1
00:00:00,000 --> 00:00:00,640
Picture this.

2
00:00:00,800 --> 00:00:05,360
You set off an autonomous Claude Code
subagent on a massive background job,

3
00:00:05,420 --> 00:00:09,840
say, a ninety minute end to end test suite
or a massive Docker build.

4
00:00:10,000 --> 00:00:13,120
You grab a coffee, come back an hour
later...

5
00:00:13,180 --> 00:00:13,920
and boom.

6
00:00:14,053 --> 00:00:17,360
It was abruptly killed right at the sixty
minute mark.

7
00:00:17,520 --> 00:00:23,920
Well, version 2.1.260 eliminates that
invisible ceiling, brought to you by Jellypod

8
00:00:24,107 --> 00:00:24,560
AI.

9
00:00:24,761 --> 00:00:27,861
Man, that sixty minute cutoff was such a
silent killer!

10
00:00:28,481 --> 00:00:32,161
You would have long running database
migrations or web scrapers running in a

11
00:00:32,181 --> 00:00:36,181
subagent, and suddenly everything just
dropped dead with zero warning.

12
00:00:36,208 --> 00:00:37,728
Oh, tell me about it mate!

13
00:00:37,768 --> 00:00:44,288
I, I, I actually had a multi repo test
pipeline fail on me mid run last week because

14
00:00:44,315 --> 00:00:45,568
of that exact limit.

15
00:00:45,848 --> 00:00:47,568
I thought I was going bonkers!

16
00:00:47,728 --> 00:00:54,368
But yeah, according to the official
release notes for Claude Code version 2.1.260,

17
00:00:54,448 --> 00:00:58,848
Anthropic removed the one hour time limit
on background commands started by

18
00:00:58,928 --> 00:00:59,648
subagents.

19
00:00:59,786 --> 00:01:04,726
So now subagent background tasks just run
indefinitely until they finish or until

20
00:01:04,806 --> 00:01:08,607
you manually stop them, which finally
matches how the main session behaves.

21
00:01:08,625 --> 00:01:09,185
Spot on.

22
00:01:09,205 --> 00:01:11,665
And the practical workflow now is super
clean, mate.

23
00:01:11,685 --> 00:01:15,185
You just dispatch your long background job
through a subagent,

24
00:01:15,205 --> 00:01:19,265
and then you can monitor the whole process
directly using commands like attach,

25
00:01:19,345 --> 00:01:20,385
logs, and stop.

26
00:01:20,505 --> 00:01:25,105
You do not have to worry about some
arbitrary timer pulling the plug while you step

27
00:01:25,153 --> 00:01:26,145
away from the terminal.

28
00:01:26,444 --> 00:01:27,345
Okay, but, uh...

29
00:01:28,164 --> 00:01:28,384
wait.

30
00:01:29,004 --> 00:01:33,484
If subagents do not auto terminate
background processes at sixty minutes anymore,

31
00:01:34,085 --> 00:01:36,364
what happens if a task gets stuck in an
infinite loop?

32
00:01:36,557 --> 00:01:39,157
Ah, now you are asking the real
questions .

33
00:01:39,742 --> 00:01:40,442
Seriously though!

34
00:01:40,703 --> 00:01:45,083
On build host machines or cloud instances,
a hung or spinning subagent task will now

35
00:01:45,122 --> 00:01:47,622
literally run forever if you do not watch
it.

36
00:01:47,682 --> 00:01:51,043
That is a quick way to burn through your
cloud budget or spike host CPU to one

37
00:01:51,083 --> 00:01:51,682
hundred percent.

38
00:01:51,833 --> 00:01:52,793
Yeah, fair point.

39
00:01:52,953 --> 00:01:56,793
You definitely cannot just fire and forget
without guardrails anymore.

40
00:01:56,792 --> 00:01:57,512
Exactly.

41
00:01:57,572 --> 00:02:02,952
You have to pair this uncapped execution
with proper process monitoring,

42
00:02:03,139 --> 00:02:09,592
or wrap your commands in explicit timeout
scripts, or even use Linux cgroups memory

43
00:02:09,649 --> 00:02:14,232
limits on your build runners so runaway
processes get caught before they wreck your

44
00:02:14,289 --> 00:02:14,792
server.

45
00:02:15,111 --> 00:02:16,652
That is solid advice, mate.

46
00:02:17,351 --> 00:02:19,551
You gotta respect the power of an infinite
runtime.

47
00:02:19,583 --> 00:02:20,143
Definitely.

48
00:02:20,255 --> 00:02:24,463
Were there any other noteworthy changes
packed into 2.1.260?

49
00:02:24,458 --> 00:02:27,498
Yeah, a couple of really neat quality of
life tweaks.

50
00:02:27,538 --> 00:02:31,338
For one, they added a text form of the
advisor command, so you can run slash

51
00:02:31,408 --> 00:02:35,738
advisor, slash advisor followed by the
model name, or slash advisor off.

52
00:02:35,882 --> 00:02:41,418
That is massive for headless setups,
Remote Control, and Agent SDK workflows where

53
00:02:41,438 --> 00:02:43,418
you do not have an interactive UI picker.

54
00:02:43,536 --> 00:02:44,396
Oh nice!

55
00:02:45,016 --> 00:02:46,877
What about safety around git worktrees?

56
00:02:47,076 --> 00:02:47,996
Did anything change there?

57
00:02:48,201 --> 00:02:49,061
Glad you asked!

58
00:02:49,782 --> 00:02:54,201
Before, if you went to delete a background
session worktree with unpushed work,

59
00:02:54,621 --> 00:02:57,641
it was pretty easy to accidentally wipe
stuff out.

60
00:02:57,741 --> 00:03:02,241
Now, Claude Code gives you a clear safety
prompt that spells out the exact branch

61
00:03:02,321 --> 00:03:06,141
name and the specific commit count before
deleting anything.

62
00:03:06,201 --> 00:03:09,562
So if you see three unpushed commits on
feature branch main,

63
00:03:09,621 --> 00:03:12,001
you get a chance to save your skin before
double deleting!

64
00:03:12,801 --> 00:03:15,441
That alone is gonna save a lot of
developers from a heart attack.

65
00:03:16,061 --> 00:03:22,342
Unlimited subagent runtimes plus safer
worktree cleanup makes 2.1.260 a really solid

66
00:03:22,401 --> 00:03:22,721
release.

67
00:03:22,750 --> 00:03:23,870
No doubt about it, mate.

68
00:03:24,030 --> 00:03:26,750
Go update your CLI and let those builds
run!