ten hours of an agent fleet working across a personal portfolio, from the run's own audit record
sanitized aggregates · no repository names · merged during the window, not credited to the run · not a certification
window duration
10h 3m
projects
269
distinct merged PRs
654
peak concurrency slots
129
Codex sessions started
921
The timeline
10:00 PM Oct 2 to 8:03 AM Oct 3 PDT · merged during the window · session starts, not active concurrency
first fleet
second fleet
dependency bots
me
other
Table view
Ten-minute bins, PDT; the first and last bins are clipped to the audit window
Bin start PDT
first fleet merges
second fleet merges
dependency bots merges
me merges
other merges
first fleet starts
second fleet starts
10:00 PM
1
0
46
4
0
65
0
10:10 PM
7
0
123
4
0
33
0
10:20 PM
4
0
13
3
0
8
0
10:30 PM
5
0
14
1
0
4
0
10:40 PM
1
0
31
1
0
0
0
10:50 PM
1
0
0
1
0
0
0
11:00 PM
5
0
4
0
0
1
0
11:10 PM
7
0
5
0
0
3
0
11:20 PM
1
0
1
0
1
2
0
11:30 PM
0
0
0
0
7
0
0
11:40 PM
0
0
0
0
3
1
0
11:50 PM
0
0
0
0
1
5
0
12:00 AM
0
0
0
0
0
6
0
12:10 AM
0
0
0
0
0
0
0
12:20 AM
0
0
0
0
0
0
0
12:30 AM
10
0
0
12
0
0
0
12:40 AM
32
0
0
30
0
2
0
12:50 AM
34
0
0
17
0
0
0
1:00 AM
2
0
0
1
0
4
0
1:10 AM
6
0
0
2
0
0
0
1:20 AM
4
0
0
1
0
5
0
1:30 AM
3
0
0
0
0
6
0
1:40 AM
0
0
0
0
0
32
0
1:50 AM
0
0
0
3
0
22
0
2:00 AM
1
0
0
0
0
45
0
2:10 AM
0
0
0
0
0
5
0
2:20 AM
0
0
0
1
0
22
0
2:30 AM
0
0
0
1
0
8
0
2:40 AM
2
0
0
0
0
17
0
2:50 AM
0
0
0
0
0
16
0
3:00 AM
1
0
0
1
0
11
0
3:10 AM
0
0
0
0
0
0
0
3:20 AM
1
0
1
0
0
0
0
3:30 AM
0
0
0
0
0
11
12
3:40 AM
0
0
0
1
0
28
0
3:50 AM
0
0
0
0
0
0
0
4:00 AM
0
0
0
0
0
21
0
4:10 AM
0
0
1
1
0
0
0
4:20 AM
0
0
0
0
0
42
0
4:30 AM
0
0
0
0
0
1
19
4:40 AM
0
0
0
0
0
0
0
4:50 AM
0
0
0
0
0
6
0
5:00 AM
0
0
0
0
0
62
35
5:10 AM
0
0
0
0
0
32
47
5:20 AM
0
0
0
0
0
31
19
5:30 AM
5
0
0
2
0
12
5
5:40 AM
0
0
0
0
0
0
27
5:50 AM
0
0
0
0
0
0
0
6:00 AM
0
9
0
0
0
51
10
6:10 AM
4
72
11
0
0
25
0
6:20 AM
6
5
33
0
0
11
0
6:30 AM
1
6
12
1
0
20
0
6:40 AM
1
1
4
0
0
9
0
6:50 AM
1
0
2
0
0
0
0
7:00 AM
1
1
1
0
0
0
0
7:10 AM
1
2
0
0
0
1
0
7:20 AM
2
2
3
0
0
0
0
7:30 AM
1
0
0
0
0
35
0
7:40 AM
0
0
0
0
0
14
0
7:50 AM
0
0
0
0
0
8
0
8:00 AM
0
0
0
0
0
4
0
Phase boundaries, PDT
Phase
Start
End
Inventory and sync
10:00 PM
12:28 AM
Verification guidance
12:28 AM
1:47 AM
Cloud readiness audit (side)
2:00 AM
2:13 AM
Authorized next steps
2:15 AM
4:35 AM
Organization review-rule change
4:52 AM
5:45 AM
Backlog triage
5:56 AM
8:03 AM
Debt at merge
Dependency pull requests merged during the window, by how long they had been open.
dependency pull request
median
Table view
Dependency age, one row per pull request
Ecosystem
Days waiting
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0
npm
0.1
npm
0.1
npm
0.2
npm
0.2
npm
0.3
npm
0.3
npm
0.9
npm
0.9
npm
1
npm
1
npm
3.6
npm
3.6
npm
3.6
npm
3.8
npm
3.8
npm
4.1
npm
4.2
npm
4.2
npm
4.4
npm
4.5
npm
4.6
npm
4.6
npm
4.6
npm
4.9
npm
5
npm
5
npm
5.1
npm
5.2
npm
5.6
npm
11.5
npm
11.7
npm
11.8
npm
11.9
npm
11.9
npm
18.7
npm
19.4
npm
19.6
npm
19.7
npm
19.7
npm
21.1
npm
21.1
npm
21.1
npm
21.2
npm
22.1
npm
22.1
npm
22.2
npm
22.2
npm
22.2
npm
22.2
npm
22.2
npm
22.3
npm
22.3
npm
22.3
npm
22.4
npm
22.4
npm
22.5
npm
22.9
npm
22.9
npm
23
npm
23
npm
23.1
npm
23.1
npm
23.2
npm
23.2
npm
23.2
npm
23.3
npm
23.4
npm
23.4
npm
23.4
npm
23.4
npm
23.4
npm
23.8
npm
23.8
npm
23.8
npm
23.8
npm
24.7
npm
24.7
npm
24.7
npm
24.7
npm
24.8
npm
24.8
npm
24.8
npm
24.8
npm
24.8
npm
24.9
npm
24.9
npm
24.9
npm
25
npm
25.3
npm
25.3
npm
25.3
npm
25.3
npm
25.3
npm
25.5
npm
25.5
npm
25.5
npm
25.6
npm
25.6
npm
25.6
npm
26
npm
26.2
npm
26.7
npm
26.9
npm
26.9
npm
27
npm
27.2
npm
27.2
npm
27.2
npm
27.2
npm
27.2
npm
27.5
npm
27.5
npm
27.5
npm
27.5
npm
27.6
npm
30
npm
30.1
npm
30.8
npm
30.8
npm
39.6
npm
40.5
npm
40.5
npm
40.5
npm
40.5
npm
40.8
npm
40.8
npm
40.8
npm
40.8
cargo
0
cargo
0
cargo
2.4
cargo
2.4
cargo
2.4
cargo
3.9
cargo
3.9
cargo
4.3
cargo
4.6
cargo
4.7
cargo
4.8
cargo
5.1
cargo
5.7
cargo
5.7
cargo
5.8
cargo
5.8
cargo
5.8
cargo
5.8
cargo
5.9
cargo
5.9
cargo
12.8
cargo
15.6
cargo
16.3
cargo
19.8
cargo
19.8
cargo
19.9
cargo
20.8
cargo
20.8
cargo
20.8
cargo
20.8
cargo
20.8
cargo
24.9
cargo
26.2
cargo
26.7
cargo
40.5
cargo
40.5
cargo
40.5
cargo
40.5
cargo
40.7
cargo
40.7
cargo
40.7
cargo
40.7
cargo
40.7
cargo
40.8
cargo
40.8
cargo
40.8
cargo
40.8
cargo
40.8
github-actions
1
github-actions
3.6
github-actions
3.9
github-actions
4.5
github-actions
4.5
github-actions
4.8
github-actions
4.9
github-actions
6
github-actions
6.2
github-actions
7.1
github-actions
12.9
github-actions
13.1
github-actions
19
github-actions
19.2
github-actions
20.1
github-actions
21.1
github-actions
25.5
github-actions
32.5
github-actions
69.9
github-actions
69.9
github-actions
70
github-actions
70.2
github-actions
104.9
uv
0.9
uv
1
uv
1
uv
14.1
uv
14.1
uv
14.1
uv
14.1
uv
14.1
uv
14.1
uv
14.2
uv
14.2
uv
14.3
uv
14.5
uv
14.5
uv
15
uv
29
uv
30
pip
0.3
pip
2.8
pip
2.8
pip
4.3
pip
4.5
pip
5
pip
18.6
pip
18.6
pip
18.6
pip
23.8
pip
25.5
go
18.3
Row summaries
Ecosystem
Count
Median days
npm
141
22.9
cargo
48
19.8
github-actions
23
13.1
uv
17
14.1
pip
11
5
go
1
18.3
Who opened what, and where it ended up
Pull requests captured for the final phase, by who opened them, and how each one ended up.
first fleet
second fleet
dependency bots
me
MERGED
CLOSED
OPEN
Table view
Captured backlog states by opener
Opener
MERGED
CLOSED
OPEN
Total
first fleet
7
0
2
9
second fleet
97
0
3
100
dependency bots
66
39
77
182
me
1
0
4
5
Project outcomes
Original primary outcome; preserved work can overlap any outcome.
outcome count
with preserved work
Table view
Project outcomes and overlapping preserved work
Outcome
Count
With preserved work
synchronized
124
49
verified unchanged
73
9
draft preserved
58
49
published pending gate
9
2
blocked
4
4
intentionally unpublished
1
0
Task length by effort
Elapsed time to checkpoint; not CPU time or exact execution time.
Table view
Elapsed task time to checkpoint, one row per task
Effort
Minutes
high
2
high
2
high
3
high
3
high
3
high
3
high
3
high
3
high
3
high
4
high
4
high
4
high
4
high
4
high
5
high
6
high
7
high
9
high
10
high
10
high
10
high
11
high
12
high
17
high
33
high
34
high
41
high
41
high
41
high
41
high
42
high
42
high
44
high
52
high
55
high
62
high
62
high
63
high
63
high
64
high
72
high
78
high
78
high
83
high
92
high
96
high
100
high
103
high
117
extra high
3
extra high
4
extra high
6
extra high
8
extra high
14
extra high
20
extra high
26
extra high
40
extra high
61
extra high
75
Languages
One primary language per maintained repository; unknown labels aggregated.
Table view
Top eight named languages; Other includes Unknown and remaining languages
Language
Count
Python
73
TypeScript
60
Rust
36
Swift
16
JavaScript
7
Shell
7
HTML
3
Makefile
3
Other
54
What broke
Early in the run. A git setting chosen years ago pruned remote-tracking refs in two repositories. The prior commit identities are recorded as UNKNOWN.
11:06 PM. One project's worktrees vanished for about forty minutes. Seven untracked files came back byte-for-byte from a backup. Who removed them is UNKNOWN.
11:43 PM. An iOS test job sat in progress for nearly five hours, never failing. The rerun passed. The cause is UNKNOWN.
Overnight. A framework bump came in at plus 15.5 percent against an unchanged 8 percent bundle cap. It was held, not waived, and later landed at plus 3 percent.
5:13 AM. A review rule needed a second human in a one-person organization. Required approvals went from one to zero, by decision, on the record.
About 7:40 AM. Two fleets shared one GitHub API quota. It ran out.
Still open. One project's dependency security audit fails on its default branch. Not fixed, because that was not authorized. Not waived, because that would not be true.
What these numbers are not
Merges are merged during the run window, not attributed as merged by the run.
Model identity at runtime was not exposed.
Session counts include every Codex session started on this machine inside the window, including the second fleet's headless dispatches; exec sessions are split by launch directory.
Cloud rollout was still running and is excluded from audit-derived series; the machine session census includes its starts inside the window.
The recorded audit window is about ten hours; the broader sweep duration is not established by these inputs.
Phase labels follow time boundaries, not causal attribution; the cloud readiness audit overlaps other work.
Missing PR creation times yield null ages; unknown dependency ecosystems yield null labels and are omitted from dependency age.
Opener classes use branch prefixes before author identity; other includes missing or ambiguous metadata.
Captured backlog outcomes exclude task-derived PRs and later arrivals; merges include all scoped native merged identities.
Maintained repositories include archived repositories and owned forks in the maintained identity set.
Preserved-work counts are historical classifications, not proof that every byte remains recoverable.
Unparsable session headers are counted separately; their start times cannot be assigned to the window or bins.