1 00:00:00,000 --> 00:00:02,750 Bret: Welcome to DevOps and Docker talk, and I'm your host, Bret. 2 00:00:03,630 --> 00:00:05,460 This episode is a special one. 3 00:00:05,520 --> 00:00:10,860 It's actually the first episode from a totally new podcast I launched called 4 00:00:10,860 --> 00:00:15,930 Agentic DevOps, and that podcast is gonna run in parallel with this one. 5 00:00:16,290 --> 00:00:19,650 So this one, the goal is still for the last six years. 6 00:00:20,205 --> 00:00:24,795 Everything related to containers, cloud native Kubernetes, and Docker, 7 00:00:24,795 --> 00:00:27,435 and the DevOps workloads around that. 8 00:00:27,975 --> 00:00:29,475 And I don't plan on changing any of that. 9 00:00:29,475 --> 00:00:31,455 We're gonna still have the same guests. 10 00:00:31,455 --> 00:00:36,135 A certain amount of those will be AI related guests, but I was seeing a trend. 11 00:00:36,135 --> 00:00:38,379 I. That I'll talk about in the show. 12 00:00:38,739 --> 00:00:44,290 And I thought that Agentic DevOps was going to be a big thing here in 2025. 13 00:00:44,290 --> 00:00:49,480 So a few months back we started working on content episodes and theming and branding. 14 00:00:49,690 --> 00:00:55,720 A whole new podcast that I recommend you check out at agenticdevops.Fm 15 00:00:56,000 --> 00:00:57,110 links in the show notes. 16 00:00:57,380 --> 00:01:00,680 And this is the first episode from that podcast that I'm just presenting 17 00:01:00,680 --> 00:01:02,270 here so that you can check it out. 18 00:01:02,556 --> 00:01:07,446 Neral and I talk theory around what we see coming and what might be 19 00:01:07,446 --> 00:01:12,126 a huge shift in how we use AI to do our jobs as DevOps engineers. 20 00:01:12,516 --> 00:01:16,506 And that intention for that show is to have more guests and to really 21 00:01:16,566 --> 00:01:20,466 dial in and focus on that very niche topic, at least for this year. 22 00:01:20,466 --> 00:01:20,976 Who knows? 23 00:01:21,026 --> 00:01:23,306 it might be a bigger deal than this show, 24 00:01:23,556 --> 00:01:27,096 so if you enjoy this episode, subscribe to that second podcast of 25 00:01:27,096 --> 00:01:29,556 mine, and now I'm gonna have two. 26 00:01:29,736 --> 00:01:30,756 So I hope you enjoy. 27 00:01:41,084 --> 00:01:45,824 Welcome to the first episode of my new podcast, a Agentic DevOps. 28 00:01:45,964 --> 00:01:46,954 this episode. 29 00:01:47,674 --> 00:01:53,374 Is kicking off what I think is going to be a big topic for my entire year, 30 00:01:53,554 --> 00:01:59,704 probably for the next few years around wrangling AI into some usable format. 31 00:02:00,094 --> 00:02:06,384 For DevOps, you probably heard of AI agents by now, or the MCP protocol. 32 00:02:06,504 --> 00:02:09,384 I guess I should just say MCP, since P stands for protocol. 33 00:02:09,714 --> 00:02:14,454 And these two things together are creating potentially something 34 00:02:14,454 --> 00:02:18,384 very useful for platform engineering, DevOps, and that stuff. 35 00:02:18,684 --> 00:02:21,174 it has so much potential that. 36 00:02:21,564 --> 00:02:26,304 In the first quarter of 2025, I kind of thought this was gonna be a big deal. 37 00:02:26,304 --> 00:02:29,754 This was gonna be, uh, if we can figure out how to keep these things 38 00:02:29,754 --> 00:02:32,394 from hallucinating and going crazy in our infrastructure, this could 39 00:02:32,394 --> 00:02:36,834 potentially be the AI shift for infrastructure that I was waiting for. 40 00:02:37,254 --> 00:02:39,294 So started this podcast. 41 00:02:39,354 --> 00:02:43,749 We recorded our first episode at KubeCon at the beginning of April, 42 00:02:43,749 --> 00:02:51,024 2025, and this is gonna be a series of very specific episodes around getting. 43 00:02:52,089 --> 00:02:56,679 Ais to do useful automation and work for DevOps, platform engineering, 44 00:02:56,859 --> 00:02:59,979 infrastructure management, cloud, you know, all those things 45 00:03:00,309 --> 00:03:02,619 beyond just writing YAML, right? 46 00:03:02,889 --> 00:03:06,579 So the, the intro for this podcast, there's a separate episode for intro. 47 00:03:06,579 --> 00:03:09,759 It kind of goes into my whole theory of why I think this is gonna be a thing. 48 00:03:10,059 --> 00:03:13,599 And this episode we really try to break down the basics and fundamentals for 49 00:03:13,599 --> 00:03:15,189 those of you that are catching up. 50 00:03:15,189 --> 00:03:16,524 Because it's a lot. 51 00:03:16,674 --> 00:03:17,994 There's a lot going on. 52 00:03:18,174 --> 00:03:22,924 It seems like We have announcements every day this year around AI agents or 53 00:03:22,924 --> 00:03:24,954 Agentic, ai, however you wanna call it. 54 00:03:25,204 --> 00:03:29,224 I am calling it Agentic DevOps, and hoping that name will stick. 55 00:03:29,674 --> 00:03:32,734 Now, this episode, since it's from the beginning of April. 56 00:03:33,439 --> 00:03:37,309 And it is technically now just getting released at the beginning of June. 57 00:03:37,819 --> 00:03:40,189 We're a little bit behind on launching this new podcast. 58 00:03:40,519 --> 00:03:42,709 Um, I think everything in it's still relevant. 59 00:03:42,709 --> 00:03:44,179 There's just been a lot more since. 60 00:03:44,389 --> 00:03:46,009 And I don't know the frequency yet. 61 00:03:46,009 --> 00:03:48,229 I don't know how often this podcast is gonna happen. 62 00:03:48,469 --> 00:03:50,569 It could be potentially every other week. 63 00:03:50,749 --> 00:03:51,619 It could be weekly. 64 00:03:51,619 --> 00:03:54,649 I just don't know yet because we are not gonna do the same thing 65 00:03:54,649 --> 00:03:56,719 here as on my usual podcast. 66 00:03:56,719 --> 00:03:59,509 If you're someone who knows that one DevOps and Docker talk that I've 67 00:03:59,509 --> 00:04:02,869 been doing the last seven years, that one is still gonna have AI in it. 68 00:04:02,869 --> 00:04:07,039 But this one is very specific and there might be a few episodes that have 69 00:04:07,039 --> 00:04:11,629 syndication or whatever you wanna call it, of the episodes on both podcasts. 70 00:04:12,184 --> 00:04:16,594 But most of the time we're gonna keep the focus of just everything, 71 00:04:16,594 --> 00:04:20,434 DevOps, everything, containers on the DevOps and Docker talk show. 72 00:04:20,494 --> 00:04:24,934 And this one is gonna be very specific around implementing useful 73 00:04:25,564 --> 00:04:31,744 AI related things for Agentic DevOps, or automating our DevOps with robots. 74 00:04:32,224 --> 00:04:35,734 So I hope you enjoyed this episode with Nirmal from KubeCon London. 75 00:04:41,457 --> 00:04:42,187 Hey, I'm Bret. 76 00:04:42,257 --> 00:04:43,337 And we're at Kon. 77 00:04:44,282 --> 00:04:45,972 We are Hi, Nirmal. 78 00:04:46,292 --> 00:04:47,072 Nirmal: I'm Nirmal Metha. 79 00:04:47,072 --> 00:04:50,972 I'm a principal specialist solution architect at AWS and these are 80 00:04:50,972 --> 00:04:55,472 my views and not of my employers, but this episode is all about 81 00:04:55,562 --> 00:04:56,042 Bret: AI 82 00:04:56,042 --> 00:04:56,822 Nirmal: agents 83 00:04:57,122 --> 00:04:59,762 Bret: for DevOps and platform engineering. 84 00:04:59,762 --> 00:04:59,792 Ooh. 85 00:04:59,852 --> 00:05:03,332 So let's just start off real quick with what is an AI agent? 86 00:05:03,362 --> 00:05:03,542 Okay. 87 00:05:03,542 --> 00:05:07,922 So we've heard of ai, we know ai, gen, AI chat, GPT. 88 00:05:08,132 --> 00:05:09,002 We've talked about. 89 00:05:09,342 --> 00:05:12,672 running LLMs, running inference on platforms. 90 00:05:12,702 --> 00:05:12,882 Yep. 91 00:05:12,942 --> 00:05:17,922 And that we are managing the workloads that provide other people services. 92 00:05:17,982 --> 00:05:18,432 Absolutely. 93 00:05:18,702 --> 00:05:21,642 So how is AI agents different than that? 94 00:05:22,202 --> 00:05:24,742 Nirmal: This is a air in terms of bleeding edge. 95 00:05:24,952 --> 00:05:25,192 Yeah. 96 00:05:25,672 --> 00:05:26,812 This is it, right? 97 00:05:26,812 --> 00:05:26,872 Yeah. 98 00:05:26,872 --> 00:05:28,132 Like we're a year ago. 99 00:05:28,132 --> 00:05:28,492 No one 100 00:05:28,492 --> 00:05:28,792 Bret: had this 101 00:05:28,792 --> 00:05:29,212 Nirmal: term 102 00:05:29,272 --> 00:05:29,902 Bret: six months ago. 103 00:05:29,902 --> 00:05:30,412 I don't think anybody's 104 00:05:30,412 --> 00:05:30,832 Nirmal: talking about it. 105 00:05:30,832 --> 00:05:31,417 I'm very few people. 106 00:05:31,487 --> 00:05:32,457 Yeah, very few people. 107 00:05:32,957 --> 00:05:36,977 and we've seen it in the news a lot of vendors and big companies 108 00:05:37,027 --> 00:05:41,637 announcing Agentic ai, that's another term's ai, so AI agents, Agentic 109 00:05:42,047 --> 00:05:49,937 It's giving your LLM, like your chat, GPT or your Claude or local LM Lama. 110 00:05:49,997 --> 00:05:50,297 Yeah. 111 00:05:51,197 --> 00:05:55,157 Access to run commands. 112 00:05:55,257 --> 00:05:56,337 On your behalf. 113 00:05:56,697 --> 00:05:57,867 Or on its behalf. 114 00:05:58,437 --> 00:05:58,617 Bret: Yeah. 115 00:05:58,617 --> 00:06:01,047 And we call those tools like that, if you hear that word. 116 00:06:01,107 --> 00:06:01,227 Tools. 117 00:06:01,227 --> 00:06:01,407 Yeah. 118 00:06:01,437 --> 00:06:04,137 That's like the generic tool, like I guess a shell. 119 00:06:04,602 --> 00:06:05,532 Could be a tool. 120 00:06:05,562 --> 00:06:06,042 Correct. 121 00:06:06,092 --> 00:06:08,552 Reading a file could be a tool. 122 00:06:08,702 --> 00:06:12,572 Accessing a remote, API of a web service is a tool. 123 00:06:12,602 --> 00:06:12,872 Yep. 124 00:06:12,902 --> 00:06:14,402 Searching could be a tool. 125 00:06:14,402 --> 00:06:18,322 And so these tools what what makes that different than what we've 126 00:06:18,322 --> 00:06:20,962 been seeing in our code editors? 127 00:06:20,992 --> 00:06:21,322 Yeah. 128 00:06:21,472 --> 00:06:22,202 How is that different? 129 00:06:22,586 --> 00:06:25,636 Nirmal: I'm a platform engineer and I want to build out an 130 00:06:25,636 --> 00:06:27,316 EKS cluster using Terraform. 131 00:06:27,316 --> 00:06:28,096 That's what we use. 132 00:06:28,396 --> 00:06:31,816 So I'll ask let's say Claude or chat GBT. 133 00:06:31,936 --> 00:06:32,116 Yeah. 134 00:06:32,176 --> 00:06:36,796 I'm a platform engineer and I want to build a production ready EKS cluster. 135 00:06:36,826 --> 00:06:37,636 Please create. 136 00:06:37,936 --> 00:06:42,166 The assets I need, and it will spit out some Terraform. 137 00:06:42,171 --> 00:06:42,436 Yaml, right? 138 00:06:42,436 --> 00:06:42,646 Yeah. 139 00:06:43,006 --> 00:06:44,231 Bret: And it's writing text. 140 00:06:44,231 --> 00:06:45,161 Nirmal: It's writing text. 141 00:06:45,161 --> 00:06:46,781 And I can, I'll double you a little button. 142 00:06:46,781 --> 00:06:47,621 I copy that. 143 00:06:47,621 --> 00:06:50,351 Put it in, or there'll be, if you're using Cursor, all these other tools, 144 00:06:50,351 --> 00:06:52,891 you can put it into some TF file. 145 00:06:52,891 --> 00:06:52,951 Yeah. 146 00:06:53,491 --> 00:06:57,001 I can then take that and I can ask the LM what's the command that I 147 00:06:57,001 --> 00:07:00,121 need to run to apply this Terraform? 148 00:07:00,578 --> 00:07:04,876 To actually stand up the, what it's, what's described in this terraform. 149 00:07:05,086 --> 00:07:07,786 It'll spit out, okay, you wanna do Terraform plan and then 150 00:07:07,786 --> 00:07:09,016 Terraform apply and all that. 151 00:07:09,016 --> 00:07:12,731 Terraform in it or whatever, and I'll just copy those commands and 152 00:07:12,761 --> 00:07:14,621 check 'em and write them myself. 153 00:07:15,671 --> 00:07:20,171 So the LLM is not executing anything on my behalf. 154 00:07:20,171 --> 00:07:20,666 On, on your behalf. 155 00:07:21,031 --> 00:07:25,051 Agent would be defining a tool set. 156 00:07:25,111 --> 00:07:30,031 So I could give, I could define a tool called Terraform or a tool 157 00:07:30,031 --> 00:07:35,881 called Shell I could describe what that tool does in natural language. 158 00:07:36,181 --> 00:07:36,421 Bret: Okay. 159 00:07:36,721 --> 00:07:41,581 Nirmal: And then I can give the LLM system a list of these 160 00:07:41,581 --> 00:07:43,111 tools and their descriptions. 161 00:07:43,411 --> 00:07:44,851 And tell it. 162 00:07:45,091 --> 00:07:45,751 Okay? 163 00:07:46,681 --> 00:07:47,731 Back to the same scenario. 164 00:07:47,731 --> 00:07:51,151 I'm a platform engineer and I want to create an EKS production cluster 165 00:07:51,151 --> 00:07:56,199 using Terraform, and I want you to create it right for me because 166 00:07:56,199 --> 00:07:58,149 it has the access to those tools. 167 00:07:58,149 --> 00:08:04,089 Now it internal reasons, okay, I need to create some Terraform. 168 00:08:04,389 --> 00:08:07,089 I need to validate it in some kind of way, and then I need. 169 00:08:07,639 --> 00:08:09,199 I need to execute this Terraform. 170 00:08:09,769 --> 00:08:11,629 Is there any tools that I have in my toolbox 171 00:08:11,629 --> 00:08:13,759 Bret: In this case, sorry the i is the, you're referring 172 00:08:13,759 --> 00:08:14,809 to yourself as the ai, right? 173 00:08:15,169 --> 00:08:15,319 Yeah. 174 00:08:15,319 --> 00:08:15,499 Sorry. 175 00:08:16,344 --> 00:08:17,809 It's no longer the human doing this, right? 176 00:08:17,809 --> 00:08:17,893 No. 177 00:08:17,893 --> 00:08:19,303 We gave it instructions and we sit back 178 00:08:19,843 --> 00:08:22,277 Nirmal: from the perspective, from the perspective of the, LLM the 179 00:08:22,277 --> 00:08:27,377 Gen AI tool itself, the LLM system that's the I in this scenario. 180 00:08:27,377 --> 00:08:27,378 Yeah. 181 00:08:28,277 --> 00:08:30,137 I, the LLM is deciding. 182 00:08:31,007 --> 00:08:38,267 The Gen NI tool is looking at its list of available tools and matching what it needs 183 00:08:38,267 --> 00:08:44,327 to it, figure it, it's reasoning about what the end goal is and it looks and 184 00:08:44,327 --> 00:08:48,437 says, there's this tool called Terraform that allows me to use infrastructure as 185 00:08:48,437 --> 00:08:51,167 code to deploy resources on the cloud. 186 00:08:51,677 --> 00:08:52,847 That sounds like what I need. 187 00:08:52,997 --> 00:08:53,507 Maybe. 188 00:08:53,647 --> 00:08:55,177 And it. 189 00:08:55,867 --> 00:08:58,567 Generates the terraform just like it did the first time around. 190 00:08:59,137 --> 00:09:02,017 It knows what command to run. 191 00:09:02,017 --> 00:09:07,067 It generates the command and then the magic here, a little box will 192 00:09:07,067 --> 00:09:10,907 show up and says, do you want me to execute this on your behalf? 193 00:09:11,117 --> 00:09:14,897 You click the button, you click the button, and then it executes that 194 00:09:14,927 --> 00:09:24,107 Terraform apply Uhhuh and it sounds very simple, but it's a very different paradigm 195 00:09:24,107 --> 00:09:29,517 in terms of thinking about how we interact with infrastructure or systems in general. 196 00:09:29,517 --> 00:09:31,227 Like broadly systems in general. 197 00:09:31,257 --> 00:09:35,247 Because we are no, like in this way of looking at it or thinking 198 00:09:35,247 --> 00:09:42,057 about it, I, as the human, are no longer executing those commands. 199 00:09:42,182 --> 00:09:42,272 I am. 200 00:09:42,302 --> 00:09:47,162 Trusting to a certain extent that the LLM can figure out what it needs 201 00:09:47,162 --> 00:09:53,752 to do and giving it a guardrail set of tools to use and execute. 202 00:09:54,052 --> 00:09:54,502 Bret: Yeah. 203 00:09:54,562 --> 00:09:57,212 And so we're giving the, we're giving the Chaos monkey XI 204 00:09:57,212 --> 00:09:58,172 mean, it's automation, right? 205 00:09:58,172 --> 00:10:00,722 We could actually classify this as just automation. 206 00:10:00,722 --> 00:10:01,622 It just happens to be. 207 00:10:02,357 --> 00:10:04,547 Figuring out what to automate in real time. 208 00:10:04,787 --> 00:10:08,657 Rather than the traditional automation where we have a very deterministic plan 209 00:10:08,657 --> 00:10:12,594 of, steps that are repeated over and over again by a GitHub action runner 210 00:10:12,624 --> 00:10:14,544 or a CI CD platform or something. 211 00:10:14,754 --> 00:10:14,844 Yeah. 212 00:10:14,844 --> 00:10:20,544 Nirmal: And the agent part is the piece of software that enables. 213 00:10:21,249 --> 00:10:22,359 The LLM to execute. 214 00:10:22,659 --> 00:10:22,929 Bret: Yeah. 215 00:10:22,979 --> 00:10:27,409 Nirmal: and pull, pulls this all together and one, so back to what I was 216 00:10:27,409 --> 00:10:30,949 talking about with the infrastructure and there was a part where I said, 217 00:10:30,979 --> 00:10:37,939 okay, how do we define what tools are available for the agent system to use? 218 00:10:37,939 --> 00:10:42,761 and how do I want the agent to call those tools? 219 00:10:43,946 --> 00:10:46,826 And reason about them, and there's a protocol called 220 00:10:46,826 --> 00:10:48,896 MCP Model Context Protocol. 221 00:10:48,996 --> 00:10:54,306 Just outlining a standard way of defining the tools, the system prompt 222 00:10:54,796 --> 00:10:56,176 for that tool and a description. 223 00:10:56,476 --> 00:10:58,836 Bret: And this is like an API where you like define the spec of an API. 224 00:10:58,836 --> 00:11:04,081 Nirmal: It's a defined spec of an API and the adoption of that API is 225 00:11:04,081 --> 00:11:05,101 Bret: just exploding right now, 226 00:11:05,221 --> 00:11:05,731 Nirmal: essentially. 227 00:11:05,821 --> 00:11:05,881 Bret: Yeah. 228 00:11:06,031 --> 00:11:08,971 So we're to, to under if you're not, okay sorry, lemme back up a second. 229 00:11:09,211 --> 00:11:12,451 That's a very valid point because that's the reason I wanted to record This's 230 00:11:12,451 --> 00:11:14,301 a I don't wanna be a hype machine. 231 00:11:14,631 --> 00:11:14,781 Correct. 232 00:11:14,781 --> 00:11:16,791 But I'm super excited right now. 233 00:11:16,951 --> 00:11:21,391 if you can see inside my, in my enthusiastic brain, I've 234 00:11:21,391 --> 00:11:24,391 only been paying attention to this for a little over a month. 235 00:11:24,706 --> 00:11:27,046 If you asked me two months ago what an AI agent was, I'd say, 236 00:11:27,286 --> 00:11:29,266 I don't know a robot that's ai. 237 00:11:29,296 --> 00:11:29,791 I don't know. 238 00:11:30,436 --> 00:11:32,716 I now think I've got a much better handle on this. 239 00:11:32,791 --> 00:11:36,556 I've been spending so much of my life right now, deep diving into this, to 240 00:11:36,556 --> 00:11:39,586 the point that you and I are talking about changing some of the focus 241 00:11:39,586 --> 00:11:41,026 this year on, on all these topics. 242 00:11:41,026 --> 00:11:41,386 Absolutely. 243 00:11:41,386 --> 00:11:44,056 Because I think this is gonna dominate the conversation. 244 00:11:44,356 --> 00:11:46,846 This is, these are, there's gonna be a lot of predictions in this and we're 245 00:11:46,846 --> 00:11:49,876 not gonna talk forever 'cause it's gonna need to be multiple episodes to 246 00:11:49,876 --> 00:11:51,556 really break down what's going on here. 247 00:11:51,556 --> 00:11:52,906 But we now have the definitions. 248 00:11:53,156 --> 00:11:55,166 AI agents, what are tools? 249 00:11:55,586 --> 00:11:58,196 The protocol behind it is essentially MCP right now. 250 00:11:58,196 --> 00:12:01,406 Although that's not necessarily gonna be the only thing, it's just the thing right 251 00:12:01,406 --> 00:12:03,896 now that we're agreeing on by one company. 252 00:12:03,956 --> 00:12:04,466 Exactly. 253 00:12:04,766 --> 00:12:10,626 Nirmal: We have to caveat this with, this is like this is early like Docker days. 254 00:12:10,656 --> 00:12:11,166 This is like 255 00:12:11,466 --> 00:12:14,016 Bret: Docker in day 60, right? 256 00:12:14,016 --> 00:12:14,106 Yes. 257 00:12:14,106 --> 00:12:17,726 Like we were like right after Python in 2013 when we gave that 258 00:12:17,726 --> 00:12:19,136 de, when he gave that demo, Solomon. 259 00:12:19,616 --> 00:12:23,966 Like we all saw it and didn't understand it fully, but it 260 00:12:23,966 --> 00:12:25,316 felt like something right. 261 00:12:25,316 --> 00:12:28,886 And like you and I both, that's why we were early docker captains, is 262 00:12:28,886 --> 00:12:31,586 we saw that as a platform shift. 263 00:12:31,804 --> 00:12:35,294 we've seen these waves before over, over our careers of many decades 264 00:12:35,484 --> 00:12:40,064 that we earned with this gray beard status with effort and toil. 265 00:12:40,364 --> 00:12:43,814 And I feel like this is maybe the moment. 266 00:12:44,624 --> 00:12:48,794 That was the moment of 2013 and that, and yeah, I'm not alone in that feeling. 267 00:12:48,854 --> 00:12:49,134 yes. 268 00:12:49,134 --> 00:12:52,674 Nirmal: And there's just to be clear, there's massive differences between 269 00:12:52,794 --> 00:12:57,024 like paradigm shifts in terms of like virtualization, cloud containers. 270 00:12:57,104 --> 00:13:02,004 And the tooling of software development and systems development 271 00:13:02,004 --> 00:13:05,964 and right systems operations, it's still in that same vein, but. 272 00:13:06,324 --> 00:13:07,154 Yeah, we're not replacing, 273 00:13:07,154 --> 00:13:09,944 Bret: this is not replacing infrastructure or containers or anything like that. 274 00:13:09,994 --> 00:13:11,914 This is just gonna change the way we work. 275 00:13:12,214 --> 00:13:12,634 Nirmal: Correct. 276 00:13:12,634 --> 00:13:15,724 And also it's broader than just like IT infrastructure. 277 00:13:15,854 --> 00:13:20,624 Like this has implications with software design or application, 278 00:13:20,624 --> 00:13:22,154 like what an application does. 279 00:13:22,254 --> 00:13:25,344 And I want to think of this as a teaser trailer. 280 00:13:25,659 --> 00:13:27,644 To subsequent new series, episode. 281 00:13:27,644 --> 00:13:27,766 A new series. 282 00:13:27,771 --> 00:13:28,449 Yeah, absolutely. 283 00:13:28,479 --> 00:13:28,779 We're gonna have to 284 00:13:28,779 --> 00:13:29,589 Bret: come up with a name. 285 00:13:29,589 --> 00:13:32,739 I'm toying around with the idea of Agentic DevOps, and just classifying 286 00:13:32,739 --> 00:13:36,709 that as the absolutely as the theme of certain levels of podcast episodes. 287 00:13:36,709 --> 00:13:37,760 You've heard it here first. 288 00:13:37,760 --> 00:13:37,904 Heard it here first. 289 00:13:37,939 --> 00:13:38,139 This 290 00:13:38,169 --> 00:13:39,439 Nirmal: is Agentic DevOps. 291 00:13:39,509 --> 00:13:41,999 Another term we're seeing is AI four ops. 292 00:13:42,299 --> 00:13:43,559 Again, this is early days. 293 00:13:43,589 --> 00:13:44,549 None of this is like 294 00:13:44,609 --> 00:13:44,789 Bret: Yeah. 295 00:13:44,789 --> 00:13:45,569 Set in stone at all. 296 00:13:45,569 --> 00:13:48,839 Yeah, and if you're at KU Con today with us, if you were here at this 297 00:13:48,839 --> 00:13:52,019 conference all week, AI was a constant topic, but it wasn't about this. 298 00:13:52,394 --> 00:13:57,164 It actually, there was only one talk in an entire week that even touched on the 299 00:13:57,164 --> 00:14:03,794 idea of using AI to do the job of an DevOps or operator or platform engineer. 300 00:14:03,854 --> 00:14:06,464 Like people are, what we're talking about at KU Con for the last three 301 00:14:06,464 --> 00:14:10,664 years has been how to run the inference and build the LLM models. 302 00:14:11,054 --> 00:14:15,164 And so we are just still using human effort to do that work. 303 00:14:15,359 --> 00:14:19,514 But this, I feel like I'm gonna draw the line in the sand and say, this is the. 304 00:14:19,974 --> 00:14:24,804 month or the definitely the year, that kicks off. 305 00:14:25,274 --> 00:14:30,214 What will be a multi-year effort of figuring out how we use automated 306 00:14:30,214 --> 00:14:33,874 LLMs Essentially with access to all the tools we want to give it with 307 00:14:33,904 --> 00:14:36,284 the proper permissions and only the permissions we want to give 308 00:14:36,284 --> 00:14:39,074 it right to do our work for us. 309 00:14:39,479 --> 00:14:42,809 In a less chaos mon monkey way, right? 310 00:14:42,809 --> 00:14:44,009 Like less chaotic way. 311 00:14:44,014 --> 00:14:44,204 Potentially. 312 00:14:44,279 --> 00:14:44,639 Potentially. 313 00:14:44,729 --> 00:14:46,709 It could, this thing can easily go off the rails. 314 00:14:46,799 --> 00:14:47,189 Absolutely. 315 00:14:47,189 --> 00:14:51,959 I will probably reference in the show notes Solomon Hike's recent talks about 316 00:14:51,969 --> 00:14:56,329 how they're now using Dagger, which is primarily A-C-I-C-D pipeline tool. 317 00:14:56,379 --> 00:15:01,179 So he's talking, and a lot of my language is actually from him iterating 318 00:15:01,179 --> 00:15:04,899 on his idea of what this might look like when we're throwing a bunch 319 00:15:04,899 --> 00:15:10,599 of crazy hallucinating AI into what we consider a deterministic world. 320 00:15:10,969 --> 00:15:11,219 Correct. 321 00:15:11,219 --> 00:15:17,399 Nirmal: I think with containers and cloud and on the infrastructure APIs we have. 322 00:15:17,894 --> 00:15:22,694 We were chipping away and really aiming at deterministic behavior 323 00:15:22,694 --> 00:15:24,566 with respect to infrastructure. 324 00:15:26,434 --> 00:15:28,646 Ironically, maybe not ironically, I don't know. 325 00:15:29,006 --> 00:15:33,776 Now we're introducing a paradigm shift that reintroduces a lot 326 00:15:33,776 --> 00:15:35,971 of non-determinism right into. 327 00:15:36,656 --> 00:15:40,376 A place that we have been fighting to non-determinism for a long time. 328 00:15:41,066 --> 00:15:43,376 Bret: We have been working to get rid of all that. 329 00:15:43,376 --> 00:15:46,226 And now we're, that's why I keep saying Chaos monkey, because we're 330 00:15:46,226 --> 00:15:47,756 throwing a wrench into the system. 331 00:15:47,766 --> 00:15:52,056 That is in some ways feels like we're going back to a world of, I don't 332 00:15:52,056 --> 00:15:53,706 know, what's the status of the system? 333 00:15:53,706 --> 00:15:54,036 I don't know. 334 00:15:54,706 --> 00:15:57,896 and this will probably be another episode, I feel like this Agentic 335 00:15:57,896 --> 00:16:00,896 approach where we're actually can have the potential to pit. 336 00:16:01,271 --> 00:16:03,341 The LLMs against each other, right? 337 00:16:03,341 --> 00:16:05,231 And have different personas of these agents. 338 00:16:05,381 --> 00:16:07,901 One is the validator, one is the tester. 339 00:16:08,051 --> 00:16:10,121 One is one is the builder. 340 00:16:10,181 --> 00:16:11,771 And they can fight amongst each other. 341 00:16:11,861 --> 00:16:12,701 And it all works out. 342 00:16:12,701 --> 00:16:15,011 It actually ha happens to actually work out better. 343 00:16:15,251 --> 00:16:19,091 And so if you're like me and for the last three years of 344 00:16:19,151 --> 00:16:21,041 understanding, ever since GPT. 345 00:16:21,551 --> 00:16:22,961 3.5 or whatever came out. 346 00:16:22,961 --> 00:16:27,671 We all saw chat GPT as a product, and then we started with GoodHub copilot and 347 00:16:27,721 --> 00:16:32,571 we started down this road As a DevOps person, I haven't had a lot to talk about 348 00:16:32,811 --> 00:16:36,531 because I'm not interested in which model is the fastest or the most accurate. 349 00:16:36,531 --> 00:16:37,191 'cause you know what? 350 00:16:37,321 --> 00:16:41,501 they all hallucinate and still even today, years later. 351 00:16:42,491 --> 00:16:45,791 Code agents and we and you can see this on YouTube, you watch basically 352 00:16:45,791 --> 00:16:49,421 thousands of videos on YouTube of people trying to use these models to 353 00:16:49,421 --> 00:16:51,911 write perfect code and they just don't. 354 00:16:52,401 --> 00:16:56,151 And so we in ops, but we look at that, I think, and the people I 355 00:16:56,151 --> 00:17:00,201 talk to even for years now are like, we're never gonna use that for ops. 356 00:17:00,241 --> 00:17:03,061 But now my opinion has changed. 357 00:17:03,121 --> 00:17:03,271 Yeah. 358 00:17:03,621 --> 00:17:03,981 Nirmal: yeah. 359 00:17:03,981 --> 00:17:07,981 And I. If you're listening to this and your gut reaction is, wait we 360 00:17:07,981 --> 00:17:10,291 have like APIs that are deterministic. 361 00:17:10,291 --> 00:17:10,891 Like you just 362 00:17:11,221 --> 00:17:11,461 Bret: Yeah. 363 00:17:11,761 --> 00:17:12,871 Nirmal: We can just call an API. 364 00:17:12,871 --> 00:17:17,431 We can have an automation tool call an API to stand up infrastructure and like, 365 00:17:17,431 --> 00:17:23,041 why do we need to recreate like another layer that makes it non-deterministic. 366 00:17:23,041 --> 00:17:27,751 And looks like an API but isn't an API and you don't really know what it might 367 00:17:27,751 --> 00:17:30,481 do or which direction it might go. 368 00:17:30,841 --> 00:17:31,021 Yeah. 369 00:17:31,051 --> 00:17:32,261 And you're feeling I don't know. 370 00:17:32,261 --> 00:17:35,291 That doesn't seem like it would solve any problems for me. 371 00:17:35,291 --> 00:17:37,301 And it seems like it might introduce a lot of problems. 372 00:17:37,501 --> 00:17:39,811 You're in the right place because that's exactly what we're gonna explore. 373 00:17:40,111 --> 00:17:40,531 Bret: Yeah. 374 00:17:40,881 --> 00:17:44,181 Nirmal: one thing for sure though is it's here, right? 375 00:17:44,616 --> 00:17:51,206 I and so I feel like as good engineers, as good system admins and operators 376 00:17:51,686 --> 00:17:53,301 Bret: are we enjoy, we love our crafts. 377 00:17:53,301 --> 00:17:54,231 We, we look at this as an. 378 00:17:54,546 --> 00:17:57,446 Art form of brain power and Right. 379 00:17:57,546 --> 00:18:00,766 Reaching for perfectionism in our YAML and in our infrastructure 380 00:18:00,766 --> 00:18:02,716 optimization and our security. 381 00:18:02,936 --> 00:18:07,946 Nirmal: And we have a healthy sense of skepticism on new tools, 382 00:18:07,946 --> 00:18:09,746 new processes, new mechanisms. 383 00:18:09,746 --> 00:18:09,956 Yeah. 384 00:18:10,016 --> 00:18:13,946 When you, when availability of your services is paramount and reliability, 385 00:18:14,306 --> 00:18:16,796 you want to introduce new things in a. 386 00:18:17,186 --> 00:18:18,476 In a prudent manner. 387 00:18:18,576 --> 00:18:22,446 And so we're gonna take that approach, but we're not going 388 00:18:22,446 --> 00:18:24,936 to dismiss that this exists. 389 00:18:24,991 --> 00:18:30,431 Clearly there's a lot of interest, energy integration happening, 390 00:18:30,761 --> 00:18:35,781 experimentation happening and some people are already starting to see value. 391 00:18:36,021 --> 00:18:36,201 Yeah. 392 00:18:36,251 --> 00:18:39,175 and we're gonna explore with you where that, goes. 393 00:18:39,205 --> 00:18:39,415 Bret (2): Yeah. 394 00:18:39,415 --> 00:18:46,585 This, just to be clear, this is KubeCon April, 2025 and almost 395 00:18:46,585 --> 00:18:48,565 no one is talking about this yet. 396 00:18:48,845 --> 00:18:52,835 It feels like it's right under the surface of a lot of conversations and 397 00:18:52,835 --> 00:18:56,075 a lot of people maybe are thinking about it, but I'm not even sure that 398 00:18:56,075 --> 00:18:58,880 we're honest with ourselves around. 399 00:18:59,645 --> 00:19:02,135 That this is coming, whether we like it or not. 400 00:19:02,345 --> 00:19:09,455 And only because, yeah, not only, but one of the large reasons is business. 401 00:19:10,055 --> 00:19:10,385 Okay. 402 00:19:10,745 --> 00:19:11,255 Lemme back up. 403 00:19:11,255 --> 00:19:15,275 You know how in a lot of organizations, Kubernetes became a mandate, right? 404 00:19:15,275 --> 00:19:18,495 So there's lots of stories that came out over the course of Kubernetes 405 00:19:18,495 --> 00:19:22,425 lifetime of teams being told that they need to implement Kubernetes. 406 00:19:22,475 --> 00:19:27,545 It didn't come from a systems engineering approach of solving a known problem. 407 00:19:27,545 --> 00:19:28,595 It came down. 408 00:19:28,880 --> 00:19:33,390 Because an executive decided that they read a CIO magazine article 409 00:19:33,390 --> 00:19:35,580 that said Kubernetes was a cool new thing and they did it right. 410 00:19:35,730 --> 00:19:37,020 I hear this all the time. 411 00:19:37,020 --> 00:19:41,280 I confirm this multiple times this week with other people, and I now feel 412 00:19:41,280 --> 00:19:43,770 like we're not talking about it yet. 413 00:19:44,190 --> 00:19:51,300 But I did hear multiple analysts say their organizations that they're working with 414 00:19:51,330 --> 00:19:57,480 expect that we are going to reduce the number of personnel in infrastructure. 415 00:19:57,780 --> 00:19:58,830 Because of ai. 416 00:19:58,955 --> 00:20:02,315 the only way that's possible is if we use agents to our 417 00:20:02,315 --> 00:20:04,295 advantage, because we can't, yeah. 418 00:20:04,295 --> 00:20:06,425 I still don't believe we're replacing ourselves. 419 00:20:06,755 --> 00:20:09,735 I don't think the agents will ever in, in the near term. 420 00:20:09,735 --> 00:20:12,955 And as far as we can see out, let's say five years they will, they 421 00:20:12,955 --> 00:20:16,255 won't be running all infrastructure in the world by themselves. 422 00:20:16,375 --> 00:20:17,815 They can't turn on servers. 423 00:20:17,995 --> 00:20:22,095 They maybe you can actually pixie boot and do a power on a POE or whatever, but. 424 00:20:22,845 --> 00:20:27,405 Like we still need someone to give them orders and rules and guidelines to go 425 00:20:27,405 --> 00:20:31,695 do the work, but to me, I'm starting to wonder if very quickly, especially 426 00:20:31,695 --> 00:20:35,475 for those bleeding organizations that are looking to squeeze out every cost 427 00:20:35,475 --> 00:20:40,425 optimization they can of their staff, that they're going to be mandated to 428 00:20:40,605 --> 00:20:46,785 not just take AI as a code gen for yaml, but to start using these agents to. 429 00:20:47,280 --> 00:20:51,375 Increase the velocity of their work . And my, one of my stories is 430 00:20:51,375 --> 00:20:55,185 over the last 30 years I do this in talks is every major shift has been 431 00:20:55,185 --> 00:20:57,675 about speed, cost reduction in speed. 432 00:20:57,915 --> 00:20:59,835 Sometimes we get 'em both at the same time. 433 00:20:59,895 --> 00:21:01,275 Sometimes they're one or the other. 434 00:21:01,275 --> 00:21:03,435 We get a cost reduction, but we don't go any faster, which is 435 00:21:03,435 --> 00:21:06,855 fine, or we're going faster, but it's not necessarily cheaper yet. 436 00:21:06,855 --> 00:21:07,215 Nirmal: Right. 437 00:21:07,515 --> 00:21:07,755 Bret: And. 438 00:21:09,060 --> 00:21:13,020 I feel like this is maybe the next one where We're gonna be feeling the 439 00:21:13,020 --> 00:21:17,010 pressure because all the devs are gonna be writing code with ai, which 440 00:21:17,010 --> 00:21:21,150 in theory is going to improve their performance, which means they're writing 441 00:21:21,150 --> 00:21:24,390 more code, shipping more, or need, or wanting to ship more code, potentially. 442 00:21:24,390 --> 00:21:27,300 And if we're not using AI ourselves. 443 00:21:27,600 --> 00:21:32,250 To automate more of these platform designs, platform build outs, 444 00:21:32,310 --> 00:21:35,430 troubleshooting when we're in production and things are problematic and we 445 00:21:35,430 --> 00:21:38,050 don't wanna spend three hours trying to find the source of the problem. 446 00:21:38,290 --> 00:21:43,510 If we're not starting to use agents to, to automate a lot of that and reduce the 447 00:21:43,510 --> 00:21:48,550 time to market, so to speak, for a certain feature or platform feature then I don't 448 00:21:48,550 --> 00:21:52,860 think these teams are gonna hire more of us to help enable the devs to deploy. 449 00:21:53,190 --> 00:21:57,180 What it could end up happening is we end up more with more shadow ops, where 450 00:21:57,180 --> 00:22:01,020 the developers are so fed up with us not speeding up to the, if they're 451 00:22:01,020 --> 00:22:03,210 gonna go 10 x we have to go 10 x. Yeah. 452 00:22:03,260 --> 00:22:05,990 If they're gonna go three x or whatever the number ends up being in the reports. 453 00:22:05,990 --> 00:22:09,255 And Gartner puts out like the AI makes it efficient, more efficient 454 00:22:09,255 --> 00:22:11,235 for developers to, to code with ai. 455 00:22:11,235 --> 00:22:13,725 And the models get better and the way they use it is better. 456 00:22:14,115 --> 00:22:17,575 And so they're shipping code faster and they can do the same speed with three 457 00:22:17,575 --> 00:22:19,383 times less developers, or they can just. 458 00:22:19,825 --> 00:22:22,993 Produce three times more work, which I think is more likely, because if it's 459 00:22:22,993 --> 00:22:25,903 the common denominator and everyone has it, then that means every company 460 00:22:25,903 --> 00:22:29,323 can execute faster and they're gonna, they're gonna want to do that because 461 00:22:29,323 --> 00:22:30,433 their competitors are doing that. 462 00:22:30,433 --> 00:22:33,193 So that's a's, that's a very loaded and long prediction. 463 00:22:34,063 --> 00:22:35,173 Nirmal: That's a hypothesis. 464 00:22:35,663 --> 00:22:36,753 It's, I think there's a lot of predict here. 465 00:22:36,753 --> 00:22:40,473 It's gonna take some time for us to even chip away at that hypothesis, 466 00:22:40,473 --> 00:22:42,033 but it's a good starting point. 467 00:22:42,133 --> 00:22:47,593 If we're, but assuming that is like the hypothesis that organizations 468 00:22:47,593 --> 00:22:51,103 are looking at to adopt these tools that's a great starting point 469 00:22:51,103 --> 00:22:53,773 for us to help you figure out. 470 00:22:54,283 --> 00:22:57,133 what they are, why they are, what they do. 471 00:22:57,163 --> 00:22:57,313 Yeah. 472 00:22:57,313 --> 00:22:58,063 And how to use them. 473 00:22:58,363 --> 00:23:01,093 Bret: This is this, by the way, a lot a little bit of that opinion 474 00:23:01,093 --> 00:23:03,553 of mine, and there's more to come 'cause I've got a lot more written 475 00:23:03,553 --> 00:23:04,423 down than we're never gonna get to. 476 00:23:04,843 --> 00:23:09,443 But a significant portion of that is actually coming from what I've learned 477 00:23:09,443 --> 00:23:14,543 this week from analyst whose job it is to figure this stuff out for their 478 00:23:14,543 --> 00:23:16,013 organization and their customers. 479 00:23:16,013 --> 00:23:16,373 Interesting. 480 00:23:16,553 --> 00:23:20,693 And so I, I am a little weighted by their. 481 00:23:21,698 --> 00:23:25,508 Almost unrealistic expectations of how fast we can do this. 482 00:23:25,508 --> 00:23:26,678 'cause we are still humans. 483 00:23:26,838 --> 00:23:30,078 An organization can't adopt AI until the humans learn how to adopt AI and 484 00:23:30,078 --> 00:23:31,728 the humans have to go at human speed. 485 00:23:31,808 --> 00:23:34,928 So we can't just flip a switch and suddenly AI is here and 486 00:23:34,928 --> 00:23:35,888 running everything for us. 487 00:23:35,888 --> 00:23:38,978 At least not until we have Iron Man's Jarvis. 488 00:23:39,008 --> 00:23:39,488 Or whatever. 489 00:23:39,488 --> 00:23:42,908 Like until we have that, we still have to learn these tools and still have 490 00:23:42,908 --> 00:23:44,528 to adapt our platforms to use them. 491 00:23:44,528 --> 00:23:44,529 Yes. 492 00:23:44,534 --> 00:23:46,088 And adapt our learning to use them. 493 00:23:46,448 --> 00:23:47,738 And that's gonna take some time 494 00:23:47,798 --> 00:23:48,308 Nirmal: and. 495 00:23:48,663 --> 00:23:51,633 I'd like to, like the parting thought for this is Okay. 496 00:23:51,633 --> 00:23:56,193 And here, like you said, there's an under the surface kind of thing happening. 497 00:23:56,223 --> 00:23:56,583 Yeah. 498 00:23:56,943 --> 00:23:57,783 So whispers, 499 00:23:57,783 --> 00:23:58,983 Bret: it's almost like murmurs and under 500 00:23:58,983 --> 00:23:59,583 Nirmal: the surface. 501 00:23:59,588 --> 00:23:59,788 Yeah. 502 00:24:00,088 --> 00:24:02,793 AI agent, AI agents, mag 503 00:24:03,093 --> 00:24:03,423 Bret: DevOps. 504 00:24:03,453 --> 00:24:03,843 Ooh. 505 00:24:04,383 --> 00:24:05,583 This is our ASMR podcast. 506 00:24:05,583 --> 00:24:06,333 Moment of the podcast. 507 00:24:08,288 --> 00:24:09,518 Nirmal: Like MCP protocol. 508 00:24:09,818 --> 00:24:10,148 Bret: Yeah. 509 00:24:10,488 --> 00:24:14,448 Nirmal: you mentioned HA proxy on the previous podcast, about load 510 00:24:14,448 --> 00:24:18,408 balancing and figuring out the street, like token utilization of 511 00:24:18,408 --> 00:24:20,958 GPUs and tokens and all that stuff. 512 00:24:21,208 --> 00:24:25,318 and we had a conversation at the solo booth and they were talking about having. 513 00:24:25,778 --> 00:24:30,008 A proxy for an MCP gateway, one of the things that we're seeing the early 514 00:24:30,008 --> 00:24:32,688 signs of is these new workloads, right? 515 00:24:32,688 --> 00:24:39,268 This agentic kind of thinking Around even just executing the agentic platform, 516 00:24:39,268 --> 00:24:43,678 if you will, And everything from looking at the tokens and optimizing 517 00:24:43,678 --> 00:24:50,268 load balancing to inference endpoints or MCP is, doesn't behave the same 518 00:24:50,268 --> 00:24:52,398 way as like just an http connection. 519 00:24:52,498 --> 00:24:53,158 Necessarily. 520 00:24:53,398 --> 00:24:53,968 And solo. 521 00:24:53,968 --> 00:24:56,368 We were talking to them and they have an MCP gateway. 522 00:24:56,648 --> 00:24:59,198 We're seeing a little bit more of a trend on AI gateways. 523 00:24:59,198 --> 00:25:04,598 Is DO the project has an AI gateway and so this is not just another workload 524 00:25:04,778 --> 00:25:06,368 and looks like just a web server. 525 00:25:06,417 --> 00:25:09,117 And the networking and everything is gonna be different. 526 00:25:09,197 --> 00:25:12,497 Not dramatically different, but We'll, but drift different 527 00:25:12,497 --> 00:25:14,177 enough that we need to be aware. 528 00:25:14,777 --> 00:25:18,377 'cause even if you're not using any of these tools, someone in your 529 00:25:18,377 --> 00:25:21,077 organization is probably gonna say, oh, we need to integrate this stuff 530 00:25:21,107 --> 00:25:23,567 into our software, to our right. 531 00:25:23,567 --> 00:25:24,947 Whatever we're delivering. 532 00:25:25,427 --> 00:25:27,407 And we'll need to know it even at that layer. 533 00:25:27,467 --> 00:25:30,887 So we're gonna also cover that component as it relates to. 534 00:25:31,427 --> 00:25:32,897 The Kubernetes ecosystem, right? 535 00:25:32,912 --> 00:25:33,542 And cloud native. 536 00:25:33,842 --> 00:25:34,202 Bret: Yeah. 537 00:25:34,322 --> 00:25:38,172 I think this, if we had to do like an elevator pitch for this podcast, it would 538 00:25:38,172 --> 00:25:46,252 be we now have a industry idea around these terms agent, and then it uses an API 539 00:25:46,252 --> 00:25:50,568 called MCP to allow us to give more work. 540 00:25:50,898 --> 00:25:55,218 To these crazy robot texting things that we have to talk to in human 541 00:25:55,218 --> 00:25:56,898 language and not with code, right? 542 00:25:56,898 --> 00:25:58,878 It's running code, but we're not talking to it with code. 543 00:25:59,148 --> 00:26:04,238 And that it can now understand all the tools we need to use and we can just give 544 00:26:04,238 --> 00:26:05,558 it a list of everything I wanted to use. 545 00:26:05,700 --> 00:26:08,820 here's my Kubernetes API, here's all my other things that I, you have 546 00:26:08,820 --> 00:26:11,070 access to, and here's my problem. 547 00:26:11,400 --> 00:26:12,420 Go solve it. 548 00:26:12,570 --> 00:26:15,330 And that paradigm. 549 00:26:15,705 --> 00:26:19,125 Three months ago, two months ago for me, I didn't know existed. 550 00:26:19,695 --> 00:26:22,695 And that's why I've been sitting on the sidelines with ai. 551 00:26:22,695 --> 00:26:26,775 Like it's cool for writing programs that mostly work in a demo. 552 00:26:27,025 --> 00:26:30,235 It's cool for adding a feature to something I already have, but it's 553 00:26:30,235 --> 00:26:34,135 not doing my job as a platform engineer or DevOps engineer. 554 00:26:34,185 --> 00:26:36,075 It's just helping me write text faster 555 00:26:36,169 --> 00:26:37,459 Then I can type into my keyboard. 556 00:26:37,509 --> 00:26:39,129 And that was not that interesting. 557 00:26:39,129 --> 00:26:41,499 That's why you didn't see a lot of me talking about that on this show, 558 00:26:41,679 --> 00:26:42,609 was it just wasn't that interesting. 559 00:26:42,759 --> 00:26:48,369 This is an interesting topic for ops and for absolutely engineers on the platform. 560 00:26:48,669 --> 00:26:48,849 Nirmal: Yep. 561 00:26:49,164 --> 00:26:49,464 Bret: So 562 00:26:49,944 --> 00:26:50,544 Nirmal: stay tuned. 563 00:26:50,634 --> 00:26:51,174 Yeah. 564 00:26:51,174 --> 00:26:54,504 And I, I love crazy texting robots. 565 00:26:54,534 --> 00:26:54,894 Crazy 566 00:26:54,894 --> 00:26:55,824 Bret: texting robots. 567 00:26:55,954 --> 00:26:57,544 Maybe that's the title. 568 00:26:57,544 --> 00:26:58,024 TBD. 569 00:27:00,244 --> 00:27:00,784 Alright. 570 00:27:00,844 --> 00:27:01,354 Alright. 571 00:27:01,354 --> 00:27:02,104 See you soon, man. 572 00:27:02,704 --> 00:27:02,914 See 573 00:27:02,914 --> 00:27:03,094 Nirmal: you. 574 00:27:03,094 --> 00:27:03,096 See you. 575 00:27:03,754 --> 00:27:04,324 Bye. 576 00:27:04,324 --> 00:27:04,384 Bye.