry rescues 56% of first-run quarantines · 88% of those actually retried
73%
first-try pass rate
14%
final-state quarantine (issues)
28%
quarantine rate (runs)
Cost · API-equivalent
fleet telemetry · per attempt · Jun 7 – Oct 29, 2026

Spend by model tier

$3,339.26 total across the window
sonnet$2338.94 · avg $1.63/attempt · n=1436 attempts
unknown tier$670.16 · avg $1.69/attempt · n=397 attempts
opus$330.16 · avg $2.28/attempt · n=145 attempts
⚠ 20% of attempt spend ($670.16 · 397 attempts) has no recorded model_tier — telemetry gap, not a model
Overhead · orchestrator + scout ops
$1,173.36
1280 cost events · additive to the tier spend above · not a model tier
Claude Code transcripts, afk worktrees only · per API call · Sep 29 – Oct 18, 2026
$0.03
median / call
$0.09
p90 / call
1,188
calls logged
whole machine, all projects: $12,203.00 across 72,221 calls
Cost dial

Per-API-call cost distribution

785
309
66
14
9
5
0
<$.05.05–.10.10–.15.15–.20.20–.30.30–.50$.50+
Model mix

Opus share by day

Sep 29 range: 0%–32% Oct 18
Throughput

Slices per day

1,455 runs
06-0706-2407-1408-0208-1709-0509-2410-09
Context cost

Tokens by type, per day

98.7M tokens · 95.1% cache reads — the context is overwhelmingly cached
cache read cache write input output