1
00:00:08,240 --> 00:00:12,320
Welcome back to Grok Bot Release Notes,
brought to you by Jellypod AI.

2
00:00:12,940 --> 00:00:19,080
Last time we covered version 0.70.2, so I
won't rehash the bot inboxes and unlinked

3
00:00:19,300 --> 00:00:19,700
X search.

4
00:00:20,360 --> 00:00:23,960
Today is about what landed between October
12 and October 16.

5
00:00:24,740 --> 00:00:26,840
The biggest change isn't a feature in the
app.

6
00:00:27,180 --> 00:00:28,280
It's how you pay for it.

7
00:00:28,988 --> 00:00:34,028
On October 12, SpaceXSI and Cursor
announced The X Subscription.

8
00:00:34,548 --> 00:00:39,028
It folds Grok, Grok Bot, Cursor, and X
Premium into one plan.

9
00:00:39,548 --> 00:00:44,668
The tiers are Free at zero dollars a
month, Plus at eight, Premium at thirty,

10
00:00:45,087 --> 00:00:47,728
Super at one hundred, and Ultra at two
hundred.

11
00:00:48,308 --> 00:00:50,648
Premium carries four times the usage of
Plus.

12
00:00:50,948 --> 00:00:54,108
Super carries fifteen times, and Ultra
fifty times.

13
00:00:54,728 --> 00:00:58,688
Teams get Premium, Super, and Ultra at the
same per seat prices,

14
00:00:58,928 --> 00:01:00,448
thirty, one hundred, and two hundred.

15
00:01:00,967 --> 00:01:05,188
That's described as cheaper than legacy
Cursor Teams, with three times more coding

16
00:01:05,268 --> 00:01:05,588
usage.

17
00:01:06,158 --> 00:01:08,838
And that Premium price replaces something
specific.

18
00:01:09,498 --> 00:01:14,658
SuperGrok, Cursor Pro, and X Premium Plus
used to cost ninety dollars a month

19
00:01:14,698 --> 00:01:15,158
combined.

20
00:01:15,658 --> 00:01:16,378
Now it's thirty.

21
00:01:17,010 --> 00:01:19,150
Here's the mechanism that matters for Grok
Bot.

22
00:01:19,590 --> 00:01:23,830
Paid plans now share one Grok model usage
pool across Grok Bot,

23
00:01:24,170 --> 00:01:26,110
Cursor, and the Grok tab on X.

24
00:01:26,630 --> 00:01:29,010
And the allowance refreshes weekly, not
monthly.

25
00:01:29,650 --> 00:01:33,290
In the official announcement, the line is
that one heavy day doesn't use up your

26
00:01:33,330 --> 00:01:33,630
month.

27
00:01:34,430 --> 00:01:36,470
Think about a routine that goes wrong
overnight.

28
00:01:37,070 --> 00:01:39,870
Under a monthly budget, it could drain the
whole team's month.

29
00:01:40,470 --> 00:01:42,830
Under a weekly pool, the damage resets in
days.

30
00:01:43,487 --> 00:01:47,067
Now the trade off, and this is where the
community is pushing back.

31
00:01:47,727 --> 00:01:52,487
Third party models are no longer included
in the flat pool on the new X plans.

32
00:01:53,127 --> 00:01:55,987
You pay for them at API list prices
through top ups.

33
00:01:56,827 --> 00:02:01,587
My interpretation, and it's only my read,
is that this nudges coding and bot

34
00:02:01,647 --> 00:02:06,087
workflows toward Grok 4.7, since that's
what the pool actually covers.

35
00:02:07,447 --> 00:02:11,327
On the Cursor forum, the announcement
thread has well over a hundred replies.

36
00:02:11,907 --> 00:02:17,507
Users named Archit, coco_max, and tangjun
are among those reporting early confusion.

37
00:02:18,107 --> 00:02:22,567
The common complaint is that heavy Grok
Bot workflows burn through the shared weekly

38
00:02:22,627 --> 00:02:24,147
quota faster than expected.

39
00:02:24,847 --> 00:02:29,327
One feature request on the forum asks for
per agent usage, meaning who worked,

40
00:02:29,767 --> 00:02:32,767
tokens, cache, and cost against the weekly
pool.

41
00:02:33,327 --> 00:02:34,867
It had six votes when we checked.

42
00:02:35,627 --> 00:02:39,767
Another thread, from an Ultra user, says
their weekly allowance hit ninety nine

43
00:02:39,807 --> 00:02:41,247
percent in three days.

44
00:02:41,925 --> 00:02:43,365
So here's how I'd use this.

45
00:02:43,925 --> 00:02:47,145
If you run several bots, treat the weekly
pool like a budget.

46
00:02:47,685 --> 00:02:50,425
Put your most expensive routines on a
schedule you can see.

47
00:02:51,065 --> 00:02:52,965
Check usage midweek, not at the end.

48
00:02:53,605 --> 00:02:57,844
And remember Grok Bot's own Usage and
Billing screen now shows your plan name and

49
00:02:57,905 --> 00:03:02,205
labels its percentage as used, which came
in with 0.70.2.

50
00:03:03,135 --> 00:03:07,795
Second change: on October 14, xAI
introduced Grok Voice 3.0,

51
00:03:08,195 --> 00:03:10,295
and it also powers Grok Bot's voice mode.

52
00:03:10,615 --> 00:03:13,295
So when you call your bot, you're now
talking to the new model.

53
00:03:13,995 --> 00:03:16,995
xAI bills it as the most intelligent
speech to speech model,

54
00:03:17,155 --> 00:03:18,555
with latency under a second.

55
00:03:19,114 --> 00:03:23,475
On Artificial Analysis's Speech to Speech
Index it scores 86.9,

56
00:03:23,815 --> 00:03:26,015
against 81.3 for Voice 2.0.

57
00:03:26,695 --> 00:03:30,075
On Full Duplex Bench, which tests
interruptions and turn taking,

58
00:03:30,455 --> 00:03:32,415
it scored 97.0 percent.

59
00:03:33,175 --> 00:03:34,155
What does that mean for you?

60
00:03:34,815 --> 00:03:37,615
Interrupting your bot mid sentence should
feel less awkward.

61
00:03:38,255 --> 00:03:43,115
For developers, voice agent calls upgrade
automatically under the model name grok

62
00:03:43,195 --> 00:03:46,155
voice latest, at the existing rate of
eight cents a minute,

63
00:03:46,375 --> 00:03:47,935
which is four dollars eighty an hour.

64
00:03:48,712 --> 00:03:51,212
SpaceXSI also gave real world numbers.

65
00:03:51,652 --> 00:03:55,832
Voice 3.0 handles millions of daily driver
requests in Tesla vehicles,

66
00:03:56,272 --> 00:04:00,772
things like calendar checks, emails,
coffee orders, and reservations through Grok

67
00:04:00,832 --> 00:04:01,072
Bot.

68
00:04:01,652 --> 00:04:05,412
And in Starlink call centers, it answers
twenty thousand calls a day,

69
00:04:05,972 --> 00:04:08,292
with about seventy percent resolved
without a human.

70
00:04:08,832 --> 00:04:12,252
Those are SpaceXSI's own figures, so read
them as company claims.

71
00:04:13,030 --> 00:04:14,010
Now the desktop client.

72
00:04:14,450 --> 00:04:20,150
Version 0.71.2, built October 14, has been
rolling out to macOS and Windows,

73
00:04:20,510 --> 00:04:23,110
and it arrived without an official
changelog entry.

74
00:04:23,830 --> 00:04:27,890
As of our check, the official log still
tops out at 0.70.2.

75
00:04:28,370 --> 00:04:32,690
So everything I'm about to say about
0.71.2 comes from forum reports,

76
00:04:32,770 --> 00:04:33,510
not release notes.

77
00:04:34,323 --> 00:04:36,243
The biggest gotcha is on Windows 11.

78
00:04:36,723 --> 00:04:40,343
A user named ratiphi reports that local
tool access is blocked,

79
00:04:40,743 --> 00:04:44,483
things like the CopyFromBox action and
local shell execution.

80
00:04:44,963 --> 00:04:49,583
The error reads, Local tools are turned
off, the user has set local tool access to

81
00:04:49,643 --> 00:04:49,963
Never.

82
00:04:50,483 --> 00:04:54,743
But in Settings, under Execution on Local
Computer, it's set to Always Allow.

83
00:04:55,263 --> 00:04:58,683
A separate thread describes the same stuck
on Never behavior on a Mac,

84
00:04:59,043 --> 00:05:00,003
for one account only.

85
00:05:00,683 --> 00:05:03,703
The workaround for now is to fall back to
in chat attachments,

86
00:05:03,783 --> 00:05:06,443
or run the work on the cloud VM until a
patch lands.

87
00:05:07,112 --> 00:05:08,052
Two smaller reports.

88
00:05:08,632 --> 00:05:14,792
Newly created bots in 0.71.2 default to
Unassigned with no category dropdown,

89
00:05:15,312 --> 00:05:19,772
and the avatar editor locks to a default
gray shape with no custom shape or accent

90
00:05:19,812 --> 00:05:20,452
color pickers.

91
00:05:21,092 --> 00:05:24,092
Then there's the more serious category,
the cloud computer.

92
00:05:24,752 --> 00:05:27,652
Some users describe their Agent Computer
becoming unreachable,

93
00:05:27,812 --> 00:05:32,652
with a missing gateway descriptor file,
leaving bots in an unresponsive loop where

94
00:05:32,732 --> 00:05:34,072
manual resets fail.

95
00:05:34,792 --> 00:05:37,132
And several users report silent rollbacks.

96
00:05:37,732 --> 00:05:40,952
One user said their cloud computer
reverted twice on October 8,

97
00:05:41,192 --> 00:05:43,032
losing about two hours of video rendering.

98
00:05:43,732 --> 00:05:48,252
Another thread claims a full disk triggers
a silent rollback and loses two days of

99
00:05:48,292 --> 00:05:48,632
work.

100
00:05:49,392 --> 00:05:53,752
A support engineer hasn't confirmed that
cause, so treat it as a user hypothesis.

101
00:05:54,552 --> 00:05:56,592
The practical lesson doesn't need
confirmation though.

102
00:05:57,152 --> 00:06:01,072
Commit code to an external git repository,
or sync to a cloud drive,

103
00:06:01,352 --> 00:06:02,132
continuously.

104
00:06:02,692 --> 00:06:04,492
Don't let the VM be the only copy.

105
00:06:05,257 --> 00:06:10,017
On the positive side, creator Matt
Palmer's October 13 walkthrough highlights a

106
00:06:10,077 --> 00:06:11,057
Shopify connector.

107
00:06:11,577 --> 00:06:15,557
His description is that your bot can track
orders, check your store's status,

108
00:06:15,677 --> 00:06:16,757
and read your metrics.

109
00:06:17,237 --> 00:06:21,377
That's from his video, not the official
log, so check your own connector list before

110
00:06:21,437 --> 00:06:22,037
counting on it.

111
00:06:22,633 --> 00:06:28,013
Last, two safeguards from the 0.70.2 notes
that matter more now that bots can pay

112
00:06:28,073 --> 00:06:29,833
with X Money or Stripe Link.

113
00:06:30,413 --> 00:06:34,173
Before an action it can't undo, like
sending, posting, buying,

114
00:06:34,213 --> 00:06:37,353
or deleting, a bot checks your own
messages for approval.

115
00:06:37,893 --> 00:06:42,673
And after a bank or brokerage connector
returns data, the bot asks before sharing it

116
00:06:42,733 --> 00:06:43,373
anywhere else.

117
00:06:44,073 --> 00:06:48,453
Also, the notes say Grok Bot can't reach a
private Tailscale MCP server yet,

118
00:06:48,873 --> 00:06:50,953
and suggest Tailscale Funnel to expose it.

119
00:06:51,613 --> 00:06:52,393
So, the takeaways.

120
00:06:52,693 --> 00:06:53,833
Watch your weekly pool.

121
00:06:54,293 --> 00:06:55,313
Try the new voice mode.

122
00:06:55,713 --> 00:07:00,453
If you're on Windows 11 and 0.71.2, expect
local tool trouble.

123
00:07:00,853 --> 00:07:03,073
And back up your work outside the VM.

124
00:07:03,733 --> 00:07:06,993
That's it for today, and this episode was
brought to you by Jellypod AI.

