1
00:00:00,079 --> 00:00:03,439
So if you are running multiple terminal
tasks in Codex right now,

2
00:00:03,980 --> 00:00:08,619
things just got a whole lot cleaner in
version zero point one fifty point zero.

3
00:00:09,191 --> 00:00:13,591
Wait, you mean we can finally stop doing
the whole copy paste dance between

4
00:00:13,591 --> 00:00:14,752
different terminal windows?

5
00:00:15,964 --> 00:00:16,724
Exactly that.

6
00:00:17,305 --> 00:00:21,724
Thanks to Jellypod sponsoring today's
daily update, we get to dive straight into

7
00:00:21,764 --> 00:00:22,044
this.

8
00:00:22,625 --> 00:00:28,565
OpenAI just dropped cross task mentions
with @ syntax in Codex zero point one fifty

9
00:00:28,704 --> 00:00:29,144
point zero.

10
00:00:29,591 --> 00:00:30,711
Oh, interesting.

11
00:00:31,292 --> 00:00:35,651
So like, how you @ mention someone in
Slack, but for running agent tasks?

12
00:00:35,958 --> 00:00:36,518
Right.

13
00:00:36,598 --> 00:00:41,078
In the prompt input, you can literally
type at task one zero two,

14
00:00:41,291 --> 00:00:43,638
or whatever task ID was generated.

15
00:00:43,798 --> 00:00:50,198
And Codex can read from that task, send a
message to it, or even spawn sub tasks off

16
00:00:50,225 --> 00:00:52,038
it directly from your terminal session.

17
00:00:52,341 --> 00:00:53,101
Oh wow.

18
00:00:53,282 --> 00:00:56,922
Okay, walk me through what that actually
looks like when you are sitting there

19
00:00:56,962 --> 00:00:57,581
writing code.

20
00:00:57,962 --> 00:00:59,441
Give me a concrete scenario.

21
00:00:59,891 --> 00:01:04,551
Okay, say you are refactoring a primary
module in one main terminal task,

22
00:01:04,631 --> 00:01:04,831
right?

23
00:01:05,431 --> 00:01:10,252
But in the background, you have a
secondary task, let us call it at task eighty

24
00:01:10,311 --> 00:01:12,632
eight, running your integration test
suite.

25
00:01:13,046 --> 00:01:16,046
Right, right, and tests start failing.

26
00:01:16,576 --> 00:01:17,316
Exactly.

27
00:01:17,836 --> 00:01:22,076
Instead of stopping what you are doing,
scrolling up through a thousand lines of

28
00:01:22,176 --> 00:01:27,156
terminal output, copying the error stack,
and pasting it back into your primary

29
00:01:27,217 --> 00:01:27,556
prompt...

30
00:01:28,176 --> 00:01:33,396
you just tell Codex, hey, check at task
eighty eight and figure out why the user

31
00:01:33,456 --> 00:01:34,616
auth test failed.

32
00:01:35,466 --> 00:01:36,986
That is so nice!

33
00:01:37,346 --> 00:01:42,186
Because manually dumping terminal logs
into context always feels so messy.

34
00:01:42,226 --> 00:01:44,927
And honestly, it burns through tokens like
crazy.

35
00:01:45,471 --> 00:01:47,752
Well, and that is the clever part here.

36
00:01:47,831 --> 00:01:53,171
Codex does not just dump the entire
transcript of at task eighty eight into your

37
00:01:53,191 --> 00:01:53,831
current window.

38
00:01:54,431 --> 00:01:59,712
It pulls structured task summaries,
keeping context tight while giving the agent

39
00:01:59,771 --> 00:02:01,252
enough info to fix the issue.

40
00:02:01,711 --> 00:02:07,931
So it turns individual terminal prompt
loops into actual multi task orchestration.

41
00:02:08,551 --> 00:02:12,252
That feels like a pretty big shift in how
developers will coordinate background

42
00:02:12,331 --> 00:02:12,712
agents.

43
00:02:13,099 --> 00:02:13,839
It really is.

44
00:02:14,439 --> 00:02:19,099
But speaking of running agents across
different tasks, version zero point one fifty

45
00:02:19,239 --> 00:02:24,439
point zero also brings some big security
hardening, especially around untrusted

46
00:02:24,500 --> 00:02:25,279
repositories.

47
00:02:25,592 --> 00:02:28,492
Ah, untrusted repos.

48
00:02:29,212 --> 00:02:33,572
Let me guess, prompt injection risks
through project config files?

49
00:02:34,032 --> 00:02:34,812
Spot on.

50
00:02:35,232 --> 00:02:39,552
In previous builds, if a project root had
an AGENTS dot md file,

51
00:02:40,052 --> 00:02:43,472
Codex would just execute those
instructions automatically.

52
00:02:43,572 --> 00:02:49,333
Now, if the project is untrusted, Codex
ignores those project level AGENTS dot md

53
00:02:49,392 --> 00:02:50,433
instructions completely.

54
00:02:50,877 --> 00:02:55,317
That is a huge relief for anyone reviewing
open source pull requests.

55
00:02:55,797 --> 00:03:01,157
You do not want a cloned repo secretly
overriding your agent's behavior behind your

56
00:03:01,177 --> 00:03:01,577
back.

57
00:03:02,177 --> 00:03:02,797
Exactly.

58
00:03:03,257 --> 00:03:09,097
Plus, managed deny read path rules now
stay strictly enforced even if you cycle your

59
00:03:09,137 --> 00:03:11,277
TUI permission modes mid session.

60
00:03:11,918 --> 00:03:14,637
Security controls stay locked down no
matter what.

61
00:03:15,007 --> 00:03:20,087
Okay, what about when an agent goes off on
a tangent or gets stuck in a loop?

62
00:03:20,607 --> 00:03:22,506
Anything new for handling interruptions?

63
00:03:22,792 --> 00:03:24,312
Actually, yes!

64
00:03:24,408 --> 00:03:26,952
They introduced Interrupt lifecycle hooks.

65
00:03:27,165 --> 00:03:32,712
So when you hit Control C to interrupt a
top level turn, Codex can automatically run

66
00:03:32,781 --> 00:03:38,072
custom shell commands or Model Context
Protocol handlers that you set up.

67
00:03:38,402 --> 00:03:39,402
Wait, really?

68
00:03:39,962 --> 00:03:44,563
So if my agent spun up a local dev server
or an active cloud execution,

69
00:03:44,942 --> 00:03:48,923
Control C can clean up those background
processes automatically?

70
00:03:50,274 --> 00:03:55,394
Yep, no more orphaned node processes
lingering in the background eating up your RAM!

71
00:03:57,397 --> 00:04:00,918
Okay, as a former tester, that speaks
directly to my soul.

72
00:04:00,977 --> 00:04:02,177
That is fantastic.

73
00:04:02,640 --> 00:04:07,279
There are a few really nice developer
quality of life polish items in this release

74
00:04:07,339 --> 00:04:07,519
too.

75
00:04:08,299 --> 00:04:13,400
Like slash copy now opens an interactive
picker where you can choose whether to copy

76
00:04:13,460 --> 00:04:17,839
the full response, just isolated code
blocks, or blockquotes.

77
00:04:18,257 --> 00:04:19,198
Oh, sweet!

78
00:04:19,698 --> 00:04:22,497
And didn't I see something about automatic
task naming?

79
00:04:22,833 --> 00:04:23,593
Yes!

80
00:04:23,713 --> 00:04:28,593
Unnamed terminal tasks now get descriptive
titles automatically based on what you

81
00:04:28,633 --> 00:04:33,873
are doing, and slash rename will actually
suggest an editable title for you.

82
00:04:34,134 --> 00:04:38,154
That makes tracking multiple at mentions
so much easier.

83
00:04:38,235 --> 00:04:42,595
Oh, and markdown links now show up as
clickable labels in supported terminals!

84
00:04:43,157 --> 00:04:43,497
Right.

85
00:04:43,737 --> 00:04:47,318
And we should definitely highlight the
follow up patch that dropped right after,

86
00:04:47,878 --> 00:04:50,497
version zero point one fifty point one.

87
00:04:50,924 --> 00:04:54,764
Oh, what did zero point one fifty point
one fix?

88
00:04:55,292 --> 00:04:58,412
It fixed remote compaction for multimodal
sessions.

89
00:04:58,592 --> 00:05:02,972
Retained images now get counted toward the
token budget by default,

90
00:05:03,025 --> 00:05:08,172
trimming older images as needed so your
context window does not quietly balloon

91
00:05:08,252 --> 00:05:11,372
during long image heavy debugging
sessions.

92
00:05:11,754 --> 00:05:13,394
That is super solid.

93
00:05:13,875 --> 00:05:18,434
Between cross task at mentions, safer
untrusted repo handling,

94
00:05:18,714 --> 00:05:24,074
and interrupt hooks, zero point one fifty
point zero is a monster release.

95
00:05:25,034 --> 00:05:25,894
It really is.

96
00:05:26,334 --> 00:05:28,714
Alright, that is the latest update for
today.

97
00:05:29,255 --> 00:05:30,674
Catch you all on the next one!