← charleslikesdata.com afk cockpit
6 repos · 900 runs · 732 issues · 636 merged · data through Jun 7 – Aug 1, 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.
98%
landed rate
n = 648 gate-passed
98%
landed rate (executor repos)
n = 6 exec repos
$3.97
API-equiv $ / merged PR
n = 636 merged
fully loaded: includes failed runs, quarantines, and scout overhead
36%
quarantine (safety) rate
n = 732 issues
19%
escalation rate
n = 900 runs
0%
rejected rate
n = 1,242 proposals
Pipeline

Stage funnel

proposed
1,242
promoted
746
60% of proposed
attempted
732
98% of promoted
passed gate
648
89% of attempted
published
637
98% of passed gate
merged
636
100% of published
By repository

Outcomes per repo

landed quarantined other
Orbit-Wars
67 · 87%
afk-agent-system
387 · 89%
afk-cockpit
106 · 94%
household-bms
48 · 96%
the-vault
11 · 0%
the-workshop
113 · 79%
Producers

Proposal quality

dropped
240
harvest
5
mis-sized
54
pending
195
rejected
1
shipped
746
superseded
1
Safety axis

Quarantine by category

261 total
other
82
no-op
40
review
22
capability
20
integration-conflict
17
build-crash
16
scope
11
setup-crash
10
test-conflict
8
circuit-breaker
7
deny-surface
7
timeout
6
coupling
4
idle-timeout
4
unrecoverable-signature
3
flaky-test
2
security
2
Retries & recovery

Runs per issue

1.23 avg runs/issue
12345
Retries & recovery

Quarantine → retry rescue

retry rescues 57% of first-run quarantines · 99% of those actually retried
72%
first-try pass rate
17%
final-state quarantine (issues)
29%
quarantine rate (runs)
Cost · API-equivalent
fleet telemetry · per attempt · Jun 7 – Aug 1, 2026

Spend by model tier

$2,227.81 total across the window
sonnet$1360.87 · avg $1.58/attempt · n=862 attempts
unknown tier$674.64 · avg $1.69/attempt · n=399 attempts
opus$192.31 · avg $2.47/attempt · n=78 attempts
⚠ 30% of attempt spend ($674.64 · 399 attempts) has no recorded model_tier — telemetry gap, not a model
Overhead · orchestrator + scout ops
$294.81
243 cost events · additive to the tier spend above · not a model tier
Claude Code transcripts, afk worktrees only · per API call · Jul 2 – Aug 1, 2026
$0.03
median / call
$0.07
p90 / call
32,587
calls logged
whole machine, all projects: $8,427.75 across 67,336 calls
Cost dial

Per-API-call cost distribution

25703
5707
879
191
83
19
5
<$.05.05–.10.10–.15.15–.20.20–.30.30–.50$.50+
Model mix

Opus share by day

Jul 2 range: 0%–62% Aug 1
Throughput

Slices per day

900 runs
06-0706-1306-1906-2707-0307-1407-2307-30
Context cost

Tokens by type, per day

2.31B tokens · 95.8% cache reads — the context is overwhelmingly cached
cache read cache write input output