1
00:00:00,000 --> 00:00:03,520
Alright, if you've been running Claude
Code in tmux lately and- and felt like your

2
00:00:03,564 --> 00:00:07,840
terminal screen was throwing a absolute
rave, you're not alone.

3
00:00:07,904 --> 00:00:11,840
This is the Claude Code Changelog, brought
to you by Jellypod AI.

4
00:00:12,000 --> 00:00:13,200
I'm Lachlan.

5
00:00:13,492 --> 00:00:14,292
And I'm James.

6
00:00:14,332 --> 00:00:17,652
And- and seriously, that- that tmux
flickering was driving me up the wall,

7
00:00:17,692 --> 00:00:18,532
Lachlan.

8
00:00:18,572 --> 00:00:23,492
But let's- let's actually dive into the
big safety shift in v2.1.200 first.

9
00:00:23,524 --> 00:00:26,612
They've flipped the default permission
mode to 'Manual'.

10
00:00:26,928 --> 00:00:27,648
Right, right.

11
00:00:27,728 --> 00:00:29,328
So no more Claude just...

12
00:00:29,512 --> 00:00:34,048
wandering off and running commands while
you're, I don't know,

13
00:00:34,084 --> 00:00:35,008
grabbing a cuppa.

14
00:00:35,141 --> 00:00:38,288
It's- it's like putting a lock on the back
gate, you know?

15
00:00:38,528 --> 00:00:43,568
If it hits one of those AskUserQuestion
dialogs, it just- it sits there.

16
00:00:43,728 --> 00:00:45,328
It pauses indefinitely.

17
00:00:45,625 --> 00:00:49,065
Yeah, it- it won't auto-continue at all
now unless you go out of your way to

18
00:00:49,113 --> 00:00:52,425
configure an explicit idle timeout via
/config.

19
00:00:52,585 --> 00:00:56,025
It's a massive deal because if you leave
your terminal open,

20
00:00:56,045 --> 00:00:59,866
you don't want an autonomous agent just
deciding, 'Hey, let's push this' without a

21
00:00:59,918 --> 00:01:00,665
human eye on it.

22
00:01:00,983 --> 00:01:01,863
Oh, absolutely.

23
00:01:01,903 --> 00:01:06,183
I mean, back in my early days, I nearly
tanked a- a client's site with a dodgy

24
00:01:06,245 --> 00:01:07,303
midnight update.

25
00:01:07,360 --> 00:01:09,783
Having an AI doing that on autopilot?

26
00:01:10,023 --> 00:01:12,743
No thanks, that gives me the absolute
collywobbles.

27
00:01:14,333 --> 00:01:14,973
Exactly.

28
00:01:15,073 --> 00:01:21,213
But then, right on its heels, we got
v2.1.201, and this is the- the real

29
00:01:21,266 --> 00:01:23,053
under-the-hood performance beauty.

30
00:01:23,186 --> 00:01:26,813
It's all about protecting prompt caching
for Sonnet 5.

31
00:01:27,108 --> 00:01:31,188
Right, because if you're working in a
massive codebase, prompt caching is what keeps

32
00:01:31,220 --> 00:01:34,228
your API bill from- from absolutely
skyrocketing, yeah?

33
00:01:34,441 --> 00:01:38,788
If you bust that cache, you're paying for
thousands of tokens over and over.

34
00:01:39,080 --> 00:01:43,080
Right, and what was happening under the
hood was that Claude Code was injecting

35
00:01:43,107 --> 00:01:48,360
these mid-conversation system role
messages-basically 'harness reminders' to keep

36
00:01:48,380 --> 00:01:49,400
the agent aligned.

37
00:01:49,500 --> 00:01:53,720
But because those system messages were
being popped in dynamically mid-chat,

38
00:01:53,752 --> 00:01:56,840
they completely invalidated the prompt
cache prefix!

39
00:01:57,157 --> 00:01:59,078
Oh, mate, that's- that's like...

40
00:01:59,128 --> 00:02:02,518
imagine you're reading a massive manual,
and every five minutes,

41
00:02:02,538 --> 00:02:05,918
someone taps your shoulder and rewrites
the- the preface of the book.

42
00:02:06,738 --> 00:02:09,878
You'd have to start reading the whole
bloody thing from page one again!

43
00:02:10,163 --> 00:02:10,963
Exactly!

44
00:02:11,123 --> 00:02:14,803
Sonnet has to re-evaluate the entire
context from scratch.

45
00:02:14,990 --> 00:02:20,883
By removing those mid-conversation
reminders in 2.1.201, the cache prefix stays

46
00:02:20,934 --> 00:02:22,163
completely untouched.

47
00:02:22,275 --> 00:02:26,883
Your latency drops, and your API usage
stays incredibly efficient.

48
00:02:27,193 --> 00:02:28,073
That is brilliant.

49
00:02:28,180 --> 00:02:28,793
Truly.

50
00:02:28,933 --> 00:02:32,313
And- and what about those- those quality
of life tweaks?

51
00:02:32,393 --> 00:02:34,153
The tmux fix?

52
00:02:34,460 --> 00:02:36,940
Right, so they added synchronized terminal
output.

53
00:02:37,040 --> 00:02:41,660
Instead of rendering the screen in messy,
staggered chunks-which is what caused that

54
00:02:41,727 --> 00:02:46,620
rapid flickering in tmux-it now
coordinates the render so it draws the terminal in

55
00:02:46,700 --> 00:02:48,540
one single, clean sweep.

56
00:02:48,833 --> 00:02:49,633
Ah, nice.

57
00:02:49,740 --> 00:02:51,233
No more rave in the terminal.

58
00:02:51,333 --> 00:02:53,873
And- and there's a new installer check
too, right?

59
00:02:53,913 --> 00:02:54,753
For- for memory?

60
00:02:55,050 --> 00:02:59,050
Yeah, it's a smart diagnostic warning for
out-of-memory, or OOM,

61
00:02:59,108 --> 00:02:59,930
conditions.

62
00:03:00,037 --> 00:03:04,090
If you're trying to install this on a
resource-constrained environment-like a tiny

63
00:03:04,210 --> 00:03:08,970
VPS or a docker container-the installer
will flag it upfront instead of letting Node

64
00:03:09,023 --> 00:03:11,610
crash silently with a cryptic stack trace.

65
00:03:11,767 --> 00:03:12,687
Incredible.

66
00:03:12,807 --> 00:03:17,687
Safer defaults, cheaper runs, and a screen
that doesn't melt your eyes.

67
00:03:17,815 --> 00:03:19,247
Good stuff.

68
00:03:19,335 --> 00:03:20,247
Talk soon, mate.

69
00:03:20,542 --> 00:03:21,102
Yep, see ya.