1
00:00:00,200 --> 00:00:06,526
Codex CLI 0.169.0 shipped on November
13th, and one line in the notes is easy to

2
00:00:06,586 --> 00:00:07,166
skim past.

3
00:00:07,787 --> 00:00:12,091
With the opt-in Cloud connection on macOS
and Linux, you can send images,

4
00:00:12,231 --> 00:00:16,255
choose a model and reasoning effort, and
archive or restore Cloud conversations,

5
00:00:16,515 --> 00:00:17,276
all from the terminal.

6
00:00:18,577 --> 00:00:19,257
That's what we're covering.

7
00:00:19,538 --> 00:00:21,880
And thanks to Jellypod AI for sponsoring
this episode.

8
00:00:22,507 --> 00:00:26,850
Last time we talked about the security and
directory trust changes in this same

9
00:00:26,870 --> 00:00:27,251
release.

10
00:00:28,091 --> 00:00:31,294
This is the other half of 0.169.0 worth
your time.

11
00:00:31,754 --> 00:00:37,118
The pull requests behind it are 59828,
60061,

12
00:00:37,638 --> 00:00:41,541
60635, and 60651.

13
00:00:42,201 --> 00:00:47,565
The first one, 59828, is titled "Add a
gated Cloud backend to the TUI."

14
00:00:48,366 --> 00:00:49,807
Gated is the key word there.

15
00:00:50,407 --> 00:00:51,988
So what changed, exactly?

16
00:00:52,889 --> 00:00:56,431
Before this, if you wanted to work with a
Codex Cloud conversation,

17
00:00:56,751 --> 00:00:57,692
you left the terminal.

18
00:00:59,133 --> 00:01:02,234
Now the terminal UI can talk to the Cloud
backend directly.

19
00:01:03,035 --> 00:01:07,918
The release notes name three things:
sending images, picking a model and reasoning

20
00:01:07,978 --> 00:01:11,100
effort, and archiving or restoring
conversations.

21
00:01:11,980 --> 00:01:17,744
There's also a related change, pull
request 60007, which starts cloud threads with

22
00:01:17,764 --> 00:01:19,025
the local GitHub project.

23
00:01:19,545 --> 00:01:22,087
So the thread begins tied to the repo
you're sitting in.

24
00:01:22,820 --> 00:01:23,821
Why does that matter?

25
00:01:24,782 --> 00:01:26,344
Mostly because of context switching.

26
00:01:26,824 --> 00:01:31,169
You can hand heavier or slower work to the
cloud, and still stay in the terminal to

27
00:01:31,209 --> 00:01:31,649
manage it.

28
00:01:32,149 --> 00:01:33,891
And the image part is the real gain.

29
00:01:34,272 --> 00:01:37,935
A screenshot of a broken layout, a
diagram, a visual bug.

30
00:01:38,416 --> 00:01:42,560
You can pass those into a cloud
conversation without opening a browser tab first.

31
00:01:43,195 --> 00:01:44,716
Now, how do you use it.

32
00:01:45,356 --> 00:01:46,357
Start by updating.

33
00:01:47,017 --> 00:01:52,100
Run npm install dash g at openai slash
codex at

34
00:01:52,340 --> 00:01:55,622
0.169.0.

35
00:01:55,622 --> 00:01:58,243
I'm saying that out loud because the
command has a version pin,

36
00:01:58,543 --> 00:02:00,464
and you want to be sure that's what you
actually get.

37
00:02:01,465 --> 00:02:03,846
After that, you need to turn on the Cloud
connection.

38
00:02:04,727 --> 00:02:08,829
I'll be straight with you: the release
notes say opt-in and gated,

39
00:02:09,109 --> 00:02:12,871
but they don't give the exact flag or
config key in the text I have.

40
00:02:13,411 --> 00:02:18,974
So check the 0.169.0 release page or your
in-app settings for the switch,

41
00:02:19,234 --> 00:02:20,495
rather than guessing at a name.

42
00:02:21,275 --> 00:02:23,958
Once it's on, the workflow is what the
notes describe.

43
00:02:24,518 --> 00:02:26,440
Send an image into a Cloud conversation.

44
00:02:26,940 --> 00:02:29,563
Choose the model and the reasoning effort
before you submit.

45
00:02:30,243 --> 00:02:33,767
Higher effort for the hard debugging task,
lower for a quick edit.

46
00:02:35,508 --> 00:02:38,311
And when a thread is stale, archive it
from the terminal.

47
00:02:38,891 --> 00:02:41,914
If you archived the wrong one, restore it,
same place.

48
00:02:42,374 --> 00:02:42,955
No dashboard.

49
00:02:43,712 --> 00:02:44,433
Caveats.

50
00:02:45,213 --> 00:02:46,614
First, platform.

51
00:02:47,194 --> 00:02:49,516
The Cloud connection is for macOS and
Linux.

52
00:02:49,996 --> 00:02:54,939
If you're on Windows, this particular
feature isn't there in 0.169.0.

53
00:02:56,400 --> 00:02:57,921
Second, it's opt-in.

54
00:02:58,502 --> 00:03:02,204
Nothing changes until you turn it on, so
if a teammate says they don't see it,

55
00:03:02,324 --> 00:03:03,345
that's probably why.

56
00:03:04,205 --> 00:03:06,607
Third, think before you attach images.

57
00:03:07,147 --> 00:03:10,990
Large images add to what you send, and I'd
expect that to matter for usage,

58
00:03:11,270 --> 00:03:14,212
though the notes don't spell out limits,
so check your own plan.

59
00:03:14,905 --> 00:03:16,869
Now a few smaller things in this release.

60
00:03:17,269 --> 00:03:20,655
You can search older tasks from the agent
command center by name,

61
00:03:21,135 --> 00:03:22,918
prompt preview, or working directory.

62
00:03:23,439 --> 00:03:26,864
That helps when you've got dozens of runs
and only remember the folder.

63
00:03:27,528 --> 00:03:33,615
Next, the "Yes, clear context and
implement" option now hands the full approved plan

64
00:03:33,715 --> 00:03:34,777
to the fresh session.

65
00:03:35,578 --> 00:03:38,761
And if that handoff fails, your current
session stays open.

66
00:03:39,002 --> 00:03:40,063
So you don't lose the plan.

67
00:03:40,924 --> 00:03:44,228
Anyone who's cleared context and then
regretted it will like that one.

68
00:03:44,923 --> 00:03:50,249
Also, in the terminal report editor, press
Control S to save a diagnostic ZIP

69
00:03:50,349 --> 00:03:52,491
locally, even when uploads are disabled.

70
00:03:53,152 --> 00:03:58,337
And interrupted turns now close cleanly,
so a delayed abort or a file approval

71
00:03:58,397 --> 00:04:00,619
cancellation won't interrupt the next
turn.

72
00:04:01,700 --> 00:04:03,903
That one should cut down on weird, phantom
stops.

73
00:04:04,592 --> 00:04:11,239
So the practical step: update to 0.169.0,
find the Cloud connection

74
00:04:11,299 --> 00:04:14,462
setting, and try it on one low risk thread
first.

75
00:04:15,143 --> 00:04:19,547
Send a screenshot, set the reasoning
effort, then archive and restore the thread.

76
00:04:20,348 --> 00:04:22,670
That tells you in five minutes whether it
fits how you work.

77
00:04:23,371 --> 00:04:24,452
That's the changelog for today.

