1
00:00:00,040 --> 00:00:01,360
Welcome to the Deep Dive 
Learner. 

2
00:00:02,120 --> 00:00:06,400
Today we are taking on a mission
to really unpack an incredibly 

3
00:00:06,400 --> 00:00:09,320
dense week of AI news. 
I mean, we're looking at 

4
00:00:09,320 --> 00:00:15,560
everything from April 6th to 
April 10th, 2026 and there is a 

5
00:00:15,560 --> 00:00:17,720
lot of ground to cover. 
Yeah, there really is. 

6
00:00:17,720 --> 00:00:20,520
It's it's been a massive week. 
Right, we've got breaking 

7
00:00:20,520 --> 00:00:23,680
reports from the new stack. 
We have some really deep 

8
00:00:23,680 --> 00:00:27,920
analysis over from late in space
and honestly some wild frontline

9
00:00:27,920 --> 00:00:29,560
dispatches from developer 
subreddits. 

10
00:00:29,560 --> 00:00:32,000
Like, you know, our clawed code 
and our clawed AI. 

11
00:00:32,080 --> 00:00:34,280
The subreddits have been a 
goldmine lately honestly. 

12
00:00:34,400 --> 00:00:36,760
Absolutely, because the theme of
what we're looking at today, 

13
00:00:36,760 --> 00:00:39,840
Lerner, is that whether you are 
a developer, you know, right 

14
00:00:39,840 --> 00:00:42,160
there in the trenches or a 
business leader trying to figure

15
00:00:42,160 --> 00:00:47,320
this out, we are so far past the
the sort of wow phase of AI 

16
00:00:47,320 --> 00:00:48,960
coding tools. 
Yeah, the honeymoon period is 

17
00:00:48,960 --> 00:00:51,000
completely over. 
Totally over. 

18
00:00:51,000 --> 00:00:56,040
We're in this really messy high 
stakes era now where AI is, I 

19
00:00:56,040 --> 00:00:58,800
mean, it's automating entire 
corporate workflows, it's 

20
00:00:58,800 --> 00:01:02,040
sparking these developer turf 
wars, and it's fundamentally 

21
00:01:02,040 --> 00:01:03,920
changing how businesses actually
operate. 

22
00:01:04,200 --> 00:01:06,760
So OK, let's unpack this, 
because to understand the big 

23
00:01:06,760 --> 00:01:09,360
picture, we actually have to 
start super small. 

24
00:01:09,480 --> 00:01:10,840
Right with the invisible 
failure. 

25
00:01:10,840 --> 00:01:14,000
Exactly like a micro level 
problem discovered by just one 

26
00:01:14,000 --> 00:01:16,840
single user on Reddit. 
Because it perfectly shows the 

27
00:01:16,880 --> 00:01:19,960
absolute danger of trusting 
these tools blindly. 

28
00:01:20,040 --> 00:01:22,680
Yeah, this was wild. 
So on April 10th, a Reddit user 

29
00:01:22,680 --> 00:01:26,320
in the R Claude code community, 
they basically found three 

30
00:01:26,320 --> 00:01:30,240
stacked bugs in Claude code. 
This is since version V2, point 

31
00:01:30,240 --> 00:01:32,600
0.64. 
Right, the new update. 

32
00:01:32,640 --> 00:01:35,320
Yeah, so developers do what 
developers do right? 

33
00:01:35,320 --> 00:01:38,120
They tweak their config files. 
But they found that if a user 

34
00:01:38,120 --> 00:01:41,080
sets their config to Always 
thinking enabled. 

35
00:01:41,440 --> 00:01:43,120
True, it was also an environment
variable. 

36
00:01:43,120 --> 00:01:45,080
Right, Yes, exactly. 
If you set the environment 

37
00:01:45,080 --> 00:01:48,160
variable, Claude could effort 
level to Max the extended 

38
00:01:48,160 --> 00:01:50,920
thinking feature. 
Just, well, it silently fails. 

39
00:01:51,040 --> 00:01:56,000
There is 0 error message, no 
warning in the terminal, it just

40
00:01:56,000 --> 00:01:58,960
quietly runs at the default 
shallow capability. 

41
00:01:59,240 --> 00:02:03,480
See, that is terrifying because,
you know, Lerner, think about 

42
00:02:03,480 --> 00:02:06,560
this like driving a car. 
You're on the highway, your 

43
00:02:06,560 --> 00:02:10,520
speedometer says you're doing 65
mph, You feel the engine, the 

44
00:02:10,520 --> 00:02:12,760
road is going by, it all seems 
totally fine. 

45
00:02:13,040 --> 00:02:15,600
But in reality, you're only 
doing 30 mph and you have 

46
00:02:15,680 --> 00:02:18,280
absolutely no idea. 
Yeah, there's no check engine 

47
00:02:18,280 --> 00:02:19,520
light whatsoever. 
Right. 

48
00:02:19,520 --> 00:02:22,640
So how did they even figure it 
out if there was no error? 

49
00:02:22,920 --> 00:02:25,480
Well, they had to prove it using
what they call a Canary 

50
00:02:25,480 --> 00:02:26,080
question. 
OK. 

51
00:02:26,160 --> 00:02:28,840
I love this part. 
It's basically a logic puzzle 

52
00:02:28,840 --> 00:02:31,200
with deliberate misdirection 
built into it. 

53
00:02:31,440 --> 00:02:35,000
If the model does a fast shallow
pass, it will always get it 

54
00:02:35,000 --> 00:02:36,880
wrong. 
It falls for the trap every 

55
00:02:36,880 --> 00:02:38,840
time. 
That was literally the only way 

56
00:02:38,840 --> 00:02:41,240
they could prove their config 
files were lying to them. 

57
00:02:41,480 --> 00:02:44,040
That's insane. 
And then fixing it. 

58
00:02:44,360 --> 00:02:46,040
They had to use a manual 
wrapper, didn't? 

59
00:02:46,040 --> 00:02:48,680
They yeah, a community member 
had to publish a custom wrapper 

60
00:02:48,680 --> 00:02:51,680
just to intercept the config 
before execution to force it to 

61
00:02:51,680 --> 00:02:54,080
actually work. 
I have to ask, I mean, is a 

62
00:02:54,080 --> 00:02:58,120
silent failure actually worse 
than a complete system crash? 

63
00:02:58,400 --> 00:03:01,000
Oh, absolutely, yes. 
I mean, what's fascinating here 

64
00:03:01,000 --> 00:03:03,680
is the core lesson. 
You cannot trust your config 

65
00:03:03,680 --> 00:03:06,040
files anymore. 
Right, you can't fix what you 

66
00:03:06,040 --> 00:03:07,800
don't know is broken. 
Exactly. 

67
00:03:08,040 --> 00:03:11,960
A loud system crash stops the 
assembly line so you can fix it.

68
00:03:11,960 --> 00:03:15,240
Yeah, a silent failure means you
keep shipping defective code. 

69
00:03:15,720 --> 00:03:18,600
We are having to completely 
shift from just looking at our 

70
00:03:18,600 --> 00:03:21,680
settings to actively verifying 
the behavior of the AI. 

71
00:03:21,680 --> 00:03:23,560
Which brings us to the scale of 
this problem. 

72
00:03:23,800 --> 00:03:26,640
Because the final and failure 
for one developer is, well, it's

73
00:03:26,640 --> 00:03:28,760
a massive headache. 
Yeah, but what happens when you 

74
00:03:28,760 --> 00:03:31,960
scale that blind trust across 
the entire software industry? 

75
00:03:32,080 --> 00:03:36,120
Right, because over on our cloud
AI, there was this post from an 

76
00:03:36,120 --> 00:03:39,120
11 year veteran developer who 
just automated, I mean roughly 

77
00:03:39,120 --> 00:03:42,720
80% of their job. 80%, that's 
just wild. 

78
00:03:42,720 --> 00:03:44,680
It's nuts. 
They use the cloud CLI in this 

79
00:03:44,680 --> 00:03:48,680
custom.net console app to just 
constantly pull the GitLab API 

80
00:03:48,680 --> 00:03:51,040
for issues. 
So the AI reads the issue, 

81
00:03:51,200 --> 00:03:54,200
writes the code and submits the 
PR the pull request. 

82
00:03:54,240 --> 00:03:56,520
And on the surface, that looks 
like the ultimate developer 

83
00:03:56,520 --> 00:03:59,040
dream, right? 
Right, but the news stack is 

84
00:03:59,040 --> 00:04:00,760
reporting a totally different 
reality. 

85
00:04:01,240 --> 00:04:04,160
They're calling it an AI 
generated PR crisis. 

86
00:04:04,640 --> 00:04:07,080
Open source maintainers are just
drowning right now. 

87
00:04:07,080 --> 00:04:11,040
They literally cannot keep up 
with the unsustainable volume of

88
00:04:11,040 --> 00:04:14,240
these AI pull requests. 
Because the failure mode is so 

89
00:04:14,240 --> 00:04:19,000
tricky, the code, well it sounds
right on 1st read it looks 

90
00:04:19,000 --> 00:04:21,519
plausible. 
Plausibility is the real danger 

91
00:04:21,519 --> 00:04:23,560
here. 
The syntax is technically 

92
00:04:23,560 --> 00:04:27,920
correct, but it's subtly bugged,
or it completely ignores project

93
00:04:27,920 --> 00:04:31,080
conventions, or, you know, 
misses the contextual nuance of 

94
00:04:31,080 --> 00:04:33,440
the code base. 
And the News Stack warned that 

95
00:04:33,440 --> 00:04:37,520
enterprise teams are next. 
So here's where it gets really 

96
00:04:37,520 --> 00:04:40,240
interesting. 
I want to push back on that 11 

97
00:04:40,240 --> 00:04:42,680
year veterans automation dream. 
OK, let's hear it. 

98
00:04:42,960 --> 00:04:45,760
If that developer automates 80% 
of their coding, they aren't 

99
00:04:45,760 --> 00:04:47,800
actually eliminating 80% of the 
work, right? 

100
00:04:48,080 --> 00:04:50,640
Aren't they just dumping 80% of 
their mental workload onto 

101
00:04:50,640 --> 00:04:53,280
whoever the poor reviewer is 
that has to check that code? 

102
00:04:53,400 --> 00:04:54,640
That is exactly what's 
happening. 

103
00:04:54,640 --> 00:04:56,840
The bottleneck of software 
engineering has officially 

104
00:04:56,840 --> 00:04:58,560
shifted. 
Wow, shifted how? 

105
00:04:58,720 --> 00:05:01,400
Well, for decades the bottleneck
was generation. 

106
00:05:01,720 --> 00:05:04,440
Writing the code. 
Now generation is practically 

107
00:05:04,440 --> 00:05:06,520
free. 
The new bottleneck is 

108
00:05:06,520 --> 00:05:08,960
evaluation. 
Because evaluating the alien 

109
00:05:08,960 --> 00:05:12,560
logic takes so much brainpower. 
Right, and the sheer volume of 

110
00:05:12,560 --> 00:05:16,080
this generated code vastly 
outpaces reviewer capacity, 

111
00:05:16,560 --> 00:05:20,120
which leads to this incredibly 
dangerous culture of rubber 

112
00:05:20,120 --> 00:05:21,760
stamping. 
You're tired. 

113
00:05:21,760 --> 00:05:24,400
The syntax looks fine, so you 
just click Approve. 

114
00:05:24,600 --> 00:05:28,600
And boom, silent failures and 
contextually blind bugs slip 

115
00:05:28,600 --> 00:05:29,880
right into production. 
Yep. 

116
00:05:30,280 --> 00:05:33,720
Which is why the giant AI labs 
are currently in an absolute 

117
00:05:33,720 --> 00:05:37,320
knife fight to be the ultimate 
solution to this exact review 

118
00:05:37,320 --> 00:05:39,080
bottleneck. 
Right, the mind tier war. 

119
00:05:39,560 --> 00:05:41,960
Let's talk about the big 
corporate moves, because on 

120
00:05:41,960 --> 00:05:44,360
April 9th, Open AI came out 
swinging. 

121
00:05:44,400 --> 00:05:47,240
They really did. 
They announced that new $100 a 

122
00:05:47,240 --> 00:05:51,080
month ChatGPT Pro tier aimed 
specifically at developers 

123
00:05:51,080 --> 00:05:53,040
hitting rate limit. 
And what blew my mind is that 

124
00:05:53,040 --> 00:05:56,360
Open AI is pitch materials 
explicitly named Claude code. 

125
00:05:56,640 --> 00:05:58,960
Like they position themselves 
dollar for dollar against 

126
00:05:58,960 --> 00:06:01,680
Claude's heavy usage pricing. 
It's incredibly aggressive. 

127
00:06:01,680 --> 00:06:04,920
It is and at the same time you 
have GitHub Copilot CLI 

128
00:06:05,240 --> 00:06:06,960
introducing this rubber duck 
feature. 

129
00:06:07,840 --> 00:06:11,360
Yes, the rubber duck. 
Which is so cool, It allows 

130
00:06:11,360 --> 00:06:14,440
developers to get a second 
opinion from a completely 

131
00:06:14,440 --> 00:06:17,040
different model family right 
there in the command line. 

132
00:06:17,040 --> 00:06:19,200
Because different models have 
different blind spots. 

133
00:06:19,200 --> 00:06:21,120
Exactly. 
Oh, and as a quick side note, 

134
00:06:21,280 --> 00:06:24,680
Simon Wilson just released this 
tool called Cleanup Clawed Code 

135
00:06:24,680 --> 00:06:27,800
Paste to fix white space when 
you copy prompts from the 

136
00:06:27,800 --> 00:06:30,040
terminal. 
Which just shows how deeply 

137
00:06:30,040 --> 00:06:33,480
embedded clawed code is in 
People's Daily workflows. 

138
00:06:33,720 --> 00:06:36,600
People are building custom 
utility scripts just to format 

139
00:06:36,600 --> 00:06:39,560
their copy pastes. 
Right, so looking at all this, 

140
00:06:39,560 --> 00:06:43,960
if Open AI is naming clawed code
directly in a pricing pitch, are

141
00:06:43,960 --> 00:06:46,280
they actually sweating real 
customer churn? 

142
00:06:46,280 --> 00:06:48,160
Oh. 
Without a doubt, this raises an 

143
00:06:48,160 --> 00:06:50,520
important question about 
philosophy really, because you 

144
00:06:50,520 --> 00:06:53,440
have two completely different 
approaches to solving complex 

145
00:06:53,440 --> 00:06:55,880
coding problems right now. 
OK, break that down for me. 

146
00:06:55,960 --> 00:06:58,360
Well, on one side you have open 
AI's approach. 

147
00:06:58,520 --> 00:07:01,320
Throw more money and more 
compute at a single massive 

148
00:07:01,320 --> 00:07:03,360
model. 
Tell it to think harder. 

149
00:07:03,400 --> 00:07:04,960
Right. 
And on the other side. 

150
00:07:05,080 --> 00:07:08,000
You have Copilot's rubber duck 
approach multi model 

151
00:07:08,000 --> 00:07:10,640
deliberation. 
Instead of 1 model thinking 

152
00:07:10,640 --> 00:07:12,120
harder, you have two different 
models. 

153
00:07:12,120 --> 00:07:15,000
Argue it out, one model right? 
It's a different model. 

154
00:07:15,000 --> 00:07:18,080
Family critiques. 
It's a structural solution to 

155
00:07:18,080 --> 00:07:20,720
the rubber stamping problem. 
So they are fighting over 

156
00:07:20,720 --> 00:07:25,000
developers right at the command 
line, but meanwhile Anthropic is

157
00:07:25,000 --> 00:07:26,440
doing something entirely 
different. 

158
00:07:26,840 --> 00:07:30,840
They are quietly executing this 
massive pincer movement to 

159
00:07:30,840 --> 00:07:33,240
capture entire corporate 
structures. 

160
00:07:33,240 --> 00:07:35,920
Yeah, they're surrounding the 
Enterprise from top to bottom. 

161
00:07:35,920 --> 00:07:39,720
So on April 8th, Anthropic 
launched managed agents in 

162
00:07:39,720 --> 00:07:41,920
public beta for builders. 
Right. 

163
00:07:42,040 --> 00:07:43,760
Anthropic runs the 
infrastructure, the 

164
00:07:43,760 --> 00:07:47,000
orchestration, all of it. 
The business just defines what 

165
00:07:47,000 --> 00:07:49,720
the agent needs to do. 
And then literally the next day,

166
00:07:49,720 --> 00:07:52,880
April 9th, they bring Claude 
Cowork into general 

167
00:07:52,880 --> 00:07:55,840
availability. 
It's massive because that is for

168
00:07:55,840 --> 00:07:58,160
the non developers, the business
buyers. 

169
00:07:58,480 --> 00:08:02,920
You can delegate tasks with 0 
coding required and it even 

170
00:08:02,920 --> 00:08:06,000
looks like cloud for Microsoft 
Office is live over at pivot dot

171
00:08:06,000 --> 00:08:09,040
claw dot AI. 
It's a brilliant two track go to

172
00:08:09,040 --> 00:08:11,160
market strategy, top down and 
bottom up. 

173
00:08:11,280 --> 00:08:14,160
So what does this all mean? 
I kind of look at it like a 

174
00:08:14,160 --> 00:08:17,440
restaurant, right? 
Manage Agents is like hiring 

175
00:08:17,440 --> 00:08:22,160
this brilliant head chef, but 
they absolutely refuse to share 

176
00:08:22,160 --> 00:08:23,720
their recipes with your kitchen 
staff. 

177
00:08:23,800 --> 00:08:24,800
That's a great way to look at 
it. 

178
00:08:24,840 --> 00:08:28,040
And then cowork is like just 
handing the customers out front 

179
00:08:28,040 --> 00:08:30,200
a magic menu where the food just
appears. 

180
00:08:30,800 --> 00:08:34,760
The trade off here is wild. 
Businesses lose all that heavy 

181
00:08:34,760 --> 00:08:37,799
infrastructure burden, but 
aren't they also completely 

182
00:08:37,799 --> 00:08:41,200
surrendering all their internal 
operational knowledge to 

183
00:08:41,200 --> 00:08:43,120
Anthropic? 
They absolutely are. 

184
00:08:43,120 --> 00:08:45,200
That is the ultimate trade off 
of this era. 

185
00:08:45,200 --> 00:08:47,920
It's a huge risk. 
It is by abstracting the 

186
00:08:47,920 --> 00:08:50,800
operational knowledge away from 
internal engineering teams, 

187
00:08:50,920 --> 00:08:52,880
businesses gain incredible 
speed. 

188
00:08:52,880 --> 00:08:55,440
I mean, they move so much 
faster, but they become 

189
00:08:55,440 --> 00:08:58,080
fundamentally dependent on 
Anthropic's ecosystem. 

190
00:08:58,080 --> 00:09:00,000
They lose their operational 
sovereignty. 

191
00:09:00,160 --> 00:09:02,360
Wow. 
OK, so if Anthropic is just 

192
00:09:02,360 --> 00:09:06,040
swallowing the enterprise hole 
like this, I have to ask what 

193
00:09:06,040 --> 00:09:09,080
kind of absolute monster of a 
model is powering all of this 

194
00:09:09,080 --> 00:09:11,280
behind the scenes? 
And more importantly, what are 

195
00:09:11,280 --> 00:09:13,760
they hiding? 
That is the $1,000,000 question 

196
00:09:13,760 --> 00:09:16,720
or well the $30 billion question
as we found out this week, 

197
00:09:16,960 --> 00:09:19,680
right? 
So announced April 7th and 8th 

198
00:09:19,800 --> 00:09:22,520
we have Project Glass Wing and 
Claude Mythos. 

199
00:09:22,520 --> 00:09:25,120
Yeah, Claude Mythos. 
It's a new general purpose 

200
00:09:25,120 --> 00:09:28,520
model, supposedly comparable to 
Opus 4.6. 

201
00:09:28,520 --> 00:09:32,240
But with additional totally 
undisclosed capabilities. 

202
00:09:32,240 --> 00:09:34,320
Exactly. 
And the kicker is they are not 

203
00:09:34,320 --> 00:09:37,800
publicly releasing it. 
It is restricted to a tiny, tiny

204
00:09:37,800 --> 00:09:40,400
set of security research preview
partners. 

205
00:09:40,400 --> 00:09:43,920
Yeah, Latent Space covered this 
and literally called it the 

206
00:09:44,400 --> 00:09:47,720
first model too dangerous to 
release since GPT 2. 

207
00:09:47,800 --> 00:09:49,360
Which is quite the statement. 
It is. 

208
00:09:49,720 --> 00:09:52,720
And during this exact same news 
cycle, it gets reported that 

209
00:09:52,720 --> 00:09:56,680
Anthropic hit $30 billion in 
annual recurring revenue. 

210
00:09:57,320 --> 00:09:58,560
So I have to get a little 
cynical here. 

211
00:09:58,560 --> 00:10:00,080
Go for it. 
Are they really holding this 

212
00:10:00,080 --> 00:10:04,400
model back just for safety or is
announcing a quote UN quote too 

213
00:10:04,400 --> 00:10:07,640
dangerous model just the 
absolute ultimate marketing 

214
00:10:07,640 --> 00:10:10,720
flicks to drop exactly when you 
hit 30 billion in ARR. 

215
00:10:11,320 --> 00:10:13,640
I love the cynicism here there 
because if we connect this to 

216
00:10:13,640 --> 00:10:15,480
the bigger picture, you're 
actually touching on something 

217
00:10:15,480 --> 00:10:18,440
profound about enterprise sales.
OK, tell me deliberate non 

218
00:10:18,440 --> 00:10:21,480
release acts as the ultimate 
safety credibility signal. 

219
00:10:21,480 --> 00:10:24,280
Oh, interesting. 
Yeah, think about conservative 

220
00:10:24,320 --> 00:10:28,560
enterprise buyers, banks, 
healthcare, government. 

221
00:10:28,960 --> 00:10:32,720
They are terrified of liability.
By holding back mythos, 

222
00:10:33,000 --> 00:10:35,960
Anthropic is telling those 
buyers we are incredibly 

223
00:10:35,960 --> 00:10:38,720
powerful, but we are also the 
most responsible adult in the 

224
00:10:38,720 --> 00:10:41,440
room. 
Safety is essentially a flex. 

225
00:10:41,440 --> 00:10:42,960
It's a product feature. 
Wow. 

226
00:10:43,840 --> 00:10:45,480
Lerner, look at how this all 
connects. 

227
00:10:45,760 --> 00:10:49,640
We went from a single silent 
configuration bug on a Reddit 

228
00:10:49,640 --> 00:10:54,400
thread all the way to a $30 
billion enterprise ecosystem 

229
00:10:54,560 --> 00:10:56,800
that Anthropic is building. 
It really highlights how 

230
00:10:56,800 --> 00:10:58,800
critical the human element still
is. 

231
00:10:58,800 --> 00:11:01,360
Exactly. 
Whether you are fighting off AI 

232
00:11:01,360 --> 00:11:03,720
generated pull requests, or 
you're deliberating between 

233
00:11:03,720 --> 00:11:06,280
models in your terminal, or 
you're adopting managed agents 

234
00:11:06,280 --> 00:11:09,240
for your business, you cannot 
just set things to autopilot. 

235
00:11:09,240 --> 00:11:09,680
You. 
Absolutely. 

236
00:11:09,760 --> 00:11:10,400
Really. 
Cannot you? 

237
00:11:10,400 --> 00:11:12,720
Have to verify the behavior. 
You have to remain the critical 

238
00:11:12,720 --> 00:11:14,960
thinker in the loop. 
And you know, that leaves us 

239
00:11:14,960 --> 00:11:17,840
with a really big lingering 
question to Mull over. 

240
00:11:18,240 --> 00:11:20,280
What's that? 
Well, if we are entirely 

241
00:11:20,280 --> 00:11:23,280
abstracting our operational 
knowledge to managed agents, 

242
00:11:23,280 --> 00:11:26,120
right? 
And if AI is generating 80% of 

243
00:11:26,120 --> 00:11:29,680
our code while we just review 
it, how does the next generation

244
00:11:29,680 --> 00:11:34,000
of junior developers actually 
learn what good looks like? 

245
00:11:34,000 --> 00:11:38,320
I mean, if the AI does all the 
doing, where do tomorrow's 

246
00:11:38,320 --> 00:11:42,000
expert reviewers even come from?
Oh wow, that is. 

247
00:11:42,000 --> 00:11:43,480
Yeah, that is terrifying to 
think about. 

248
00:11:43,600 --> 00:11:45,920
If you never build a muscle, how
do you know if the machine is 

249
00:11:45,920 --> 00:11:47,480
doing it right? 
Exactly. 

250
00:11:47,800 --> 00:11:50,120
Well, Lerner, that is something 
to really chew on this week. 

251
00:11:50,120 --> 00:11:52,520
Thank you so much for joining us
for this deep dive. 

252
00:11:52,520 --> 00:11:54,440
Yeah, thanks for listening 
everyone. 

253
00:11:54,560 --> 00:11:55,720
We will see you on the next one.