1
00:00:00,000 --> 00:00:03,569
Nirmal Mehta: non-technical users
starting to use these tools to build

2
00:00:03,569 --> 00:00:06,529
business applications that aren't devs.

3
00:00:06,889 --> 00:00:07,250
Right?

4
00:00:07,250 --> 00:00:09,590
They don't know anything about CI/CD.

5
00:00:09,829 --> 00:00:11,690
They don't know anything
about infrastructure.

6
00:00:11,900 --> 00:00:15,940
They don't know anything about code
and they're gonna start building apps.

7
00:00:16,227 --> 00:00:22,677
if you thought Shadow IT was a big
problem back when cloud was new, oh boy.

8
00:00:22,857 --> 00:00:27,737
Shadow IT with gen AI is like,
that's a whole brave new world

9
00:00:27,797 --> 00:00:28,967
that we're entering into.

10
00:00:32,502 --> 00:00:35,972
Bret AI July 2025: Welcome to the
Agentic DevOps podcast, and I am

11
00:00:35,972 --> 00:00:37,262
one of your hosts, Bret Fisher.

12
00:00:37,262 --> 00:00:39,962
My co-host Nirmal Mehta
is back in the studio.

13
00:00:40,322 --> 00:00:44,852
His studio, not my studio, and we are
recording what we're calling season two.

14
00:00:44,852 --> 00:00:48,572
It's actually been almost six
months since our last episode, and

15
00:00:48,572 --> 00:00:53,552
a lot has happened and we quite
frankly, have just been overwhelmed.

16
00:00:53,972 --> 00:00:55,712
I think a lot of us are
feeling this right now.

17
00:00:56,082 --> 00:01:00,552
The last three months since
December have been a total shift.

18
00:01:00,582 --> 00:01:05,532
I'm looking at this as more of the
epoch of AI is good enough now, that's

19
00:01:05,532 --> 00:01:07,192
the summary of this entire podcast.

20
00:01:07,300 --> 00:01:12,550
Where are we last left off in August,
we were talking about AI as a helpful

21
00:01:12,550 --> 00:01:15,970
assistant, but yes, you have to spend
a lot of time with prompting and

22
00:01:15,970 --> 00:01:17,530
managing hallucinations and all that.

23
00:01:18,850 --> 00:01:21,820
I'm not really doing much of that anymore,
and for the last three months, I haven't

24
00:01:21,820 --> 00:01:26,230
experienced much of any, uh, what I
would call significant hallucinations.

25
00:01:26,230 --> 00:01:29,890
I certainly have seen it make wrong
decisions and do things I didn't intend.

26
00:01:30,865 --> 00:01:33,145
But it's, it's amazing
how much better these are.

27
00:01:33,195 --> 00:01:40,815
Truly Opus, and some of these newer models
like GPT 5.4 and 4.5 and 4.6 of Opus and

28
00:01:40,815 --> 00:01:44,865
Sonnet have fundamentally changed the
game for me and a whole lot of us online.

29
00:01:44,865 --> 00:01:45,735
So we talk about that.

30
00:01:45,735 --> 00:01:49,065
We go through some of the headlines of the
last six months, but we're now starting

31
00:01:49,095 --> 00:01:53,955
season two, and we've already got multiple
other episodes lined up around the world

32
00:01:53,985 --> 00:02:00,585
of AI and its intersection with engineers
doing automation and DevOps like things.

33
00:02:00,585 --> 00:02:01,825
So, I'm excited.

34
00:02:01,855 --> 00:02:03,385
Uh, we've got a whole bunch of shows.

35
00:02:03,745 --> 00:02:06,575
I mean, we've never had seasons
before, but why not Season two.

36
00:02:06,625 --> 00:02:07,825
Let's, let's call it that.

37
00:02:08,095 --> 00:02:10,485
And, I'm looking forward to
getting these episodes out to you.

38
00:02:10,485 --> 00:02:15,585
So this is the first of many and
we go pretty random and quick

39
00:02:15,590 --> 00:02:18,405
through a lot of the content we're
gonna talk about in this episode.

40
00:02:18,405 --> 00:02:21,525
So here we are, 2026.

41
00:02:22,695 --> 00:02:23,415
Everything AI.

42
00:02:23,939 --> 00:02:24,509
Bret: Hi, Nirmal.

43
00:02:25,009 --> 00:02:25,249
Welcome

44
00:02:25,323 --> 00:02:26,723
Nirmal Mehta: Hi, thank you.

45
00:02:26,783 --> 00:02:32,323
I'm Nirmal Meha, and I'm a principal
specialist solution architect and

46
00:02:32,323 --> 00:02:38,853
containers tech lead at AWS and,
longtime friend and co-host, with

47
00:02:38,853 --> 00:02:43,673
Bret on all the different shows that
Bret has created over the years.

48
00:02:43,753 --> 00:02:49,653
I'm super excited to be back here
for the next quote unquote, season

49
00:02:49,683 --> 00:02:51,513
of the Agentic DevOps podcast.

50
00:02:51,823 --> 00:02:52,813
Just a quick note.

51
00:02:52,863 --> 00:02:56,453
The opinions that I will be sharing
on the show are my opinions and not

52
00:02:56,603 --> 00:02:58,943
of my employer, Amazon Web Services.

53
00:02:59,033 --> 00:03:00,383
And with that.

54
00:03:00,883 --> 00:03:04,773
Boy, do we have a long list
of things to catch up on?

55
00:03:05,319 --> 00:03:07,779
Bret: we can't even
catch up on six months.

56
00:03:07,829 --> 00:03:07,989
Ok,

57
00:03:08,063 --> 00:03:08,693
Nirmal Mehta: No.

58
00:03:08,709 --> 00:03:11,779
Bret: so if this is your first Agentic
DevOps podcast, welcome to the show.

59
00:03:11,879 --> 00:03:12,179
Nirmal Mehta: Yeah.

60
00:03:12,179 --> 00:03:12,749
Welcome.

61
00:03:12,999 --> 00:03:15,879
Bret: And Nirmal and I started this
podcast a year ago, next month.

62
00:03:16,379 --> 00:03:19,739
We were in London, I believe,
London Cobe Con, CloudNativecon.

63
00:03:21,023 --> 00:03:21,513
Nirmal Mehta: Correct.

64
00:03:21,699 --> 00:03:22,709
Bret: We started it there.

65
00:03:23,329 --> 00:03:25,759
It felt like the right time to start it.

66
00:03:26,259 --> 00:03:27,239
Nirmal Mehta: It really did.

67
00:03:27,324 --> 00:03:31,284
Bret: We were talking about it, but it
felt like murmurs, like everyone in the

68
00:03:31,284 --> 00:03:33,864
industry was focused on how to run AI.

69
00:03:34,364 --> 00:03:34,654
Nirmal Mehta: Yeah.

70
00:03:35,150 --> 00:03:36,970
Bret: Using AI, and agents were new.

71
00:03:37,240 --> 00:03:38,740
The agents were a new idea.

72
00:03:39,190 --> 00:03:42,775
And MCP was less than six
months old at that point.

73
00:03:43,275 --> 00:03:43,565
Nirmal Mehta: Yeah.

74
00:03:43,820 --> 00:03:44,990
Bret: I think we talked
about that first one.

75
00:03:44,990 --> 00:03:47,720
Like we mentioned A2A, like a
brand new protocol for agent.

76
00:03:47,720 --> 00:03:48,050
Agent or

77
00:03:48,125 --> 00:03:48,575
Nirmal Mehta: Correct.

78
00:03:48,575 --> 00:03:49,130
Yeah.

79
00:03:49,175 --> 00:03:50,225
century ago.

80
00:03:50,270 --> 00:03:51,650
Bret: I know it feels like forever.

81
00:03:51,920 --> 00:03:53,240
Well, none of that matters anymore.

82
00:03:53,540 --> 00:03:54,200
We're gonna get into it.

83
00:03:54,500 --> 00:03:55,910
But we had some shows over the summer.

84
00:03:55,910 --> 00:03:58,820
We had some guests, we had Laura Tacho on.

85
00:03:59,120 --> 00:03:59,390
We

86
00:03:59,720 --> 00:04:00,230
Nirmal Mehta: Yeah.

87
00:04:00,330 --> 00:04:05,850
Bret: Solomon Hikes on talking about
sort of the world of CI in DevOps, in AI.

88
00:04:05,850 --> 00:04:08,160
Like how that relates to
AI and how they see it.

89
00:04:08,500 --> 00:04:09,670
They have since pivoted.

90
00:04:09,670 --> 00:04:10,720
We might talk about that.

91
00:04:10,990 --> 00:04:13,330
But this is gonna be the kickoff episode.

92
00:04:13,330 --> 00:04:14,270
So I wouldn't call it a reboot.

93
00:04:14,410 --> 00:04:14,770
Nirmal Mehta: Yeah.

94
00:04:15,240 --> 00:04:15,640
Bret: I wouldn't call it

95
00:04:15,640 --> 00:04:15,670
Nirmal Mehta: a

96
00:04:15,670 --> 00:04:15,700
Bret: comback.

97
00:04:15,700 --> 00:04:16,090
Nirmal Mehta: No,

98
00:04:16,590 --> 00:04:17,430
Bret: I've been here for years.

99
00:04:17,830 --> 00:04:18,690
that's the LL Cool J quote.

100
00:04:18,830 --> 00:04:21,620
just for the kids that don't
know that 80's, 90's music.

101
00:04:21,830 --> 00:04:25,430
and so let's back up a second
and remember where we were.

102
00:04:25,460 --> 00:04:29,120
Our last episode was Laura
Tacho in really around August.

103
00:04:29,120 --> 00:04:30,530
I think we aired it in September.

104
00:04:30,530 --> 00:04:32,275
We recorded maybe around August

105
00:04:33,815 --> 00:04:34,165
Nirmal Mehta: About,

106
00:04:34,915 --> 00:04:40,545
Bret: of AI not being a a productivity
booster for business when it comes to

107
00:04:40,545 --> 00:04:44,985
business goals, business outcomes of
the software teams, and how AI wasn't

108
00:04:44,985 --> 00:04:50,535
actually trailblazing this amazing
improvement in business productivity.

109
00:04:50,805 --> 00:04:54,945
It helps with coding and speeds up
coding, but that's a small portion of

110
00:04:54,945 --> 00:04:59,595
a developer's week and that is only
one part of the software lifecycle.

111
00:04:59,895 --> 00:05:05,475
So we kind of had some sort of bad
news around, hey, maybe five or 10%,

112
00:05:05,505 --> 00:05:07,335
you know, business productivity.

113
00:05:07,815 --> 00:05:13,065
And then third quarter happened,
and third quarter we got more news.

114
00:05:13,065 --> 00:05:16,935
we saw more studies saying,
yeah, AI has had, I think the

115
00:05:16,935 --> 00:05:19,125
Yale study, which is debatable.

116
00:05:19,125 --> 00:05:22,125
I don't think everybody agreed with it,
but I think the headline there was AI

117
00:05:22,125 --> 00:05:24,405
has had zero effects on jobs so far.

118
00:05:24,765 --> 00:05:27,785
So there are definitely different camps.

119
00:05:28,085 --> 00:05:34,515
You've got one camp on the red pill side
that is describing it as 10x developer

120
00:05:34,515 --> 00:05:36,705
and just amazing productivity benefits.

121
00:05:37,095 --> 00:05:40,905
But everyone that I know that says
that is an independent developer.

122
00:05:42,805 --> 00:05:47,465
Nirmal and I were at a great conference
together and this conference was a group

123
00:05:47,465 --> 00:05:52,875
of friends that are all builders, they're
all SaaS makers, developer consultants,

124
00:05:52,925 --> 00:05:58,325
making products, working for big and
small companies sometimes, but they're all

125
00:05:58,775 --> 00:06:01,270
highly invested in using AI for their job.

126
00:06:01,999 --> 00:06:02,289
Nirmal Mehta: Yeah.

127
00:06:02,335 --> 00:06:05,815
Bret: One thing that Nirmal and I come
from is we come from the ops world

128
00:06:05,815 --> 00:06:10,495
where these people tend to come more
from the pure dev, pure builder world.

129
00:06:10,525 --> 00:06:10,815
Nirmal Mehta: Yeah.

130
00:06:11,165 --> 00:06:15,295
Bret: From enterprise teams and big
consulting companies and complication and

131
00:06:15,505 --> 00:06:18,430
legacy and monolith, like that's our world

132
00:06:18,480 --> 00:06:18,770
Nirmal Mehta: Yeah.

133
00:06:19,015 --> 00:06:20,395
The operations side.

134
00:06:21,325 --> 00:06:22,015
Yes.

135
00:06:22,066 --> 00:06:24,976
Bret: I do have my friends and
I do spend that time making,

136
00:06:24,976 --> 00:06:26,566
I am an independent creator.

137
00:06:26,566 --> 00:06:30,846
I am a solo developer and I
have solo developer friends.

138
00:06:30,906 --> 00:06:36,029
And lemme tell you, these two camps are
completely different when it comes to AI.

139
00:06:36,059 --> 00:06:40,109
Like it's two sides of the same coin
on AI, but they're completely different

140
00:06:40,199 --> 00:06:45,199
experiences and opinions based on the size
of the team and the size of the company.

141
00:06:45,699 --> 00:06:46,179
What do you think

142
00:06:46,304 --> 00:06:48,444
Nirmal Mehta: And, and there's
like, there's a lot there.

143
00:06:48,544 --> 00:06:53,264
for me, it was not only attending,
that solo developer retreat with

144
00:06:53,264 --> 00:06:54,734
you, Bret, which was amazing.

145
00:06:55,164 --> 00:06:59,174
but I followed that up, by attending
the Pragmatic Summit in San

146
00:06:59,174 --> 00:07:01,794
Francisco, just a few days later.

147
00:07:01,894 --> 00:07:02,224
Bret: back to

148
00:07:02,269 --> 00:07:04,489
Nirmal Mehta: of this recording,
that was just two weeks ago.

149
00:07:04,989 --> 00:07:09,359
And, from my perspective, you're
right Bret, both of us are kind of,

150
00:07:09,859 --> 00:07:14,089
our experience and our expertise is
grounded in operations system and

151
00:07:14,089 --> 00:07:21,499
administration, SRE DevOps, engineering,
DevOps practices more on the Ops.

152
00:07:21,559 --> 00:07:21,889
Right?

153
00:07:21,889 --> 00:07:26,819
I would say like we're lowercase
dev, big case Ops people,

154
00:07:27,269 --> 00:07:27,449
Bret: That's a

155
00:07:27,523 --> 00:07:28,303
Nirmal Mehta: and we've.

156
00:07:28,324 --> 00:07:28,444
Bret: Yeah.

157
00:07:28,944 --> 00:07:33,504
Nirmal Mehta: We've spent our
careers bridging, you know, we're

158
00:07:33,504 --> 00:07:35,244
DevOps practitioners, right?

159
00:07:35,364 --> 00:07:36,384
First and foremost.

160
00:07:36,384 --> 00:07:41,294
And we've been bridging this
tension that's been between the

161
00:07:41,474 --> 00:07:43,844
developers and the operations folks.

162
00:07:43,844 --> 00:07:47,924
And of course, those lines have
gotten very blurry with cloud

163
00:07:47,924 --> 00:07:52,214
native tooling and, the modern
organizational structures these days.

164
00:07:52,714 --> 00:07:56,944
in my day to day, I typically
talk to platform engineers,

165
00:07:57,094 --> 00:08:01,994
operations folks, engineering
teams that are, like SRE teams.

166
00:08:02,494 --> 00:08:07,264
I don't talk to developers,
especially not solo developers much.

167
00:08:07,444 --> 00:08:09,964
And so for, for me, for me,

168
00:08:10,294 --> 00:08:13,834
Bret: contracts and
enterprise Amazon contracts.

169
00:08:14,039 --> 00:08:16,744
Nirmal Mehta: for me, these two
events back to back were kind of

170
00:08:16,744 --> 00:08:19,644
like an expedition from me into this.

171
00:08:19,884 --> 00:08:23,964
The world of developers, like the
current state of developers, it

172
00:08:23,964 --> 00:08:28,194
was like a exploration of that
world that I, I'm not usually in.

173
00:08:28,284 --> 00:08:33,084
And it was fascinating and I kind
of agree, Bret, like we've been

174
00:08:33,084 --> 00:08:39,654
working for over a decade on bridging
DevOps together culturally, right?

175
00:08:39,654 --> 00:08:45,944
Containerization, cloud native automation,
and making developers lives easier and

176
00:08:45,944 --> 00:08:51,194
also making operations folks' lives
easier and bridging that tension, right?

177
00:08:51,244 --> 00:08:53,014
that classic throat over the fence.

178
00:08:53,514 --> 00:08:58,784
I think you're kind of right with
respect to Agentic systems, AI.

179
00:08:59,854 --> 00:09:05,014
kind of reigniting some of that
tension because the adoption,

180
00:09:05,114 --> 00:09:10,194
what I saw, and granted this was
definitely a group of leading edge

181
00:09:10,194 --> 00:09:12,414
adopters of this technology for sure.

182
00:09:12,834 --> 00:09:18,494
But what I saw there was, as a
developer, the adoption and the usage

183
00:09:18,494 --> 00:09:24,244
of these tools, at least in some
part of the SLDC lifecycle, right?

184
00:09:24,244 --> 00:09:27,744
The inner loop development
was pretty phenomenal.

185
00:09:27,794 --> 00:09:32,664
I just saw really amazing things,
and just productivity boosts.

186
00:09:32,664 --> 00:09:37,034
I don't want to guess on the
percentage, but very clear, direct

187
00:09:37,034 --> 00:09:38,864
re, you know, outcomes there.

188
00:09:39,364 --> 00:09:44,044
again, this is the Age Agentic DevOps
podcast and our mission on this podcast,

189
00:09:44,044 --> 00:09:46,714
one, is we wanted to focus on the.

190
00:09:47,214 --> 00:09:48,654
More of the operations side.

191
00:09:48,684 --> 00:09:52,464
How are these AI and agentic
systems affecting the operations?

192
00:09:52,464 --> 00:09:58,634
Like everything kind of from when the
code is built, to like the testing

193
00:09:58,694 --> 00:10:02,774
CI/CD and then actually running a
production and then the operations

194
00:10:02,774 --> 00:10:04,874
of those applications in production.

195
00:10:05,374 --> 00:10:11,444
And it was interesting to see how the
acceleration of these tools, these tools

196
00:10:11,444 --> 00:10:16,844
accelerating application development
was now highlighting in very stark

197
00:10:16,844 --> 00:10:23,404
terms the need for organizations to
have automation from the CI/CD all the

198
00:10:23,404 --> 00:10:27,234
way through to their infrastructure
in the same way that we've been

199
00:10:27,234 --> 00:10:29,334
talking about for the last 20 years.

200
00:10:29,724 --> 00:10:36,624
Because now there's a wave, a
tsunami, I think of hundred x

201
00:10:36,624 --> 00:10:38,484
more applications gonna be built.

202
00:10:38,984 --> 00:10:39,584
starting now.

203
00:10:39,584 --> 00:10:40,784
The wave has already started.

204
00:10:41,284 --> 00:10:47,124
And I don't think as DevOps
practitioners, as operations folks, I

205
00:10:47,124 --> 00:10:50,554
don't think we're ready to handle this.

206
00:10:50,774 --> 00:10:55,334
wave of new applications and this new way
of software development that's happening.

207
00:10:55,834 --> 00:10:59,374
I think there's like even more tension,
but the fundamentals are still the same.

208
00:10:59,374 --> 00:11:04,184
That's the takeaway I got is
like you now more than ever, you

209
00:11:04,184 --> 00:11:06,074
organization needs automation.

210
00:11:06,074 --> 00:11:11,084
It needs solid DevOps practices,
it needs solid documentation and

211
00:11:11,114 --> 00:11:16,504
you know, efficient operations and
I'll say automation again and likely

212
00:11:16,544 --> 00:11:18,864
we'll need to adopt agentic tools.

213
00:11:18,924 --> 00:11:21,244
We'll talk about this a
little bit more, but I think.

214
00:11:21,744 --> 00:11:28,884
We will have to adopt agentic tooling to
even just operate these systems at the

215
00:11:28,884 --> 00:11:33,294
scale to meet the needs of these, the
tsunami of applications that are coming.

216
00:11:33,794 --> 00:11:37,459
Bret: Yeah, I mean, it really talks,
you know, it talks, this leans

217
00:11:37,459 --> 00:11:41,179
into like, is your CI/CD debt?

218
00:11:41,179 --> 00:11:42,169
We talk about code debt,

219
00:11:42,498 --> 00:11:42,788
Nirmal Mehta: Yeah.

220
00:11:43,189 --> 00:11:45,859
Bret: but ops, debt, SRE
debt, DevOps debt, these are

221
00:11:45,859 --> 00:11:47,659
all same, similar problems.

222
00:11:48,019 --> 00:11:53,419
And if your systems are not enabled
enough that a developer can spin up a

223
00:11:53,419 --> 00:11:58,909
new idea, iterate on it, put it into
staging, put it into production without

224
00:11:58,909 --> 00:12:00,739
your involvement, if they're not enabled.

225
00:12:01,084 --> 00:12:03,454
We might label that as platform
engineering, but if they're not enabled

226
00:12:03,454 --> 00:12:06,954
for that, you're suddenly gonna have
a lot more work on your plate if

227
00:12:06,954 --> 00:12:10,194
you're having to deal with everything
that the developer's creating.

228
00:12:10,224 --> 00:12:14,234
' cause if the developer can code
10 x faster, let's just say they

229
00:12:14,234 --> 00:12:17,594
can, or three x whatever, two x,
it's gonna be faster than today.

230
00:12:17,954 --> 00:12:19,124
They're coding more.

231
00:12:19,214 --> 00:12:24,284
We know that the negatives around like
the business productivity, that all

232
00:12:24,284 --> 00:12:27,404
starts to fall down because the code
is only a small portion of the problem.

233
00:12:27,404 --> 00:12:28,754
Like we, they create code now.

234
00:12:28,754 --> 00:12:33,464
We dev people, DevOps people have a new
problem, which is they're shipping more

235
00:12:33,464 --> 00:12:36,404
updates to the same code and they're
probably gonna be shipping a lot more

236
00:12:36,434 --> 00:12:38,804
new repos as they spin up things.

237
00:12:38,804 --> 00:12:42,464
And this is what we saw at our, retreat
before you went to San Francisco.

238
00:12:42,964 --> 00:12:47,524
Is that a solo creator can now be
working on three or four projects at

239
00:12:47,524 --> 00:12:53,104
the same time, or a solo dev in a team
can now, you know, things that before,

240
00:12:53,344 --> 00:12:56,884
well that would be a couple of weeks
of work and I'd have to get approvals.

241
00:12:57,324 --> 00:12:59,844
Assuming that you don't have to
go through approvals and all these

242
00:12:59,844 --> 00:13:03,924
different processes, you can iterate
on an idea very quickly, especially

243
00:13:03,924 --> 00:13:06,774
in those first early days of an idea.

244
00:13:07,104 --> 00:13:10,284
And what I'm finding as a
DevOps engineer is I'm creating

245
00:13:10,284 --> 00:13:12,054
way more tooling for myself.

246
00:13:12,534 --> 00:13:15,264
And so let's go back to the third
quarter because we're gonna step

247
00:13:15,264 --> 00:13:16,884
through a couple of major shifts in

248
00:13:17,003 --> 00:13:17,543
Nirmal Mehta: Yeah.

249
00:13:17,544 --> 00:13:19,944
Bret: paying attention
that have enabled all this.

250
00:13:20,243 --> 00:13:20,903
Nirmal Mehta: Yes.

251
00:13:20,994 --> 00:13:23,784
Bret: enable it for the DevOps
engineers, not just the developers.

252
00:13:24,214 --> 00:13:27,554
Nirmal Mehta: if you remember, some of
our earlier episodes, we were kind of

253
00:13:27,554 --> 00:13:31,744
highlighting that, you know, if we, a
year ago when we were using these tools,

254
00:13:31,844 --> 00:13:36,464
I think we all had like a healthy dose of
skepticism and we should still maintain

255
00:13:36,464 --> 00:13:40,804
that healthy dose of skepticism, where we
saw well, these tools are hallucinating a

256
00:13:40,804 --> 00:13:43,544
lot, they're making simple applications.

257
00:13:44,024 --> 00:13:44,414
Bret: I spend all my

258
00:13:44,564 --> 00:13:44,804
Nirmal Mehta: Like,

259
00:13:45,044 --> 00:13:45,314
Bret: Yeah.

260
00:13:45,344 --> 00:13:48,984
Nirmal Mehta: the output, had to be
shepherded looked at very closely.

261
00:13:49,034 --> 00:13:53,734
and I think a healthy skepticism
look at that was, you know, these

262
00:13:53,734 --> 00:13:57,874
models are really cool and there's
cool stuff that's enabled by that.

263
00:13:58,234 --> 00:14:02,464
And some of these operations, like
with MCP servers, that's awesome

264
00:14:02,514 --> 00:14:03,924
as that was starting to roll out.

265
00:14:04,424 --> 00:14:09,484
But it's not gonna replace this
whole entire function or, you

266
00:14:09,484 --> 00:14:11,664
know, the quality is not there yet.

267
00:14:12,164 --> 00:14:18,114
and if the last time you used
any of these AI tools was, let's

268
00:14:18,114 --> 00:14:21,734
say six months ago or older,

269
00:14:22,234 --> 00:14:28,004
Then you might not realize something that
happened in October and, in the last six

270
00:14:28,004 --> 00:14:30,704
months or however it's been since October.

271
00:14:31,064 --> 00:14:38,034
and that is these new models that came
out, especially, and we heard this from, I

272
00:14:38,034 --> 00:14:40,134
can't even count how many people told me.

273
00:14:40,534 --> 00:14:43,594
this was like announced at the
Pragmatic Summit as if it was like

274
00:14:43,774 --> 00:14:46,194
just an obvious paradigm shift.

275
00:14:46,524 --> 00:14:53,024
But when Opus 4.6 from Claude came
out from Anthropic, Claude, Opus

276
00:14:53,024 --> 00:14:58,344
4.6 came out and sonnet 4.6, and
some of these other models too.

277
00:14:58,574 --> 00:15:01,994
Gemini 3.1, pro 5.3 Codex,

278
00:15:02,494 --> 00:15:02,704
Bret: right.

279
00:15:02,808 --> 00:15:03,798
Nirmal Mehta: like basically the,

280
00:15:03,844 --> 00:15:04,714
Bret: this year, but yeah.

281
00:15:05,214 --> 00:15:08,084
Nirmal Mehta: let's just say
like the latest, Cohort of

282
00:15:08,084 --> 00:15:09,494
models that have been released.

283
00:15:09,564 --> 00:15:10,824
Bret: while you were,
let me recap for you.

284
00:15:10,854 --> 00:15:14,844
While we were at KubeCon, We
had multiple state of the arts.

285
00:15:14,844 --> 00:15:16,884
We're gonna call it the soda
models, frontier models, whatever.

286
00:15:17,384 --> 00:15:19,574
Released like within the same week.

287
00:15:19,904 --> 00:15:22,514
One of them was Opus 4.5, right?

288
00:15:22,514 --> 00:15:27,184
We had Sonnet, I mean we were
on Sonnet three five a year ago

289
00:15:27,684 --> 00:15:28,034
Nirmal Mehta: right.

290
00:15:28,120 --> 00:15:30,570
Bret: Sonnet and Opus 4.6.

291
00:15:31,070 --> 00:15:31,850
Many people.

292
00:15:31,970 --> 00:15:35,350
Yeah, like you, so many people
have told me are game changer.

293
00:15:35,350 --> 00:15:36,520
It is so much better.

294
00:15:36,550 --> 00:15:38,260
The hallucinations are way less.

295
00:15:38,560 --> 00:15:42,160
We now have the AI guiding us
through gaps in the context.

296
00:15:42,160 --> 00:15:45,910
I just demoed a couple hours ago on
the live stream how if I tried to

297
00:15:45,910 --> 00:15:49,810
build an ECS cluster and it's walking
me through, answering the questions

298
00:15:49,810 --> 00:15:52,410
without me asking it's way better,

299
00:15:52,530 --> 00:15:52,750
Nirmal Mehta: Yes,

300
00:15:53,355 --> 00:15:54,885
Bret: much better than six months ago.

301
00:15:54,935 --> 00:15:55,425
Nirmal Mehta: correct.

302
00:15:55,621 --> 00:15:58,981
Bret: I think everyone who's done
this or tried this is agreeing.

303
00:15:59,131 --> 00:16:02,671
I think like you're saying, everyone
who's still using the cheaper models

304
00:16:02,671 --> 00:16:06,421
or maybe not the Claude models, I
mean, it's not that like Claude's got.

305
00:16:07,486 --> 00:16:08,716
is the cream of the crop right now.

306
00:16:08,716 --> 00:16:09,526
It's everyone's favorite.

307
00:16:09,526 --> 00:16:13,386
That's a combination of
the Claude code CLI and in

308
00:16:13,386 --> 00:16:15,336
combination with the Opus models.

309
00:16:15,726 --> 00:16:19,161
But can still do this on Gemini Pro 3.1.

310
00:16:19,196 --> 00:16:25,256
You can still do this on code, open
AI, you know, chat g or GPT 5.3.

311
00:16:25,256 --> 00:16:28,436
Now, Codex, if you specifically look
up the Codex, but if you're using

312
00:16:28,436 --> 00:16:31,436
either one of those three, I just
know that I have the most experience

313
00:16:31,436 --> 00:16:32,606
with Opus, so I can speak to that.

314
00:16:32,936 --> 00:16:36,026
But what you're saying is
both of us at the same time in

315
00:16:36,026 --> 00:16:37,406
different groups of people across

316
00:16:37,565 --> 00:16:38,055
Nirmal Mehta: Correct.

317
00:16:38,066 --> 00:16:43,366
Bret: the industry, within two
months or even just a month of that

318
00:16:43,366 --> 00:16:48,366
November release, everyone I know
who was what I would call blue pill,

319
00:16:48,396 --> 00:16:51,756
meaning we're not really convinced
AI is gonna do that much for us.

320
00:16:51,806 --> 00:16:55,536
and maybe you can get the function right,
but that's the best you're gonna do.

321
00:16:55,816 --> 00:16:57,496
I'm probably gonna have to change
something in that function.

322
00:16:57,496 --> 00:17:00,106
You're probably gonna make unnecessary
lines of code that you don't need.

323
00:17:00,196 --> 00:17:02,656
And these are all still true,
but they're way less common.

324
00:17:02,656 --> 00:17:06,586
And now suddenly that combined with
the lead release of skills, which

325
00:17:06,586 --> 00:17:11,266
also happened in the fall in October,
Claude announced the idea of skills

326
00:17:11,266 --> 00:17:16,096
that has taken on a whole another level
of hype because it's truly easy to use

327
00:17:16,096 --> 00:17:20,836
and I am actively creating new skills
every week for the last two months.

328
00:17:21,256 --> 00:17:25,516
So we are now in a world suddenly
where six months ago I wasn't

329
00:17:26,016 --> 00:17:29,166
sure where AI fit in my world.

330
00:17:29,556 --> 00:17:34,896
And now every conversation for
everything I do in DevOps starts with AI.

331
00:17:34,926 --> 00:17:41,796
there's nothing I'm doing anymore where
I'm not aligned with an AI assistant who

332
00:17:41,796 --> 00:17:47,786
is helping me walk through my design or
my GitHub workflow creation or, you know,

333
00:17:47,786 --> 00:17:51,986
whether I'm making a landing page for
something or whether I'm deciding whether,

334
00:17:52,086 --> 00:17:55,326
The CVEs I wanna patch are even necessary.

335
00:17:55,326 --> 00:17:57,126
there's just so many different things.

336
00:17:57,126 --> 00:18:02,276
I'm doing AI first now and I think in
the last 24 hours, which I've spent

337
00:18:02,776 --> 00:18:08,611
probably over 10 hours in the last 48
hours talking and waiting on an AI.

338
00:18:08,611 --> 00:18:13,696
And it has only gotten I think,
one, maybe two things wrong.

339
00:18:13,696 --> 00:18:14,836
And they weren't even really wrong.

340
00:18:14,836 --> 00:18:17,146
They just work the way I wanted 'em to.

341
00:18:17,506 --> 00:18:21,616
And that is across dozens and
dozens of prompts across multiple

342
00:18:21,616 --> 00:18:24,646
projects GitHub Actions, on web

343
00:18:24,955 --> 00:18:25,245
Nirmal Mehta: Yeah.

344
00:18:25,596 --> 00:18:27,916
Bret: GoLang tooling for my DevOps.

345
00:18:28,366 --> 00:18:31,156
Like I'm just doing so much
stuff all at the same time.

346
00:18:31,276 --> 00:18:35,836
So you and I now have to bring that,
like to me the goal is to bring that

347
00:18:35,836 --> 00:18:38,656
experience to the enterprise team, right?

348
00:18:38,825 --> 00:18:39,115
Nirmal Mehta: Yeah.

349
00:18:39,436 --> 00:18:40,396
Bret: the possibilities.

350
00:18:40,666 --> 00:18:44,206
Nirmal Mehta: I just wanna,
highlight this and make a very

351
00:18:44,706 --> 00:18:46,626
exclamation point for our listeners.

352
00:18:46,656 --> 00:18:50,306
first of all, thank you all for the
folks that have supported us and,

353
00:18:50,676 --> 00:18:52,716
shouted out that they enjoy the show.

354
00:18:53,216 --> 00:18:58,826
They trust us too give them some
to ground what we've been seeing.

355
00:18:58,946 --> 00:19:02,776
And so I think there's two
things here as takeaways.

356
00:19:03,276 --> 00:19:10,636
One is if you've tried these tools and
models in the past, like six months ago or

357
00:19:10,636 --> 00:19:17,136
older and you haven't touched them again
since then, I highly recommend you at

358
00:19:17,136 --> 00:19:20,916
least try some of the latest SOTA models,

359
00:19:21,406 --> 00:19:22,946
and maintain.

360
00:19:22,946 --> 00:19:26,726
Another thing that we talked about
early on in the podcast episodes is

361
00:19:27,146 --> 00:19:29,336
also maintain a healthy skepticism.

362
00:19:29,386 --> 00:19:32,596
Everyone's use cases and what
they're doing is very different.

363
00:19:33,096 --> 00:19:35,136
they still are LLM models.

364
00:19:35,166 --> 00:19:38,406
They still act in the LM model way.

365
00:19:38,906 --> 00:19:45,156
and yes, the quality in terms of output
that we're seeing for whatever we're

366
00:19:45,156 --> 00:19:48,036
working on is increased dramatically.

367
00:19:48,536 --> 00:19:53,996
as engineers, as practitioners,
as operators of production systems

368
00:19:54,496 --> 00:19:59,196
maintain a healthy skepticism,
which means understanding how to

369
00:19:59,196 --> 00:20:05,046
use these tools, understand where,
their limits, their capabilities.

370
00:20:05,496 --> 00:20:11,496
I would highly recommend against just
dismissing these tools completely.

371
00:20:12,216 --> 00:20:17,736
Bret: the number of people that we respect
in the industry have in the last three

372
00:20:17,736 --> 00:20:20,406
months gone what I called AI red pilled.

373
00:20:20,436 --> 00:20:25,156
This is a Matrix, this is back to the
ultimate choice of Morpheus and Neo

374
00:20:25,156 --> 00:20:29,356
in the Matrix where he offered him two
pills, the red pill and the blue pill.

375
00:20:29,356 --> 00:20:33,796
The blue pill meant you'd wake up at
home, nothing would've changed you,

376
00:20:33,856 --> 00:20:36,706
you know, the reality hasn't been
altered, and you can just live your life

377
00:20:36,706 --> 00:20:37,996
pretending that nothing ever happened.

378
00:20:38,371 --> 00:20:40,681
That's the skeptics, right?

379
00:20:41,181 --> 00:20:45,491
The number of people that, but have,
are no longer taking the blue pill and

380
00:20:45,491 --> 00:20:49,841
are now on the red pill, which is, let's
see far, how far the rabbit hole goes.

381
00:20:50,261 --> 00:20:54,011
And let's try this thing everywhere.

382
00:20:54,011 --> 00:20:54,941
We can try it.

383
00:20:55,361 --> 00:20:57,821
You can still have skepticism
in trying a new tool, right?

384
00:20:57,825 --> 00:20:58,115
Nirmal Mehta: Yeah.

385
00:20:58,151 --> 00:20:59,711
Bret: skeptics when we
first started Docker.

386
00:20:59,711 --> 00:21:01,091
We were like, this is really that good.

387
00:21:01,221 --> 00:21:04,781
and then we saw people that would
try to use too much Docker, right?

388
00:21:04,781 --> 00:21:05,891
Like you use Docker everywhere.

389
00:21:05,891 --> 00:21:08,351
Like we thought it was gonna be the
replacement for brew on our local

390
00:21:08,351 --> 00:21:11,741
machines and we were gonna run every
local CLI tool in a Docker container.

391
00:21:11,771 --> 00:21:13,501
'cause it's convenient and it's isolated

392
00:21:13,540 --> 00:21:15,435
Nirmal Mehta: I mean, we
we're kind of going there

393
00:21:15,976 --> 00:21:17,026
Bret: We didn't, yeah, we

394
00:21:17,295 --> 00:21:18,195
Nirmal Mehta: with agent tools.

395
00:21:18,196 --> 00:21:18,526
Bret: far.

396
00:21:18,975 --> 00:21:19,875
Nirmal Mehta: Yeah, that's true.

397
00:21:20,336 --> 00:21:23,096
Bret: I thought containers were gonna
be like I thought Docker files were

398
00:21:23,096 --> 00:21:24,776
gonna be our build documentation.

399
00:21:24,776 --> 00:21:27,476
Like it was gonna be everything
for CI inside a Dockerfile.

400
00:21:27,740 --> 00:21:28,030
Nirmal Mehta: True.

401
00:21:28,226 --> 00:21:29,486
Bret: were stages and Docker files.

402
00:21:29,710 --> 00:21:30,000
Nirmal Mehta: Yeah.

403
00:21:30,026 --> 00:21:31,256
Bret: thought I was totally red pilled.

404
00:21:31,436 --> 00:21:33,116
Obviously it's always
somewhere in the middle,

405
00:21:33,655 --> 00:21:34,015
Nirmal Mehta: Yeah.

406
00:21:34,076 --> 00:21:37,466
Bret: I like that you wanna be skeptical,
but I am not issuing the same skepticism.

407
00:21:37,966 --> 00:21:38,506
Nirmal Mehta: I think,

408
00:21:38,536 --> 00:21:38,956
Bret: mean,

409
00:21:39,016 --> 00:21:42,126
Nirmal Mehta: think what I mean
by skepticism here is have a

410
00:21:42,216 --> 00:21:46,796
systems engineering approach
to these new tools, right?

411
00:21:46,846 --> 00:21:49,686
don't just read a blog post and.

412
00:21:50,376 --> 00:21:55,506
Consume the hype and say, oh,
it's like a hundred x productivity

413
00:21:55,506 --> 00:21:59,666
gain across the board, but
really start to use these tools.

414
00:21:59,666 --> 00:22:03,611
I think the broader message
here is, don't ignore it.

415
00:22:04,111 --> 00:22:06,631
Get good at using these tools

416
00:22:07,171 --> 00:22:07,561
Bret: Right.

417
00:22:07,591 --> 00:22:10,861
Nirmal Mehta: think about how these
tools can be used in your day to

418
00:22:10,861 --> 00:22:17,211
day, in the automation of the systems
and, orchestration and automation

419
00:22:17,211 --> 00:22:22,351
and documentation that you might be
responsible for in your job or your role.

420
00:22:22,801 --> 00:22:30,471
And also think about it as like an artist
would, a canvas, a brush paint where.

421
00:22:30,971 --> 00:22:35,831
You could use these tools to explore lots
of innovation and creativity as well.

422
00:22:36,281 --> 00:22:41,811
but skepticism in the sense that
use it, see how it's being used.

423
00:22:42,291 --> 00:22:45,681
Don't just take like LeVar Burton
said, don't take just our word for it.

424
00:22:46,701 --> 00:22:49,576
Bret: I don't think anybody, we're not
even using the words vibe coding anymore.

425
00:22:49,576 --> 00:22:52,986
those words are a year old and we
now understand the nuance and it's

426
00:22:52,986 --> 00:22:56,866
sort of a meme at this point, but no
one's saying Vibe code your DevOps.

427
00:22:56,926 --> 00:22:57,286
Right.

428
00:22:57,591 --> 00:22:57,951
Nirmal Mehta: Correct.

429
00:22:57,951 --> 00:22:58,551
Yes.

430
00:22:58,836 --> 00:23:00,786
Bret: That is why I'm making courses.

431
00:23:00,786 --> 00:23:06,226
That's why I've been working for a year
now to build this DevOps world in my

432
00:23:06,241 --> 00:23:10,861
community and in my content, because
this is gonna take years and years.

433
00:23:10,861 --> 00:23:13,651
It's, this is every one of these big
waves like Docker, like Kubernetes.

434
00:23:13,786 --> 00:23:14,116
Nirmal Mehta: Yeah.

435
00:23:14,221 --> 00:23:17,311
Bret: It's at least 10 years before
ma. The majority of companies have

436
00:23:17,311 --> 00:23:20,731
understood it and adopted it, but I
think there's three main areas for people

437
00:23:20,731 --> 00:23:22,111
to focus on in the world of DevOps.

438
00:23:22,441 --> 00:23:25,411
The first area that everybody should
be doing right now, regardless of

439
00:23:25,411 --> 00:23:29,461
your skillset, regardless of whether
you're allowed to at work, just

440
00:23:29,461 --> 00:23:33,601
do it at home, is to try, even if
you can only get the free models.

441
00:23:33,811 --> 00:23:40,891
GitHub free account provides free
copilot, which provides free LLMs for

442
00:23:40,891 --> 00:23:43,411
you to use with all of the copilot tools.

443
00:23:43,411 --> 00:23:44,281
That's web.

444
00:23:44,781 --> 00:23:45,711
VS Code.

445
00:23:45,711 --> 00:23:48,981
That's command line TUIs, like
you can use all that stuff.

446
00:23:48,981 --> 00:23:52,321
There's free Claude, there's
free, GPT, there's free Gemini.

447
00:23:52,651 --> 00:23:54,061
You can use these today.

448
00:23:54,361 --> 00:23:58,711
But you, if you can just get a
$20 plan with any of those things,

449
00:23:58,711 --> 00:24:01,621
whether it's GitHub copilot, whether
it's Claude, whether it's Codex.

450
00:24:02,331 --> 00:24:06,621
just spend $20, even if you just spend
it for a month to do some learning, if

451
00:24:06,621 --> 00:24:08,061
you can't get your company to pay for it.

452
00:24:08,891 --> 00:24:12,421
think anyone that I would sit down
with and help them through the

453
00:24:12,421 --> 00:24:17,351
basics of getting started would
see really fast benefits to their

454
00:24:17,351 --> 00:24:19,121
personal workflow for DevOps.

455
00:24:19,201 --> 00:24:24,371
'cause when we think of DevOps, we think
I'm largely writing YAML or shell scripts,

456
00:24:24,491 --> 00:24:27,161
or maybe if I'm fancy I'm doing Golang or

457
00:24:27,261 --> 00:24:27,591
Nirmal Mehta: Right.

458
00:24:27,697 --> 00:24:29,437
Bret: I'm doing some
automation for my GitHub.

459
00:24:29,647 --> 00:24:31,297
I'm probably spending
a lot of time in Git.

460
00:24:31,477 --> 00:24:35,227
I need to do git commits and
GitHub things, or GitLab things.

461
00:24:35,632 --> 00:24:40,282
And almost all of that can immediately
be benefited just by having, Claude

462
00:24:40,302 --> 00:24:45,217
Code installed on your machine or the
chat GPT desktop is a fantastic tool,

463
00:24:45,607 --> 00:24:49,597
and you don't have to necessarily be
a developer to do all of this to see

464
00:24:49,597 --> 00:24:52,097
the benefits, but it's a study buddy.

465
00:24:52,097 --> 00:24:53,117
It's a work buddy.

466
00:24:53,117 --> 00:24:54,407
It's there and it's reliable.

467
00:24:54,407 --> 00:24:58,847
The better, the higher end and more
expensive models you can find more

468
00:24:58,847 --> 00:25:00,437
likely will give you better results,

469
00:25:00,487 --> 00:25:00,577
Nirmal Mehta: Yes.

470
00:25:00,727 --> 00:25:05,147
Bret: seek on open, open weighted
models or closed source models.

471
00:25:05,387 --> 00:25:08,687
But I can tell you that for the last
three months, everyone I know and

472
00:25:08,777 --> 00:25:11,477
across dozens and dozens of people,
big or small, they're all talking

473
00:25:11,477 --> 00:25:13,277
about Opus as the model to use.

474
00:25:13,577 --> 00:25:17,057
Some of them still use
Gemini and still use GPT 5.3.

475
00:25:17,507 --> 00:25:20,147
Those still work, but they're
all in love with Opus.

476
00:25:20,197 --> 00:25:23,917
That's the first wave is like
get used to local tooling.

477
00:25:24,337 --> 00:25:24,757
Nirmal Mehta: Mm-hmm.

478
00:25:24,847 --> 00:25:30,967
Bret: push yourself to try using
AI as a buddy to help you anytime

479
00:25:30,967 --> 00:25:32,467
you need to create some content.

480
00:25:32,467 --> 00:25:34,267
Content meaning YAML or whatever.

481
00:25:34,667 --> 00:25:39,802
next phase is getting AI to
help you automate your CI.

482
00:25:40,952 --> 00:25:41,552
Nirmal Mehta: Yes.

483
00:25:41,742 --> 00:25:45,212
Bret: we are still very early days,
but in the fall of last year GitHub

484
00:25:45,212 --> 00:25:49,592
announced a bunch of different
things, especially at GitHub Universe.

485
00:25:49,692 --> 00:25:52,062
I'm sure Amazon also launched
months of different things on

486
00:25:52,062 --> 00:25:54,702
Bedrock, but I unfortunately
don't pay attention to everyone.

487
00:25:55,182 --> 00:25:58,292
But I can tell you that on the GitHub
side, and maybe you can, compare this to

488
00:25:58,292 --> 00:26:02,992
what's happening on Bedrock, and other
functionality AWS, but like at GitHub.

489
00:26:03,492 --> 00:26:06,572
They launched, not only keeping
track with all the models.

490
00:26:06,602 --> 00:26:10,142
'cause if you didn't know GitHub copilot
models is kind of like open router

491
00:26:10,142 --> 00:26:17,112
where you can use everything from Opus
to Gemini to DeepSeek, to GPT, to Qwen,

492
00:26:17,492 --> 00:26:22,652
to, tiny Little Pie and tiny little,
there's dozens of models that you use

493
00:26:22,652 --> 00:26:26,702
all through GitHub, and you only need
a GitHub account to get started there.

494
00:26:26,702 --> 00:26:27,752
You don't even need to spend money.

495
00:26:28,082 --> 00:26:32,672
But if you pay for the GitHub copilot,
I think it starts at like $17 a month,

496
00:26:33,152 --> 00:26:36,692
you can get access to all the bigger
models, and I use that every month.

497
00:26:36,722 --> 00:26:40,462
And I can easily go days or weeks in
the month on just a $20 plan without

498
00:26:40,462 --> 00:26:45,712
hitting limits, and you can start to
experiment something that teased us with

499
00:26:45,712 --> 00:26:47,212
in the fall, but officially launched.

500
00:26:47,712 --> 00:26:51,302
I think actually last month that
they're calling Agentic Workflows.

501
00:26:51,562 --> 00:26:54,787
Luckily, Nirmal and I were on point
with the name of our podcast, Agentic

502
00:26:54,792 --> 00:26:59,382
DevOps Podcast because they're calling
this functionality Agentic Workflows.

503
00:26:59,812 --> 00:27:03,632
And it's a little bit different
than just throwing an LLM prompt

504
00:27:03,722 --> 00:27:05,582
in your workflows or pipelines.

505
00:27:05,792 --> 00:27:09,272
It's using AI to create workflow.

506
00:27:09,602 --> 00:27:12,452
So you create a markdown
file with some front matter.

507
00:27:12,752 --> 00:27:16,022
You give it the front matter as
sort of like the deterministic

508
00:27:16,022 --> 00:27:19,472
stuff, and then you write a prompt
of what you would like it to do.

509
00:27:19,952 --> 00:27:23,882
And then the AI through
GitHub Copilot will create the

510
00:27:23,932 --> 00:27:25,912
workflow for you and run it.

511
00:27:26,332 --> 00:27:32,032
This is one of their ideas for how
we might really complex or even

512
00:27:32,532 --> 00:27:36,522
workflows that have an undetermined,
non-deterministic nature to them.

513
00:27:36,852 --> 00:27:39,102
A good example might be triaging.

514
00:27:39,597 --> 00:27:42,327
Why your GitHub Actions failed.

515
00:27:42,627 --> 00:27:46,257
So it has to go read the logs,
find the failure, then go look

516
00:27:46,257 --> 00:27:47,277
up what the failure might be.

517
00:27:47,277 --> 00:27:51,897
Or maybe it just knows built into the
model, and then it might suggest on the

518
00:27:51,897 --> 00:27:57,117
existing PR it might open a new PR based
on what you want it to do, and it will

519
00:27:57,117 --> 00:28:02,537
start to help automate fixing problems
essentially that before if you're trying

520
00:28:02,537 --> 00:28:06,707
to just create a very deterministic way
without a Code Rabbit or a GitHub Copilot,

521
00:28:07,097 --> 00:28:09,197
those might be very hard to do, right?

522
00:28:09,197 --> 00:28:12,947
There would've to be some grips involved,
some set and ox, like you would definitely

523
00:28:12,947 --> 00:28:15,707
have to have a very long and thorough app.

524
00:28:16,107 --> 00:28:17,967
it would, that's what it probably
would turn into an action

525
00:28:17,967 --> 00:28:19,137
that's almost like its own app.

526
00:28:19,137 --> 00:28:22,227
And we've had those, we've seen those
before, but now you can start to

527
00:28:22,227 --> 00:28:23,367
create these custom ones yourself.

528
00:28:23,367 --> 00:28:24,327
So that's stage two.

529
00:28:24,657 --> 00:28:30,417
I think stage three in the last stage
is where have an always running AI

530
00:28:30,867 --> 00:28:32,967
that's able to respond to events.

531
00:28:33,987 --> 00:28:34,977
an operator would.

532
00:28:35,852 --> 00:28:36,312
Nirmal Mehta: Right.

533
00:28:36,357 --> 00:28:37,917
Bret: the most mature area.

534
00:28:38,217 --> 00:28:41,097
So in terms of the maturity model
that I'm building out for my

535
00:28:41,097 --> 00:28:44,257
courses and for my community, it's
starting with the local tools.

536
00:28:44,257 --> 00:28:47,377
You and I saw that like the individual
developers are very productive on

537
00:28:47,377 --> 00:28:49,417
their local machine with these tools.

538
00:28:49,917 --> 00:28:53,027
as you get really comfortable with that
stuff locally and you figure out what

539
00:28:53,027 --> 00:28:56,777
YAML and TOML does this thing write
well, and where do I need to learn how

540
00:28:56,777 --> 00:28:58,607
to hold its hand a little bit more?

541
00:28:59,027 --> 00:28:59,987
You're gonna learn all that.

542
00:29:00,077 --> 00:29:03,887
And then as you get better at using
what we call now the agent harness,

543
00:29:04,037 --> 00:29:07,897
which is just really the UI that
you're using with your agents, but

544
00:29:07,897 --> 00:29:09,667
I'm using in the open code harness.

545
00:29:09,667 --> 00:29:13,257
Others, might be using it like Claude
Code or Codex or a different harness,

546
00:29:13,757 --> 00:29:13,977
Nirmal Mehta: Yep.

547
00:29:14,073 --> 00:29:15,073
Shout out to Kiro.

548
00:29:15,573 --> 00:29:16,473
Bret: shout out to Kiro.

549
00:29:16,533 --> 00:29:16,923
Yeah.

550
00:29:17,073 --> 00:29:19,563
another harness, like there's a
lot of harnesses and everyone's got

551
00:29:19,563 --> 00:29:23,223
their favorite and they all sort of
are merging around the same ideas

552
00:29:23,583 --> 00:29:25,663
of letting the agent continue to go.

553
00:29:26,053 --> 00:29:29,493
Longer and longer, like we're
getting better at the AI's, running

554
00:29:29,493 --> 00:29:32,463
for longer periods of time, doing
multiple steps without hallucinating.

555
00:29:32,463 --> 00:29:33,273
That's getting better.

556
00:29:33,797 --> 00:29:34,017
Nirmal Mehta: Yes.

557
00:29:34,143 --> 00:29:37,893
Bret: time, we've invented this idea
of skills that really allow you to

558
00:29:37,893 --> 00:29:42,633
create a SOP, a standard operating
procedure for each type of work you do.

559
00:29:42,783 --> 00:29:46,293
And I just talked in the livestream
today about, as you get better with your

560
00:29:46,293 --> 00:29:51,213
local tooling, you have your own work
that you typically do over and over.

561
00:29:51,453 --> 00:29:55,813
Most of us don't create YAML and
TOML, out of thin air, right?

562
00:29:55,813 --> 00:29:58,993
If we're doing DevOps, we're
usually copy and pasting a lot.

563
00:29:59,233 --> 00:30:03,973
We have standards in the team repo that
we set up with standards in them that we

564
00:30:03,973 --> 00:30:05,623
use our official templates from, right?

565
00:30:05,623 --> 00:30:09,193
We've all got different processes
for how we know that this particular

566
00:30:09,193 --> 00:30:12,793
cloud formation has to be, that
these adhere to these standards, pass

567
00:30:12,793 --> 00:30:14,623
these lins, like you've got all that.

568
00:30:14,773 --> 00:30:18,763
It's probably in your confluence or
in your GitHub repo itself, you can

569
00:30:18,763 --> 00:30:20,773
very quickly turn that into a skill.

570
00:30:21,273 --> 00:30:22,733
That skill is just markdown.

571
00:30:22,893 --> 00:30:27,233
And it's like you treat it like
you're teaching a junior engineer to

572
00:30:27,233 --> 00:30:30,803
automate or to do some part of work
that you don't wanna do yourself.

573
00:30:30,953 --> 00:30:36,613
And so I have had great success in
the last month the standard operating

574
00:30:36,613 --> 00:30:40,303
procedures that I have for how I
create a reusable workflow in GitHub

575
00:30:40,303 --> 00:30:43,993
Actions or how I would go about
creating some sort of terraform.

576
00:30:44,353 --> 00:30:45,673
I know what I care about.

577
00:30:45,673 --> 00:30:47,113
I'm particular about certain things.

578
00:30:47,113 --> 00:30:50,113
I have certain standards
those maybe are in my head.

579
00:30:50,143 --> 00:30:51,403
We call that tribal knowledge.

580
00:30:51,433 --> 00:30:54,493
I'm a tribe of one, so I have
tribal knowledge to myself.

581
00:30:54,973 --> 00:30:57,853
But you're gonna put that in a
markdown and I just call it a SOP now.

582
00:30:57,883 --> 00:30:59,143
Like the skills are SOPs.

583
00:30:59,263 --> 00:31:03,313
I had SOPs before in my notion
with my team that we are automating

584
00:31:03,313 --> 00:31:04,993
through skills and in AI today.

585
00:31:05,383 --> 00:31:06,373
And I think that.

586
00:31:07,258 --> 00:31:09,778
be a bigger portion of my
courses that I anticipated.

587
00:31:10,018 --> 00:31:13,918
' Cause once you can get good to that,
you're onto something because what

588
00:31:13,918 --> 00:31:20,488
you've done is you've documented
a non-deterministic workflow an AI

589
00:31:20,488 --> 00:31:24,298
that runs locally on your machine,
you're only one step away from

590
00:31:24,298 --> 00:31:25,588
that running in a GitHub action.

591
00:31:26,088 --> 00:31:32,268
And that's what GitHub calls Continuous
AI, which is AI running in your

592
00:31:32,268 --> 00:31:37,638
automation, doing various things in a
non-deterministic way that you couldn't

593
00:31:37,638 --> 00:31:42,498
otherwise do with the previous generation
of what we're calling deterministic

594
00:31:42,498 --> 00:31:46,788
workflows, which is programmatic
if then statements, L statements.

595
00:31:46,998 --> 00:31:47,238
Like

596
00:31:47,432 --> 00:31:47,912
Nirmal Mehta: based.

597
00:31:47,978 --> 00:31:48,308
Bret: Yeah.

598
00:31:49,448 --> 00:31:50,708
So I need to have a diagram for this.

599
00:31:50,708 --> 00:31:51,278
I don't.

600
00:31:51,428 --> 00:31:53,218
but this is, I think
where we're all going.

601
00:31:53,218 --> 00:31:55,678
And I think that if you're not at
the beginning, at least of this

602
00:31:55,858 --> 00:31:57,838
series of steps and maturity model.

603
00:31:58,318 --> 00:32:00,058
I think now is the time, baby.

604
00:32:00,118 --> 00:32:03,388
Like we're a year into gentech
DevOps and today is the day.

605
00:32:03,777 --> 00:32:04,827
Nirmal Mehta: Yes, totally.

606
00:32:05,247 --> 00:32:08,487
And I think it's somewhat optional to do.

607
00:32:08,847 --> 00:32:13,712
I don't think it's gonna be optional
for, for our jobs in the near future.

608
00:32:14,212 --> 00:32:17,152
And I think it's not gonna be optional.

609
00:32:17,392 --> 00:32:24,422
I think we're gonna need these tools
to support the proliferation of

610
00:32:24,752 --> 00:32:33,362
CI/CD workflows that are gonna be
needed to deploy, And maintain all

611
00:32:33,362 --> 00:32:36,742
of these new applications that are
being built with these Agentic tools.

612
00:32:37,242 --> 00:32:39,457
I want to kind of
highlight something here.

613
00:32:39,957 --> 00:32:46,977
What we saw at these events in the
last month were developers accelerating

614
00:32:47,477 --> 00:32:51,527
the, their application development
lifecycle, the inner loop, and then

615
00:32:51,527 --> 00:32:55,397
they weren't making decisions on
how to deploy these applications.

616
00:32:55,897 --> 00:32:57,817
Claude Code was doing that.

617
00:32:57,907 --> 00:33:01,997
the agents were trying to figure
out, they let their agents

618
00:33:01,997 --> 00:33:03,497
figure out how to deploy this.

619
00:33:03,997 --> 00:33:08,947
And so those SOPs, the concept of
SOPs and skills are gonna be super

620
00:33:08,947 --> 00:33:13,767
important because, as operators, we're
gonna be responsible for creating

621
00:33:13,767 --> 00:33:15,747
the platforms and the guardrails.

622
00:33:16,247 --> 00:33:21,797
That are the landing zones for these
applications and also the system that

623
00:33:22,297 --> 00:33:28,827
helps operate all of these new pipelines
and workflows that have to be built to

624
00:33:29,327 --> 00:33:33,327
deal with a, you know, what I call the
tsunami of applications that are coming

625
00:33:33,687 --> 00:33:37,107
to you, to all of us as operators and

626
00:33:37,273 --> 00:33:39,803
Bret: gonna be creating like this.

627
00:33:39,853 --> 00:33:41,263
Nirmal Mehta: I think
we're gonna be create,

628
00:33:41,558 --> 00:33:41,918
Bret: Yeah.

629
00:33:41,968 --> 00:33:43,318
Nirmal Mehta: Yeah, I think that's true.

630
00:33:43,318 --> 00:33:45,388
I think the word DevOps is gonna be.

631
00:33:45,888 --> 00:33:47,268
Even more DevOpsy.

632
00:33:48,778 --> 00:33:53,168
like we're really gonna be developer
operators, with respect to building

633
00:33:53,168 --> 00:33:58,548
applications that are doing all of
those operational tasks, for us,

634
00:33:58,878 --> 00:34:02,718
which then leads to the next thing,
and I know this is maybe a little bit

635
00:34:02,718 --> 00:34:07,308
skipping in our outline today, but that
means that we're gonna have a lot of

636
00:34:07,308 --> 00:34:12,938
agents running and doing things on our
behalf, running kind of in parallel.

637
00:34:13,438 --> 00:34:19,398
And so, if you've been kind of just seeing
what's going on in the last couple months,

638
00:34:19,898 --> 00:34:26,708
I think the term agent orchestration is
a term to look out for and I think as

639
00:34:26,708 --> 00:34:34,998
operators, as platform engineers, as SREs,
we're gonna have to be able to answer how.

640
00:34:35,418 --> 00:34:41,318
And where, and how all these
agents are gonna run in a secure

641
00:34:41,818 --> 00:34:46,798
environment, with all of the, well
architected principles around it.

642
00:34:47,298 --> 00:34:53,128
And, that's another area of tooling
that is just coming out right,

643
00:34:53,128 --> 00:34:57,368
is, and I think, one can make the
case that something like OpenClaw

644
00:34:58,018 --> 00:35:00,888
is like agent orchestration layer.

645
00:35:01,388 --> 00:35:07,168
I think Kubernetes is very well placed
to be an agent orchestration layer.

646
00:35:07,668 --> 00:35:13,638
And, I think all of our cloud native
Kubernetes knowledge and containerization

647
00:35:13,638 --> 00:35:15,698
knowledge, is not obsolete.

648
00:35:16,628 --> 00:35:21,868
It's gonna be very well placed to be
the harness and the orchestration for

649
00:35:21,868 --> 00:35:26,038
all of these agents as they're running,
doing all of these tasks, both on the

650
00:35:26,038 --> 00:35:32,298
developer side and on the operations
side, and also your business agents.

651
00:35:32,298 --> 00:35:33,468
We haven't even talked about that.

652
00:35:33,468 --> 00:35:38,593
But I think another thing that we
saw is non-technical users starting

653
00:35:38,593 --> 00:35:43,053
to use these tools to build business
applications that aren't devs.

654
00:35:43,413 --> 00:35:43,773
Right?

655
00:35:43,773 --> 00:35:46,113
They don't know anything about CI/CD.

656
00:35:46,353 --> 00:35:48,213
They don't know anything
about infrastructure.

657
00:35:48,423 --> 00:35:52,463
They don't know anything about code
and they're gonna start building apps.

658
00:35:52,883 --> 00:36:00,493
And, if you thought Shadow IT was a big
problem back when cloud was new, oh boy.

659
00:36:00,673 --> 00:36:05,553
Shadow IT with gen AI is like,
that's a whole brave new world

660
00:36:05,613 --> 00:36:06,873
that we're entering into that.

661
00:36:07,236 --> 00:36:10,523
Bret: which really just goes to the fact
that we're gonna need more automation.

662
00:36:10,609 --> 00:36:13,129
There's a great graphic from
our friends at Geo Coio.

663
00:36:13,629 --> 00:36:18,219
this is the dev loop, or I guess it's
really the inner plus outer loop.

664
00:36:18,319 --> 00:36:20,059
This one, you saw this at the retreat.

665
00:36:20,489 --> 00:36:22,859
so this is by Michelle
Hanson, of Geo Coio.

666
00:36:22,889 --> 00:36:24,219
And it's a very simple diagram.

667
00:36:24,219 --> 00:36:27,189
in the first one, it's
traditional projects before AI.

668
00:36:27,189 --> 00:36:28,624
And it's basically a timeline.

669
00:36:28,984 --> 00:36:34,844
And one fifth of this timeline is
scoping before development, and then

670
00:36:34,844 --> 00:36:36,674
three fifths of the timeline is coding.

671
00:36:36,704 --> 00:36:38,114
And then one fifth at the end is qa.

672
00:36:38,614 --> 00:36:43,214
And what they're describing as
with AI, maybe you can get it

673
00:36:43,214 --> 00:36:45,734
done in 60% of the time as before.

674
00:36:45,794 --> 00:36:50,404
So, so you're saving 40% of your
time overall, but the coding

675
00:36:50,404 --> 00:36:53,434
is highly compressed to less
than one fifth of the time.

676
00:36:53,434 --> 00:36:54,424
This is just approximation.

677
00:36:54,474 --> 00:36:55,314
these aren't exactly like

678
00:36:55,418 --> 00:36:55,768
Nirmal Mehta: Right,

679
00:36:56,934 --> 00:36:59,904
Bret: metrics or studies or anything
like that, but the scoping is

680
00:36:59,904 --> 00:37:02,604
doubled and the QA is almost doubled.

681
00:37:02,784 --> 00:37:05,544
And so the two sides, the
planning at the beginning.

682
00:37:06,044 --> 00:37:08,264
Which we now have plan
modes in our agents,

683
00:37:08,314 --> 00:37:08,664
Nirmal Mehta: right.

684
00:37:08,739 --> 00:37:11,199
Bret: now have these concepts
of plan agents and build agents.

685
00:37:11,649 --> 00:37:12,489
Nirmal Mehta: And specs

686
00:37:12,529 --> 00:37:13,669
Bret: helping you create the spec.

687
00:37:13,669 --> 00:37:13,909
Yeah.

688
00:37:13,959 --> 00:37:14,249
Nirmal Mehta: yeah.

689
00:37:14,514 --> 00:37:15,744
Bret: plan with all the tasking.

690
00:37:16,244 --> 00:37:18,704
And then at the end, you're
gonna need a longer qa.

691
00:37:19,634 --> 00:37:22,454
QA to me means I'm gonna
need more automation.

692
00:37:22,454 --> 00:37:27,434
I'm gonna need code ql, I'm gonna need
way more AI and automation running

693
00:37:27,434 --> 00:37:31,664
against every commit because I'm gonna
suddenly get a bunch of projects thrown

694
00:37:31,664 --> 00:37:36,644
at me that weren't as well, vetted
by humans as they were in the past.

695
00:37:36,644 --> 00:37:40,794
Not if AI is making lines of code,
that means that humans may have at

696
00:37:40,794 --> 00:37:45,564
best read the code at best, but they
certainly didn't write every line

697
00:37:45,564 --> 00:37:47,064
of code, so they didn't rewrite it.

698
00:37:47,064 --> 00:37:48,864
They didn't rethink
about every single line.

699
00:37:48,864 --> 00:37:52,224
They didn't toil over every
line, which maybe in the long

700
00:37:52,224 --> 00:37:53,484
run is actually a good thing.

701
00:37:53,844 --> 00:37:58,254
But what that really means for me as a CI
person is that I'm probably gonna be more

702
00:37:58,254 --> 00:38:00,114
responsible, and I've had already two.

703
00:38:00,114 --> 00:38:02,394
Now, in the last month,
two different teams tell me

704
00:38:02,644 --> 00:38:06,514
They're having to push back as
DevOps engineers to the developers

705
00:38:06,814 --> 00:38:10,924
because the developers are shipping
more vibe coded product that they're

706
00:38:10,924 --> 00:38:15,244
having more difficulty because
their testing isn't catching the

707
00:38:15,244 --> 00:38:17,344
problems that the AIs are creating.

708
00:38:17,824 --> 00:38:20,174
And so these apps are getting
shipped to production.

709
00:38:20,354 --> 00:38:24,224
They fail in production, and then the
developers go, it works on my machine.

710
00:38:25,034 --> 00:38:28,154
and this is the age old problem,
but it's happening again.

711
00:38:28,859 --> 00:38:31,289
Because we're shipping more
lines of code in a commit.

712
00:38:31,289 --> 00:38:35,399
We're shipping more lines in a pull
request, and not as many eyeballs

713
00:38:35,399 --> 00:38:37,109
have studied every single line.

714
00:38:37,359 --> 00:38:41,339
the leading edge teams in this field of AI
right now, the ones that are pushing the

715
00:38:41,339 --> 00:38:44,399
limit, aren't looking at the code anymore.

716
00:38:44,789 --> 00:38:46,529
Like they're shipping production code.

717
00:38:46,529 --> 00:38:50,149
That's only, that's written by
AI, and then reviewed by AI.

718
00:38:50,449 --> 00:38:53,749
So, and you can have three
different AIs review it, right?

719
00:38:53,759 --> 00:38:54,479
you can have

720
00:38:55,054 --> 00:38:55,344
Nirmal Mehta: Yeah.

721
00:38:55,399 --> 00:38:57,949
Bret: my friend Aaron Francis,
created a new tool called Counselors,

722
00:38:58,009 --> 00:38:59,119
which does that exact thing.

723
00:38:59,369 --> 00:39:00,569
you give it a problem.

724
00:39:00,769 --> 00:39:03,739
it has three or more AIs it will
go and ask, and then it will

725
00:39:03,739 --> 00:39:06,649
summarize the differences and
what they agreed on and what they

726
00:39:06,649 --> 00:39:09,429
disagreed on in a summary for you.

727
00:39:09,429 --> 00:39:12,479
And It's literally called the counselors,
And we don't have that specifically for

728
00:39:12,479 --> 00:39:14,309
DevOps yet, but I can see that coming.

729
00:39:14,744 --> 00:39:20,294
Where we're going to have a council
review, the PR, because maybe this

730
00:39:20,294 --> 00:39:26,204
month opus is a little bit more loosey
goosey, like Opus likes the vibe coding

731
00:39:26,204 --> 00:39:27,584
this month in the latest version.

732
00:39:27,644 --> 00:39:30,494
And maybe Gemini is more
picky by default, right?

733
00:39:30,494 --> 00:39:33,824
this reminds me of the nineties and
two thousands with antivirus software

734
00:39:33,824 --> 00:39:37,774
where we had, the best antivirus teams
that, like I ran an antivirus team

735
00:39:37,774 --> 00:39:39,954
for a, enterprise of 7,000 mailboxes.

736
00:39:40,344 --> 00:39:44,574
we never could trust just McAfee
or, you know, or just one scanner.

737
00:39:44,574 --> 00:39:46,614
So we had the scanner of scanners.

738
00:39:46,614 --> 00:39:50,544
We had a single product
that had like eight or nine.

739
00:39:50,724 --> 00:39:54,774
It had Kaspersky it had all these
different scanners you had to be

740
00:39:54,774 --> 00:39:58,794
able to get an email through all
of them it would get to the inbox

741
00:39:59,129 --> 00:39:59,489
Nirmal Mehta: I think so.

742
00:39:59,549 --> 00:39:59,769
Yep.

743
00:39:59,814 --> 00:40:02,864
Bret: we had very opinionated,
and the same is true the CVE

744
00:40:02,864 --> 00:40:04,094
scanners today, like some teams.

745
00:40:04,409 --> 00:40:06,029
Scan with two different CV scanners.

746
00:40:06,419 --> 00:40:07,979
Some teams, have their favorite.

747
00:40:08,429 --> 00:40:10,979
But anyway, this image that I'm
describing to the audio audience.

748
00:40:11,479 --> 00:40:13,539
the second thing, do you
have any thoughts about

749
00:40:13,614 --> 00:40:14,274
Nirmal Mehta: So

750
00:40:14,409 --> 00:40:14,619
Bret: tell you

751
00:40:14,664 --> 00:40:18,294
Nirmal Mehta: yeah, I think I saw this
quote we ship faster but break more.

752
00:40:18,744 --> 00:40:21,384
So that's like the era that
we're in right now, right?

753
00:40:21,414 --> 00:40:26,034
Like we're shipping faster,
but everything's breaking

754
00:40:26,034 --> 00:40:27,414
more than it used to.

755
00:40:28,179 --> 00:40:28,419
Bret: yeah.

756
00:40:28,419 --> 00:40:30,859
We've got a, and that
leans on the DevOps engine,

757
00:40:31,009 --> 00:40:31,519
Nirmal Mehta: Correct?

758
00:40:31,729 --> 00:40:32,419
Yes.

759
00:40:32,435 --> 00:40:33,005
Bret: automation.

760
00:40:33,005 --> 00:40:36,275
That's, and there's so many talks
over the last year that I could quote,

761
00:40:36,325 --> 00:40:36,655
Nirmal Mehta: And.

762
00:40:37,060 --> 00:40:40,180
Bret: Han's doctor at the Gradle
Summit, I think the CEO of

763
00:40:40,180 --> 00:40:42,040
Gradle, he talked about this.

764
00:40:42,190 --> 00:40:46,380
He's basically saying a wave of
problems coming to DevOps and

765
00:40:46,380 --> 00:40:48,140
SRE and Ops from all of this.

766
00:40:48,140 --> 00:40:50,030
And we better prepare because it's come.

767
00:40:50,110 --> 00:40:50,570
Nirmal Mehta: Yes,

768
00:40:50,571 --> 00:40:51,861
Bret: to your team, it's coming soon.

769
00:40:52,250 --> 00:40:55,060
Nirmal Mehta: And the other thing
here, just going like highlighting

770
00:40:55,090 --> 00:41:03,200
that diagram, scope and Q&A scope
is documentation, it's context.

771
00:41:03,700 --> 00:41:09,090
And, we haven't had this conversation
yet, but I think in the new architecture

772
00:41:09,590 --> 00:41:13,450
there's going to be a context layer.

773
00:41:13,950 --> 00:41:18,320
there's gonna be agent orchestration,
but the's a context layer there

774
00:41:18,820 --> 00:41:23,590
As operators are gonna be responsible
for main, like creating, run,

775
00:41:23,650 --> 00:41:28,770
running, maintaining, because when
no one, no human's looking at the

776
00:41:28,770 --> 00:41:33,490
code, the code is still important
because it is the application, but

777
00:41:33,490 --> 00:41:39,890
it's, importance is goes down a
little bit compared to the specs, the

778
00:41:39,890 --> 00:41:45,500
requirements, the context, your business
documentation and your Jira tickets.

779
00:41:45,500 --> 00:41:55,310
And like the context and the intent of
your developers or your business users

780
00:41:55,810 --> 00:41:58,630
is more important than the line of code.

781
00:41:58,976 --> 00:41:59,396
Bret: Mm-hmm.

782
00:41:59,440 --> 00:42:04,870
Nirmal Mehta: And I think that was another
theme that came out of our conversations

783
00:42:04,870 --> 00:42:11,520
the last month is that, Having systems
that maintain and version control that

784
00:42:11,520 --> 00:42:17,150
context are going to be part of our
architectures moving forward as well.

785
00:42:17,650 --> 00:42:17,940
Bret: Yeah.

786
00:42:18,055 --> 00:42:19,515
Nirmal Mehta: And, what that looks like.

787
00:42:19,755 --> 00:42:20,385
Not sure.

788
00:42:20,385 --> 00:42:24,255
I think it's gonna be an amalgamation
of a lot of things, but this also means

789
00:42:24,255 --> 00:42:29,985
that with skills and these operational
agents, these i AI ops agents on the

790
00:42:29,985 --> 00:42:32,705
SRE side, that's still true there.

791
00:42:33,205 --> 00:42:34,345
It's the context.

792
00:42:34,375 --> 00:42:37,495
You know, it's your runbooks, it's
your SOPs, it's your troubleshooting

793
00:42:37,765 --> 00:42:41,510
steps, which your in incident response,
which are your product documentation.

794
00:42:42,010 --> 00:42:45,960
If you're sitting there listening to
this and, you've been fighting your

795
00:42:45,960 --> 00:42:51,510
management about having, story points,
in that sprint to update your docs

796
00:42:51,510 --> 00:42:56,330
and like never getting enough time,
This is your hour, this is your time.

797
00:42:56,420 --> 00:43:01,980
Because organizations that will
be successful with these tools are

798
00:43:01,980 --> 00:43:07,460
the ones that have automation in
place or building toward that have

799
00:43:07,460 --> 00:43:12,430
good documentation practices, have
good, operational practices and

800
00:43:12,550 --> 00:43:17,160
organizations that have been deferring
that for another time in the future.

801
00:43:17,160 --> 00:43:20,070
That tech and like have not
paid that technical debt,

802
00:43:20,570 --> 00:43:22,040
that you're gonna be behind.

803
00:43:22,160 --> 00:43:23,000
Just straight up.

804
00:43:23,000 --> 00:43:25,700
You're gonna be behind, you're
not gonna be able to leverage

805
00:43:26,000 --> 00:43:29,080
these tools because that's the
main interaction with these tools.

806
00:43:29,080 --> 00:43:29,440
Now.

807
00:43:29,500 --> 00:43:31,930
it's not like programming a line of code.

808
00:43:31,980 --> 00:43:33,690
it's giving them context and intent.

809
00:43:34,190 --> 00:43:36,980
Bret: you know, when we talk about these
things like the skills and the agent

810
00:43:36,980 --> 00:43:41,570
files and the rules and the commands, I
was talking on the stream today about the

811
00:43:41,570 --> 00:43:42,890
difference between skills and commands.

812
00:43:42,890 --> 00:43:45,980
And I think skills, I mean,
commands are probably mostly dead.

813
00:43:45,980 --> 00:43:49,460
I think skills are gonna win and like
that, I think that model is people are

814
00:43:49,460 --> 00:43:52,520
really understand liking that model
and it's working well for people.

815
00:43:53,020 --> 00:43:58,210
But we're essentially just recreating the
software lifecycle we had documentation

816
00:43:58,210 --> 00:44:03,530
in wikis and doc systems, and we had
standard operating procedures in another

817
00:44:03,530 --> 00:44:06,680
system where they were very meticulous
and they had lots of screenshots and like,

818
00:44:06,920 --> 00:44:08,600
this is how the humans get things done.

819
00:44:08,600 --> 00:44:09,440
These are our standards.

820
00:44:09,770 --> 00:44:13,880
And we would have, you know, mermaid
graphics or various other tools, you

821
00:44:13,880 --> 00:44:15,590
know, Visio or whatever you might have.

822
00:44:15,965 --> 00:44:19,925
For diagramming workflows or diagramming
system designs or network designs.

823
00:44:20,225 --> 00:44:23,885
Like these are all things
we would do in planning.

824
00:44:23,885 --> 00:44:26,615
And then, then after we'd ship
something, we'd realize, especially

825
00:44:26,615 --> 00:44:29,825
like when you hire new people, you'd
realize, oh man, like we're, there's

826
00:44:29,825 --> 00:44:31,475
a gap in our knowledge for them.

827
00:44:31,525 --> 00:44:31,815
Nirmal Mehta: Yeah.

828
00:44:32,065 --> 00:44:33,235
Bret: I realize we missed this process.

829
00:44:33,235 --> 00:44:34,255
Let's go create that sop.

830
00:44:34,445 --> 00:44:39,515
these are all things that the
individual or small team developers

831
00:44:39,515 --> 00:44:41,585
are figuring out how to do for the AI.

832
00:44:41,615 --> 00:44:44,675
'cause it turns out, if the AI
has all of that, he can perform or

833
00:44:44,675 --> 00:44:47,085
she, or the it, I guess it's an it.

834
00:44:47,205 --> 00:44:50,025
The, it will now look at all that stuff.

835
00:44:50,445 --> 00:44:54,195
And the simplest way for us to do that
today is to just put that in the repo.

836
00:44:54,645 --> 00:44:57,605
The problem for like us DevOps
people is that we don't.

837
00:44:58,105 --> 00:45:02,635
In very few place teams that I've worked
with is all of their DevOps knowledge.

838
00:45:02,665 --> 00:45:04,015
In the repo it's, I mean

839
00:45:04,130 --> 00:45:04,480
Nirmal Mehta: Right.

840
00:45:04,550 --> 00:45:05,160
It's all over.

841
00:45:05,215 --> 00:45:06,295
Bret: a lot of TOML, yeah

842
00:45:06,390 --> 00:45:06,680
Nirmal Mehta: Yeah.

843
00:45:06,955 --> 00:45:08,425
Bret: but they have documentation systems.

844
00:45:08,425 --> 00:45:10,045
They have Slack, they
have tribal knowledge.

845
00:45:10,045 --> 00:45:11,635
They have all these
different various places.

846
00:45:11,635 --> 00:45:15,205
They have some stuff that's basically
in the dashboards, like you imply

847
00:45:15,205 --> 00:45:17,755
knowledge through a dashboard
layout and what things are important

848
00:45:17,755 --> 00:45:18,925
and which metrics we care about

849
00:45:18,955 --> 00:45:19,305
Nirmal Mehta: Right.

850
00:45:19,411 --> 00:45:23,931
Bret: these are all over the place,
but the reason it's all in the AI repos

851
00:45:23,931 --> 00:45:27,171
today or the AI enabled code repos today.

852
00:45:27,321 --> 00:45:29,811
It's not because I think that's where
it's all gonna be in five years.

853
00:45:29,991 --> 00:45:33,201
It's because that's the quickest way we
can figure out how to dump text into the

854
00:45:33,201 --> 00:45:35,541
AI context so that the AI is way smarter.

855
00:45:35,811 --> 00:45:38,961
And we're eventually gonna figure
out whether it's MCP that connects

856
00:45:38,961 --> 00:45:39,981
all these things together.

857
00:45:40,221 --> 00:45:44,631
Like we might very quickly end up
moving past markdown in directories

858
00:45:44,901 --> 00:45:51,591
to, Hey, if you just point this to your
MCP to Confluence, then you can just

859
00:45:51,591 --> 00:45:55,071
point it to the different, you know,
URLs of all the DevOps documentation

860
00:45:55,071 --> 00:45:56,781
and you don't need the agent file,

861
00:45:56,815 --> 00:45:57,305
Nirmal Mehta: Runtime.

862
00:45:58,191 --> 00:45:58,581
Bret: Yeah.

863
00:45:59,361 --> 00:46:00,351
we're gonna figure all that out.

864
00:46:00,351 --> 00:46:03,111
These are, those are like the standards
and the practices and the conventions,

865
00:46:03,111 --> 00:46:04,281
I think we're all gonna figure out.

866
00:46:04,281 --> 00:46:07,941
But like right now, since we're in the
bleeding edge, it just happens to all be

867
00:46:08,011 --> 00:46:11,091
markdown with SMO front matter and a repo.

868
00:46:11,271 --> 00:46:12,711
So go check that stuff out.

869
00:46:12,931 --> 00:46:15,751
the last thing I wanna talk about before
we wrap up, 'cause I know we're hitting

870
00:46:15,751 --> 00:46:20,426
our limit, but this is the story I tell
to people that I, this is my vision now.

871
00:46:20,426 --> 00:46:23,126
Like I didn't really have this
vision nailed down a year ago.

872
00:46:23,126 --> 00:46:23,516
I think.

873
00:46:24,016 --> 00:46:25,816
think everything I was saying,
I have to go back and listen to

874
00:46:25,816 --> 00:46:28,566
our first episode 'cause I feel
like everything we're predicting.

875
00:46:29,066 --> 00:46:30,056
Has come to pass.

876
00:46:30,266 --> 00:46:32,366
Like it's coming faster than I thought.

877
00:46:32,666 --> 00:46:33,566
Thought like, I

878
00:46:33,725 --> 00:46:33,945
Nirmal Mehta: Yes,

879
00:46:34,445 --> 00:46:34,595
Bret: it

880
00:46:34,640 --> 00:46:35,300
Nirmal Mehta: that's for sure.

881
00:46:35,705 --> 00:46:37,925
Bret: before we had reliable
models for DevOps work.

882
00:46:38,425 --> 00:46:41,185
this is not about replacing our jobs.

883
00:46:41,365 --> 00:46:42,685
Like everyone's concerned about it.

884
00:46:42,685 --> 00:46:45,685
I'm concerned about it, especially
as a content creator because course

885
00:46:45,685 --> 00:46:47,015
sales are down I wouldn't say no one.

886
00:46:47,135 --> 00:46:50,315
People are buying courses less and
they certainly don't need courses

887
00:46:50,315 --> 00:46:54,515
that teach them the manual anymore,
which is why my whole course vibe is

888
00:46:54,515 --> 00:46:59,255
completely shifted to advisory and best
practices and architecture type stuff.

889
00:46:59,255 --> 00:47:02,405
Not, Hey, this is how you need
to learn every single command.

890
00:47:02,435 --> 00:47:05,405
'cause you're not probably gonna type
the commands here very much longer.

891
00:47:06,755 --> 00:47:10,715
I started 30 years ago and this slide
I'm showing on screen for those audio

892
00:47:10,715 --> 00:47:14,945
listeners is about the major shifts
that about once a decade, although

893
00:47:14,945 --> 00:47:20,315
these are increasing now to about two
a decade of major shifts in it I see

894
00:47:20,315 --> 00:47:24,965
as an operator and a DevOps engineer
that have been happening my entire

895
00:47:24,965 --> 00:47:27,125
career and are gonna continue to happen.

896
00:47:27,125 --> 00:47:31,225
And it just so happens that managing
agents, which are a bunch of LLMs

897
00:47:31,435 --> 00:47:32,905
is just the next phase of it.

898
00:47:33,025 --> 00:47:36,495
And in the nineties and this graphic,
what it's supposed to represent is

899
00:47:36,495 --> 00:47:39,945
at the top is in the nineties I got
to be a part of the mainframe to

900
00:47:39,945 --> 00:47:44,105
PC wave where I was literally like
shutting down a mainframe or mini

901
00:47:44,105 --> 00:47:46,805
frames or Unix servers that were, stuff

902
00:47:46,895 --> 00:47:47,315
Nirmal Mehta: X.

903
00:47:47,525 --> 00:47:48,665
Bret: to pc.

904
00:47:48,695 --> 00:47:50,765
Yeah, shifting to PCs that
didn't even have mice yet.

905
00:47:50,765 --> 00:47:52,655
Like they were literally
dos with word Perfect.

906
00:47:53,105 --> 00:47:53,465
Right.

907
00:47:53,495 --> 00:47:56,165
And then we eventually got mice
and we got windows, and then

908
00:47:56,165 --> 00:47:57,605
we got Windows 95 and then.

909
00:47:58,040 --> 00:48:02,360
Like we started to do distributed
computing and that enabled me to go from

910
00:48:02,360 --> 00:48:05,810
managing one large mainframe, which I went
to voc, two years of vocational school

911
00:48:05,810 --> 00:48:11,210
on how to manage Honeywell and HP and
Solaris and these kind of things, right?

912
00:48:11,710 --> 00:48:13,540
but I was only able to manage
a couple of these things.

913
00:48:13,540 --> 00:48:17,560
We had like four dudes to
manage five mainframes and many

914
00:48:17,560 --> 00:48:18,790
frames, and that was it, right?

915
00:48:18,790 --> 00:48:19,930
So it was like a one-to-one.

916
00:48:20,290 --> 00:48:23,950
And then in, in the early two
thousands, we invented the

917
00:48:23,950 --> 00:48:26,170
technology as VMs, virtual machines.

918
00:48:26,170 --> 00:48:30,250
And I can remember those days very
distinctly because everyone told me at

919
00:48:30,250 --> 00:48:34,510
the, my enterprise, I was working at a
large city, half a million people, and.

920
00:48:35,010 --> 00:48:36,900
Man managing a city of
a half million people.

921
00:48:36,960 --> 00:48:37,800
And I remember

922
00:48:37,840 --> 00:48:37,960
Nirmal Mehta: I.

923
00:48:38,310 --> 00:48:41,190
Bret: the meeting where we were
just trying to describe how VMs work

924
00:48:41,250 --> 00:48:43,710
and none of the system engineers
thought it was a good idea.

925
00:48:43,710 --> 00:48:46,140
Everyone thought it was gonna
wreck, shop and crash things,

926
00:48:46,140 --> 00:48:46,920
and it would be horrible.

927
00:48:46,950 --> 00:48:50,700
And we were having to give
analogies, like talking about like

928
00:48:50,700 --> 00:48:53,940
the blade that goes inside the
mainframe is actually a different

929
00:48:53,940 --> 00:48:55,770
compute that you can shard it off.

930
00:48:55,800 --> 00:48:58,410
And this is kind of like
that, but in software.

931
00:48:58,500 --> 00:49:00,300
And everyone's like, no,
it's a horrible idea.

932
00:49:00,300 --> 00:49:01,590
We're running a kernel and a kernel.

933
00:49:01,590 --> 00:49:03,660
Well anyway, VMs are the standard.

934
00:49:03,660 --> 00:49:08,640
They have been for decades, and it
enabled us to easily manage 10 or

935
00:49:08,640 --> 00:49:10,170
even a hundred machines ourself.

936
00:49:10,350 --> 00:49:14,340
Then you fast forward to the cloud now
system engineers and DevOps people can

937
00:49:14,340 --> 00:49:16,800
manage hundreds of servers themselves.

938
00:49:16,800 --> 00:49:17,800
They, that's the dawn

939
00:49:17,875 --> 00:49:19,165
Nirmal Mehta: say thousands at this point,

940
00:49:19,165 --> 00:49:19,735
Bret: thousands.

941
00:49:19,945 --> 00:49:21,949
Well, we had to create
new tools for that, right?

942
00:49:21,949 --> 00:49:22,369
That when we

943
00:49:22,518 --> 00:49:22,758
Nirmal Mehta: right?

944
00:49:22,909 --> 00:49:24,529
Bret: that in 2010, we didn't have

945
00:49:24,558 --> 00:49:24,918
Nirmal Mehta: Yes.

946
00:49:24,979 --> 00:49:25,309
Bret: tools.

947
00:49:25,848 --> 00:49:26,178
Nirmal Mehta: Right.

948
00:49:26,229 --> 00:49:28,509
Bret: then Ansible shows up,
then Terraform shows up, and

949
00:49:28,509 --> 00:49:29,829
now we're able to do thousands.

950
00:49:30,099 --> 00:49:33,489
Then containers show up and
we stop talking about servers.

951
00:49:33,579 --> 00:49:38,469
And now one engineer can manage
if not thousands, if not tens

952
00:49:38,493 --> 00:49:39,758
Nirmal Mehta: Ends of thousands,

953
00:49:39,789 --> 00:49:40,329
Bret: workloads.

954
00:49:40,688 --> 00:49:41,228
Nirmal Mehta: correct.

955
00:49:41,349 --> 00:49:44,169
Bret: happens in the 2010s
to, you know, Kubernetes takes

956
00:49:44,169 --> 00:49:45,489
us all the way up to 2020.

957
00:49:45,989 --> 00:49:49,229
And then we get things like off
shooting from that, like Wasm,

958
00:49:50,069 --> 00:49:52,049
serverless, lambda, stuff like that.

959
00:49:52,529 --> 00:49:54,359
But that is not replacing engineers.

960
00:49:54,389 --> 00:49:57,809
None of those things in
my job replaced anyone.

961
00:49:57,869 --> 00:50:00,779
I literally took the same two
guys that were managing hardware

962
00:50:00,779 --> 00:50:02,099
installations in the data center.

963
00:50:02,399 --> 00:50:06,779
They were so worried about
their jobs in 2005, and I gave

964
00:50:06,779 --> 00:50:08,279
them jobs managing the VMs.

965
00:50:08,279 --> 00:50:11,879
We had to retrain them, retool
'em, but now they had VMs to manage

966
00:50:12,119 --> 00:50:13,229
instead of physical hardware.

967
00:50:13,529 --> 00:50:14,819
And the same thing happened.

968
00:50:14,819 --> 00:50:15,509
We went to the cloud.

969
00:50:15,509 --> 00:50:18,479
We had to take everybody that
was managing SSHS to a bunch

970
00:50:18,479 --> 00:50:19,694
of Linux servers in a closet.

971
00:50:19,949 --> 00:50:21,299
And teach them cloud tooling.

972
00:50:21,569 --> 00:50:23,909
And now we have better cloud
tooling so we can manage even more.

973
00:50:24,179 --> 00:50:27,029
We had to teach people Docker so they
could manage thousands of workloads

974
00:50:27,029 --> 00:50:28,229
instead of thousands of servers.

975
00:50:28,649 --> 00:50:32,609
And I think that AI today, if we keep
reiterating this over and over, that I

976
00:50:32,609 --> 00:50:38,009
need to change this slide now that we're
just gonna be managing a series of agents,

977
00:50:38,789 --> 00:50:40,859
gonna be directing agents all day long.

978
00:50:40,859 --> 00:50:46,289
I'm already doing this I'm like, I need
you to create this GitHub action workflow.

979
00:50:46,289 --> 00:50:50,309
Please use my skill that outlines
my favorite things that I always

980
00:50:50,309 --> 00:50:52,109
want in my GitHub action workflows.

981
00:50:52,379 --> 00:50:55,589
Please help me meet these goals
and ask me any questions along

982
00:50:55,589 --> 00:50:56,669
the way that you need answered.

983
00:50:57,029 --> 00:50:59,519
Turns out the new opus
models do that really well.

984
00:50:59,519 --> 00:51:02,689
It comes back and asks you questions with
a nice little menu system that you get to

985
00:51:02,689 --> 00:51:07,249
choose what it recommends versus what you
think or do I wanna opt in like to a hand

986
00:51:07,249 --> 00:51:09,109
filled, crafted answer to the question?

987
00:51:09,559 --> 00:51:12,049
It interviews me just like a
junior engineer would when they

988
00:51:12,049 --> 00:51:13,879
go, well, you didn't teach me.

989
00:51:14,269 --> 00:51:17,779
How you expect the security
to be in this YAML file.

990
00:51:17,779 --> 00:51:20,264
And I would realize, oh yes,
let me teach you young Padawan.

991
00:51:20,594 --> 00:51:22,664
That's the same thing I'm
doing to AI right now.

992
00:51:23,054 --> 00:51:27,134
And I'm able to do that thing is going
off and spinning up a new workload that

993
00:51:27,134 --> 00:51:28,954
is building a new, workflow for me.

994
00:51:29,014 --> 00:51:32,524
And then I hop over to a new
terminal where I'm SSH into a

995
00:51:32,524 --> 00:51:36,334
server, and some people are going
crazy and actually installing

996
00:51:36,334 --> 00:51:39,484
these harnesses, like on running
servers to troubleshoot the server.

997
00:51:39,484 --> 00:51:43,084
I wouldn't go that far, but maybe
I could use my local one to SSH

998
00:51:43,084 --> 00:51:46,814
into a server and it could send
SSH commands one at a time maybe.

999
00:51:46,844 --> 00:51:50,554
I feel like probably gonna be a
day where we have a tiny little

1000
00:51:50,554 --> 00:51:54,244
agent harness that's like a single
binary that doesn't touch anything.

1001
00:51:54,244 --> 00:51:55,504
It has no right access.

1002
00:51:55,594 --> 00:51:57,604
And we can maybe throw that
on a server temporarily.

1003
00:51:57,604 --> 00:51:59,824
Maybe it runs in a container
and then we pull it off when

1004
00:51:59,824 --> 00:52:00,754
we're done and no one ever knew.

1005
00:52:01,564 --> 00:52:02,974
essentially there to help us troubleshoot.

1006
00:52:03,964 --> 00:52:07,384
I mean, you have probably heard
of the net shoot Docker container

1007
00:52:07,384 --> 00:52:09,334
that our friend Nikolai has created

1008
00:52:09,348 --> 00:52:09,918
Nirmal Mehta: Yes,

1009
00:52:10,024 --> 00:52:10,294
Bret: now.

1010
00:52:10,294 --> 00:52:13,114
Well, that's gonna get replaced
with an AI that has all those tools

1011
00:52:13,114 --> 00:52:14,494
built in and just goes and does the

1012
00:52:14,598 --> 00:52:15,108
Nirmal Mehta: correct.

1013
00:52:15,214 --> 00:52:15,364
Bret: us.

1014
00:52:15,364 --> 00:52:15,664
Right?

1015
00:52:15,773 --> 00:52:16,263
Nirmal Mehta: Correct.

1016
00:52:17,074 --> 00:52:21,154
Bret: it's all this really means is
I'm managing more infrastructure, more

1017
00:52:21,154 --> 00:52:24,874
automation that I have to do because
there's more code being shipped.

1018
00:52:24,874 --> 00:52:29,014
More and more people making code
now because now we're talking, we're

1019
00:52:29,014 --> 00:52:30,274
hearing about product managers.

1020
00:52:30,274 --> 00:52:33,534
They're able to code again because
they're able to actually make

1021
00:52:33,534 --> 00:52:36,714
commits that they wouldn't otherwise
have time to make because they can

1022
00:52:36,714 --> 00:52:38,004
do it in a fraction of the time.

1023
00:52:38,434 --> 00:52:39,454
this is just gonna keep happening.

1024
00:52:39,454 --> 00:52:42,514
So I'm gonna just help
everybody feel comfortable

1025
00:52:42,514 --> 00:52:44,734
that your job isn't going away.

1026
00:52:44,854 --> 00:52:50,004
It's just changing maybe faster than
it would in a normal 10 year arc. Maybe

1027
00:52:50,004 --> 00:52:51,714
this is gonna be like a five year arc

1028
00:52:51,764 --> 00:52:51,994
Nirmal Mehta: Yeah.

1029
00:52:52,954 --> 00:52:54,364
I mean, we're already two years in.

1030
00:52:54,364 --> 00:52:54,934
Bret: using AI.

1031
00:52:55,024 --> 00:52:55,324
Yeah,

1032
00:52:55,469 --> 00:52:55,759
Nirmal Mehta: Yeah.

1033
00:52:55,869 --> 00:52:55,929
Bret: I

1034
00:52:56,118 --> 00:52:56,928
Nirmal Mehta: I think you're right.

1035
00:52:56,988 --> 00:52:58,758
This is like, this is a five year arc.

1036
00:52:58,858 --> 00:52:59,008
Bret: Yeah,

1037
00:52:59,508 --> 00:53:01,533
Nirmal Mehta: so that was well put.

1038
00:53:02,033 --> 00:53:07,903
I think on an earlier episode, I
reiterated, you know, this won't

1039
00:53:08,403 --> 00:53:14,003
replace you, but someone who
knows how to use these tools will

1040
00:53:14,503 --> 00:53:14,773
Bret: Yeah.

1041
00:53:14,888 --> 00:53:19,288
Nirmal Mehta: if you need some impetus
when you're sitting there trying to figure

1042
00:53:19,288 --> 00:53:25,768
out what question to answer for yourself,
with these tools, think about how would

1043
00:53:25,768 --> 00:53:33,138
I use these tools to do some small aspect
of my current job day to day, but also

1044
00:53:33,138 --> 00:53:39,188
think about how would I use these tools
to operate an environment with instead

1045
00:53:39,188 --> 00:53:45,968
of a hundred containers or a thousand
containers, maybe 10,000 containers or a

1046
00:53:45,968 --> 00:53:48,528
hundred thousand application containers

1047
00:53:49,028 --> 00:53:49,318
Bret: Yeah.

1048
00:53:49,392 --> 00:53:51,732
Nirmal Mehta: How would I
use these agents to do that?

1049
00:53:52,232 --> 00:53:57,252
And maybe that's a good question,
for exploring these tools is

1050
00:53:57,252 --> 00:53:58,902
like getting to that answer.

1051
00:53:59,262 --> 00:54:03,902
I would use these tools in this
way, to operate systems at that

1052
00:54:03,902 --> 00:54:05,812
scale because I think you're right.

1053
00:54:05,812 --> 00:54:11,142
we've seen this trend from mainframe
to containers and serverless and cloud,

1054
00:54:11,642 --> 00:54:18,202
that one person's scope of responsibility
has just expanded, but the surface

1055
00:54:18,202 --> 00:54:20,602
area of applications has also expanded.

1056
00:54:21,102 --> 00:54:25,522
And so there is a need, and
it won't eliminate your job.

1057
00:54:26,022 --> 00:54:30,482
It's just the person that knows how to
use these tools will be doing your job.

1058
00:54:31,228 --> 00:54:31,683
Bret: Right, right.

1059
00:54:31,952 --> 00:54:34,722
Nirmal Mehta: So hopefully at the end
of this we're almost close to the end

1060
00:54:34,722 --> 00:54:36,612
of our context window for this episode.

1061
00:54:37,092 --> 00:54:40,392
hopefully we've convinced you
that you could still maintain

1062
00:54:40,392 --> 00:54:45,812
that skepticism about, the output
of LLMs and these tools, but.

1063
00:54:46,312 --> 00:54:50,422
Please do yourself a favor and start
using 'em and start getting used

1064
00:54:50,422 --> 00:54:54,502
to them and start understanding
how to use these tools effectively.

1065
00:54:54,772 --> 00:54:56,362
'cause it's just like any other skill.

1066
00:54:56,762 --> 00:55:00,032
it's, you know, you have to learn, you
have to exercise that muscle and you have

1067
00:55:00,032 --> 00:55:02,782
to understand what it can and can't do.

1068
00:55:03,282 --> 00:55:07,682
with that, I mean, we got halfway
through our list today, so tune

1069
00:55:07,682 --> 00:55:09,302
in again for another episode.

1070
00:55:09,302 --> 00:55:12,002
We've got so much more,

1071
00:55:12,212 --> 00:55:13,052
Bret: So what you

1072
00:55:13,172 --> 00:55:14,522
Nirmal Mehta: context to explore.

1073
00:55:14,552 --> 00:55:15,682
Bret: coming, want me to

1074
00:55:15,892 --> 00:55:16,702
Nirmal Mehta: Yeah, let's,

1075
00:55:16,852 --> 00:55:17,452
Bret: that are coming up.

1076
00:55:17,782 --> 00:55:18,622
Nirmal Mehta: yeah, let's do it.

1077
00:55:18,742 --> 00:55:21,832
Bret: D, we're gonna have Solomon,
hes at Dagger back on the call.

1078
00:55:22,222 --> 00:55:26,632
We're gonna have, the co-founders of
Mineral, which is a GitHub Actions AI.

1079
00:55:27,532 --> 00:55:31,972
Troubleshooter that essentially I've
been using for now a month that it just

1080
00:55:31,972 --> 00:55:37,702
stares at my GitHub all day long its own
running 24 7, looking for ways to help.

1081
00:55:37,912 --> 00:55:40,222
It doesn't do code commits
in the classic way.

1082
00:55:40,222 --> 00:55:41,002
that's a different thing.

1083
00:55:41,002 --> 00:55:42,772
It's not trying to like build my apps.

1084
00:55:43,042 --> 00:55:50,552
What it's doing is looking for failed
GitHub workflows or, CVEs that Abott

1085
00:55:50,552 --> 00:55:55,472
has suggested that I need to implement
through a PR and it's reminding

1086
00:55:55,472 --> 00:55:59,342
me that those aren't there or it's
recommending strategies for improving

1087
00:55:59,342 --> 00:56:01,082
the speed of my GitHub workflows.

1088
00:56:01,442 --> 00:56:05,222
it's highlighting whether the CVEs
are necessary to get fixed in a

1089
00:56:05,222 --> 00:56:07,862
particular depend bot or renovate PR.

1090
00:56:08,052 --> 00:56:11,172
It's, there is so much stuff that
it's doing, but it's essentially like.

1091
00:56:11,532 --> 00:56:17,142
I look at it as like the janitor, like
the custodian of my repos, and it solves a

1092
00:56:17,142 --> 00:56:21,912
very specific problem set that I'm excited
to have them on to talk about because I'm

1093
00:56:22,052 --> 00:56:22,632
Nirmal Mehta: Sounds awesome.

1094
00:56:22,782 --> 00:56:25,392
Bret: I feel like I'm currently a
free customer, but it's providing

1095
00:56:25,392 --> 00:56:26,982
use to me as an individual.

1096
00:56:26,982 --> 00:56:28,932
I can only imagine how helpful
it's gonna be for a team.

1097
00:56:29,182 --> 00:56:30,532
it's kinda like an automated Jira.

1098
00:56:30,562 --> 00:56:34,282
Like it makes the tickets, the fix for
the ticket, and then if you just say

1099
00:56:34,282 --> 00:56:37,442
approve it fixes it with a PR or whatever
it needs to do to fix that thing.

1100
00:56:37,882 --> 00:56:38,162
Nirmal Mehta: cool.

1101
00:56:38,388 --> 00:56:40,308
Bret: And so we've got SRE ones coming up.

1102
00:56:40,308 --> 00:56:44,528
I'm gonna try to get someone from
any shift on, which is an SRE type of

1103
00:56:44,528 --> 00:56:46,158
tool that's similar to the mineral.

1104
00:56:46,578 --> 00:56:50,868
we're gonna try to get someone on from
the leadership of GitHub on, soon.

1105
00:56:51,138 --> 00:56:54,198
And so we're gonna actually have
someone talking about GitHub Actions

1106
00:56:54,228 --> 00:56:56,598
that actually controls the teams
that are working on that there.

1107
00:56:56,938 --> 00:57:00,148
we're gonna, I'm working to get someone
from GitHub next on, because GitHub

1108
00:57:00,148 --> 00:57:03,688
next is where a lot of this innovation
is happening in the world of what

1109
00:57:03,688 --> 00:57:09,028
they're calling continuous AI or Agentic
workflows, which is all this stuff they're

1110
00:57:09,028 --> 00:57:10,858
doing with co-pilot inside of GitHub.

1111
00:57:11,248 --> 00:57:13,448
and then that's probably only
half the list I have that

1112
00:57:13,448 --> 00:57:14,258
we're currently working on.

1113
00:57:14,258 --> 00:57:16,328
So we've got more episodes coming.

1114
00:57:16,328 --> 00:57:21,498
I'm super excited to finally get into
some real data around how do I implement

1115
00:57:21,498 --> 00:57:25,338
a workflow that's not gonna hallucinate
or be at risk of a prompt injection or

1116
00:57:25,338 --> 00:57:27,168
it's gonna leak credentials through MCP.

1117
00:57:27,298 --> 00:57:31,433
I'm super ready to get into the nerdy
details, and that's what I think

1118
00:57:31,433 --> 00:57:32,453
this year is gonna be all about.

1119
00:57:32,953 --> 00:57:33,323
Nirmal Mehta: Awesome.

1120
00:57:33,353 --> 00:57:38,723
And with that, please let your
colleagues, your friends, your family,

1121
00:57:38,723 --> 00:57:41,473
your neighbor, you know, or your mom,

1122
00:57:41,543 --> 00:57:42,518
Bret: Agentic DevOps

1123
00:57:42,763 --> 00:57:44,593
Nirmal Mehta: to subscribe.

1124
00:57:44,593 --> 00:57:49,333
And like the Agentic DevOps podcast, you
can check us out at Agentic DevOps dot

1125
00:57:49,333 --> 00:57:51,613
fm if that's not the way you found us.

1126
00:57:51,893 --> 00:57:52,673
please share,

1127
00:57:52,713 --> 00:57:53,073
Bret: favorite

1128
00:57:53,163 --> 00:57:56,203
Nirmal Mehta: we'd love to hear from
you, if there's a specific topic that

1129
00:57:56,203 --> 00:57:57,553
you would like us to cover or go.

1130
00:57:58,253 --> 00:58:00,143
dive deep into, please let us know.

1131
00:58:00,473 --> 00:58:04,453
We've got a long list and, can't
wait to dive into that in this

1132
00:58:04,453 --> 00:58:06,883
season of the Agentic DevOps podcast.

1133
00:58:07,383 --> 00:58:07,713
Bret: yeah.

1134
00:58:07,773 --> 00:58:09,363
Find us both on LinkedIn.

1135
00:58:10,233 --> 00:58:11,103
he's Nirmal Mehta.

1136
00:58:11,103 --> 00:58:12,153
I'm Bret Fisher.

1137
00:58:12,153 --> 00:58:13,143
We're on LinkedIn.

1138
00:58:13,143 --> 00:58:14,253
We're on Blue Sky.

1139
00:58:14,253 --> 00:58:15,483
I'm still on Twitter.

1140
00:58:15,813 --> 00:58:16,563
You can find us on our
discord@discord.gg/devops.

1141
00:58:20,623 --> 00:58:23,263
All the links are in the show
notes for what we talked about

1142
00:58:23,263 --> 00:58:26,943
today, and we will see you on the
next episode of Agentic DevOps.

1143
00:58:27,693 --> 00:58:28,143
Chow.

1144
00:58:28,662 --> 00:58:31,002
Bret AI July 2025: Thanks for joining
us, and I'll see you in the next episode.