1 00:00:00,360 --> 00:00:03,720 Fundamentally, the interfaces that we as software 2 00:00:03,720 --> 00:00:06,000 engineers use today aren't built for this. 3 00:00:06,210 --> 00:00:11,490 They're built to do one thing very deeply at a time writing code, right. 4 00:00:12,465 --> 00:00:14,655 Now our interfaces need to change, and the 5 00:00:14,655 --> 00:00:17,235 environments in which we do that work needs to change. 6 00:00:17,415 --> 00:00:19,635 My laptop is built to do one thing at a time. 7 00:00:20,085 --> 00:00:22,995 I mean, anyone who's tried to run different Python versions on 8 00:00:22,995 --> 00:00:27,225 one machine knows what I'm talking about, and so are my IDs. 9 00:00:27,225 --> 00:00:31,035 So my environments and my interfaces need to change. 10 00:00:31,695 --> 00:00:35,625 To get the productivity out of these agents, and that's very fundamentally 11 00:00:35,625 --> 00:00:40,425 what ona does is it gives you, ma, as many of these environments as 12 00:00:40,425 --> 00:00:43,935 you need perfectly set up for the task at hand and it gives you an 13 00:00:43,935 --> 00:00:48,915 interface that MA that helps you find flow and joy in doing more things. 14 00:00:49,440 --> 00:00:49,980 In parallel, 15 00:00:55,500 --> 00:00:57,270 welcome to Screaming in the Cloud. 16 00:00:57,330 --> 00:01:02,520 I'm Corey Quinn, and my guest today has been on the show before. 17 00:01:02,520 --> 00:01:08,070 Chris Weichel is the CTO of ona, which we have not spoken about on the show 18 00:01:08,070 --> 00:01:12,870 before, because once upon a time until recently, they were known as Git Pod. 19 00:01:13,260 --> 00:01:14,880 Chris, thank you for returning. 20 00:01:15,899 --> 00:01:17,130 Thank you for having me again. 21 00:01:17,670 --> 00:01:21,600 This episode is brought to you by ona, formerly GIT Pod. 22 00:01:21,690 --> 00:01:24,810 Are you tired of coding agents pushing your S3 bucket to the wrong 23 00:01:24,810 --> 00:01:28,410 AWS account or having them commit your entire downloads folder 24 00:01:28,410 --> 00:01:31,560 because they thought your tax documents were part of the code base? 25 00:01:31,650 --> 00:01:34,860 All while using privacy policies that basically say that your 26 00:01:34,860 --> 00:01:38,250 customer data deserves a nice vacation on random cloud servers. 27 00:01:38,565 --> 00:01:42,795 Introducing ona where your coding agents run in isolated sandboxes. 28 00:01:42,795 --> 00:01:44,925 Securely within your own VPC. 29 00:01:45,355 --> 00:01:49,335 ona lets you run agents at scale with workflows, bulk open 30 00:01:49,335 --> 00:01:52,365 requests that finally tackle that Java migration that you started 31 00:01:52,365 --> 00:01:56,055 in 2019 or automatically fix CDEs when your scans find them. 32 00:01:56,485 --> 00:01:58,695 ona also supports private AI models. 33 00:01:58,705 --> 00:02:02,664 Through Amazon Bedrock that your corporate overlords might even approve of. 34 00:02:02,995 --> 00:02:08,454 Head to ona, that's O a.com and get $200 of free credits using the 35 00:02:08,454 --> 00:02:12,234 code screaming in the cloud because your laptop wasn't designed 36 00:02:12,234 --> 00:02:15,910 to babysit over caffeinated rogue coding agents with root access. 37 00:02:17,240 --> 00:02:19,395 As, as you might have picked up from that intro. 38 00:02:19,395 --> 00:02:23,235 I, I have a leading question I would like to begin with. 39 00:02:23,565 --> 00:02:27,315 Um, GIT Pod was an interesting name of the company 40 00:02:27,315 --> 00:02:30,075 because it was, it was, oh, it was like GitHub. 41 00:02:30,585 --> 00:02:31,665 No, actually. 42 00:02:31,665 --> 00:02:36,795 And you sort of got to a point of understanding what it was, and now it's all 43 00:02:36,795 --> 00:02:39,795 that work we did to teach you what that word was and that it was pronounced with 44 00:02:39,795 --> 00:02:45,225 a hard G instead of a soft, well now we're changing it again to something else. 45 00:02:45,540 --> 00:02:45,870 Why? 46 00:02:46,500 --> 00:02:48,510 There's a, a number of reasons. 47 00:02:48,510 --> 00:02:53,940 The one is that GI Pod as a name really doesn't make sense anymore. 48 00:02:54,480 --> 00:02:57,540 One, we famously left Kubernetes, so the pod part 49 00:02:57,540 --> 00:03:02,340 is out and Git isn't at the center of it all. 50 00:03:02,340 --> 00:03:05,670 You know, it's, it's a very important piece of technology for sure, but so much 51 00:03:05,670 --> 00:03:11,010 of what he can do with GI part now owner isn't centered around Git anymore. 52 00:03:11,010 --> 00:03:13,350 So the name really has become a bit of a misnomer. 53 00:03:14,010 --> 00:03:17,250 And to be frank, the amount of times we've been confused for 54 00:03:17,250 --> 00:03:21,420 GitHub or GitLab or spelled with a capital P for no apparent 55 00:03:21,420 --> 00:03:24,510 reason, I'm just very glad we can leave all that behind us. 56 00:03:24,855 --> 00:03:26,295 So hence the, hence the rename. 57 00:03:26,355 --> 00:03:29,145 I am always a little bit leery of company 58 00:03:29,175 --> 00:03:31,665 renames, and that is in many ways unfair to you. 59 00:03:31,665 --> 00:03:36,345 The one that sticks out in my mind was Mesosphere after they 60 00:03:36,345 --> 00:03:41,145 renamed for 2D, two IQ, and I. Even now, I had to look that up to 61 00:03:41,145 --> 00:03:44,025 make sure I was getting those letters correctly, and it turns out 62 00:03:44,025 --> 00:03:48,765 that the correct name is now acquired by Nutanix, so, oh, okay. 63 00:03:49,005 --> 00:03:49,575 It's. 64 00:03:49,990 --> 00:03:52,930 Brand equity is super freaking hard. 65 00:03:52,960 --> 00:03:53,650 It is. 66 00:03:53,740 --> 00:03:56,830 It takes a long time to teach people things and okay, we're 67 00:03:56,830 --> 00:04:00,880 going to be changing our name, our logos, et cetera is hard. 68 00:04:00,955 --> 00:04:03,970 I, I saw that Facebook was able to do that to meta and 69 00:04:03,970 --> 00:04:06,460 I would've bet anything that, uh, that, that wouldn't 70 00:04:06,460 --> 00:04:08,530 have worked because it's been how many years since. 71 00:04:09,450 --> 00:04:12,870 Google did that with Alphabet and every time in a newspaper 72 00:04:12,870 --> 00:04:16,260 article to this day that we see anything about it, it is 73 00:04:16,260 --> 00:04:20,400 alphabet, parentheses, Google's parent company, close parent. 74 00:04:20,730 --> 00:04:24,060 It's, it's one of those things where sometimes it sticks, but 75 00:04:24,060 --> 00:04:27,240 usually it feels like it's going to have that parenthetical forever. 76 00:04:27,390 --> 00:04:28,045 What's your sense on this one? 77 00:04:28,830 --> 00:04:32,280 My take on this is as a company, if you wanna rename, if you're, 78 00:04:32,280 --> 00:04:34,620 if you're small enough, it doesn't matter because no one knows you. 79 00:04:34,650 --> 00:04:37,409 If you're big enough, it's, everyone's gonna hear about it. 80 00:04:37,530 --> 00:04:38,940 So, you know, it's fine if you do. 81 00:04:39,150 --> 00:04:40,290 And then there's sort of the trough in the 82 00:04:40,290 --> 00:04:42,990 middle where, um, it's a bit hit or miss. 83 00:04:43,350 --> 00:04:47,040 I think for us, the main reason we did that is because we're really at 84 00:04:47,040 --> 00:04:50,340 the precipice of a pretty fundamental change on how software is written. 85 00:04:50,980 --> 00:04:53,590 With that, like ONA isn't just a rename, it's 86 00:04:53,590 --> 00:04:56,110 really a refounding of what it is that we do. 87 00:04:56,110 --> 00:04:58,210 It isn't a pivot, you know, it's not like we're doing something 88 00:04:58,210 --> 00:05:01,750 else, but it marks a new chapter on this trajectory that 89 00:05:01,750 --> 00:05:04,300 we've been on since, since the inception of the company. 90 00:05:04,990 --> 00:05:08,830 And with that, we also want to be known for leading. 91 00:05:10,350 --> 00:05:12,870 Uh, leading where we're going as, as software engineering 92 00:05:12,870 --> 00:05:16,650 as a whole, and so the new name signifies that ambition. 93 00:05:16,920 --> 00:05:21,300 Normally I would discount this to be direct as, oh, well 94 00:05:21,300 --> 00:05:25,380 everything is changing about software engineering, is it though? 95 00:05:25,410 --> 00:05:29,430 But I've been beating code into submission for longer 96 00:05:29,435 --> 00:05:32,610 than is appropriate given how terrible my code still is. 97 00:05:33,000 --> 00:05:34,710 And I, I think that it is. 98 00:05:34,790 --> 00:05:38,120 It difficult to make the straight-faced assertion that nothing 99 00:05:38,120 --> 00:05:42,170 is different about writing code in 2025 than it was back in 2020. 100 00:05:42,560 --> 00:05:45,020 The world has foundationally changed. 101 00:05:45,200 --> 00:05:49,040 You can debate where AI is making inroads versus not, but one 102 00:05:49,040 --> 00:05:52,430 area in which it has excelled has been in code generation. 103 00:05:52,760 --> 00:05:53,510 Absolutely. 104 00:05:53,510 --> 00:05:53,960 And. 105 00:05:54,860 --> 00:05:56,995 The, the way we think about this is really, we've gone 106 00:05:56,995 --> 00:06:00,925 through three waves of how we write code, and the, the very 107 00:06:00,925 --> 00:06:03,835 first one is where we've essentially artisanally handcrafted 108 00:06:03,835 --> 00:06:06,775 every single line, say for code generation and auto complete. 109 00:06:07,225 --> 00:06:09,505 And this is how we've been writing code for really 110 00:06:09,505 --> 00:06:13,135 the longest time, certainly since I can remember and. 111 00:06:14,085 --> 00:06:18,465 A few years ago when AI first entered the, the scene, we started to have 112 00:06:18,465 --> 00:06:21,735 co coded as like copilot and the likes that gave us better auto complete. 113 00:06:22,455 --> 00:06:25,635 And they, you know, made the time shorter that it, they reduced the time 114 00:06:25,635 --> 00:06:28,845 that it took to write code, but they didn't fundamentally change the pattern. 115 00:06:29,115 --> 00:06:31,215 You know, it was still a human sitting there typing 116 00:06:31,215 --> 00:06:34,005 stuff and hitting tap tap every once in a while to get 117 00:06:34,005 --> 00:06:36,825 better code, or, I dunno, better, but more at least. 118 00:06:37,335 --> 00:06:37,815 And. 119 00:06:38,115 --> 00:06:39,315 Not too long ago. 120 00:06:39,585 --> 00:06:41,355 Essentially, agents entered the scene and they very 121 00:06:41,355 --> 00:06:44,055 fundamentally changed the pattern because now it's no 122 00:06:44,055 --> 00:06:47,175 longer humans writing code, but it's machines writing code. 123 00:06:47,415 --> 00:06:50,175 And to what extent and how much and how well that's all debatable. 124 00:06:50,175 --> 00:06:51,555 Like we can happy to talk about that. 125 00:06:51,855 --> 00:06:54,765 But certainly the truth of the matter is that now we have 126 00:06:54,765 --> 00:06:58,875 these things that can write and modify code, modified code for 127 00:06:58,875 --> 00:07:04,035 us at a level of abstraction that's arguably a level up from. 128 00:07:04,740 --> 00:07:06,840 The programming languages we've been using thus far. 129 00:07:06,960 --> 00:07:09,719 And so that's a very fundamental change in, in how software is being written. 130 00:07:09,719 --> 00:07:14,700 Not unlike, you know, changes from assembly code to higher level 131 00:07:14,700 --> 00:07:18,690 languages to see and the likes to object oriented languages to now. 132 00:07:19,245 --> 00:07:22,755 I mean, you know, it's almost a beaten sent sentence or 133 00:07:22,755 --> 00:07:25,005 beaten saying that English is a new programming language. 134 00:07:25,005 --> 00:07:25,995 I don't believe that to be true. 135 00:07:26,325 --> 00:07:29,385 That's not, that's not the thing, because we're bad at that one too. 136 00:07:30,045 --> 00:07:31,605 Yeah, exactly. 137 00:07:31,605 --> 00:07:33,915 But certainly we, um, I mean, me too, like 138 00:07:33,945 --> 00:07:36,405 the under specifications is a key problem. 139 00:07:36,750 --> 00:07:39,570 And that is still so, so I'm not, I'm not saying this is a 140 00:07:39,570 --> 00:07:41,760 new language, this is a new abstraction, but it is a way we 141 00:07:41,760 --> 00:07:44,730 communicate now about code that's a very fundamental to a machine. 142 00:07:45,030 --> 00:07:48,240 It's a very fundamentally different way how, how we interact with code. 143 00:07:48,390 --> 00:07:50,970 We, I, I keep observing that I don't know how to live in this 144 00:07:50,970 --> 00:07:53,520 current world that we're in because we spent enough money and 145 00:07:53,520 --> 00:07:55,950 made the computers expensive and powerful enough that they are 146 00:07:55,950 --> 00:07:59,280 simultaneously capable of doing what we mean instead of what we say. 147 00:07:59,665 --> 00:08:01,555 And are bad at math while they do it. 148 00:08:01,735 --> 00:08:06,355 So it's this, I, I don't fully understand this world I find myself in and 149 00:08:06,355 --> 00:08:09,145 I'm starting to wonder, does this mean that I've finally lived too long? 150 00:08:09,265 --> 00:08:12,895 And maybe other people would argue that I definitely have, but 151 00:08:12,895 --> 00:08:15,685 it's like I have young children and they, I, I like, how do I 152 00:08:15,685 --> 00:08:18,295 explain to them how computers work on a month to month basis? 153 00:08:18,295 --> 00:08:19,615 It's, it's shifting under me. 154 00:08:20,475 --> 00:08:22,425 It certainly moves very, very quickly. 155 00:08:22,425 --> 00:08:25,665 I mean, we're recording this at, at the time that we 156 00:08:25,665 --> 00:08:28,455 are recording this, literally Sauna 4.5 just dropped 157 00:08:29,205 --> 00:08:29,385 Yeah. 158 00:08:29,385 --> 00:08:32,715 Within the last hour of us whacking the record button, so 159 00:08:32,715 --> 00:08:34,845 we have no idea whether it's good, whether it's bad, who 160 00:08:34,845 --> 00:08:37,275 supports At the moment it's just anthropic out there alone. 161 00:08:37,275 --> 00:08:38,265 I'm sure All the Me too. 162 00:08:38,265 --> 00:08:39,255 We support this now. 163 00:08:39,255 --> 00:08:42,885 We'll compiling in as we literally speak, but 164 00:08:42,885 --> 00:08:45,135 it's, it is weird because state of the art. 165 00:08:45,145 --> 00:08:47,095 Is still moving rapidly. 166 00:08:47,095 --> 00:08:48,625 It's not the meteoric growth curve. 167 00:08:48,625 --> 00:08:49,824 It's been over the last couple of years. 168 00:08:49,824 --> 00:08:52,165 Things have slowed down now, but it is definitely 169 00:08:52,165 --> 00:08:54,355 still showing the ability to surprise us. 170 00:08:54,835 --> 00:08:55,555 Oh, absolutely. 171 00:08:55,555 --> 00:09:00,025 And you know, the, the half hour before this show, I literally had ONA ads. 172 00:09:00,365 --> 00:09:02,045 Sauna 4.5 support to itself. 173 00:09:02,735 --> 00:09:03,425 See, okay. 174 00:09:03,425 --> 00:09:06,425 The first product I've heard of supporting it is you Good work. 175 00:09:06,425 --> 00:09:08,015 Your timing is excellent. 176 00:09:08,345 --> 00:09:12,155 Now, I, I have to ask in a bit of confession of my own, we are 177 00:09:12,155 --> 00:09:15,185 in the process of renaming our company from the Duck Bill Group 178 00:09:15,185 --> 00:09:19,085 to Simply Duck Bill as we expand into a services offering as well 179 00:09:19,115 --> 00:09:22,535 as pure ser as to software offering as well as pure services. 180 00:09:22,805 --> 00:09:25,955 It, the group does not really carry the same weight and 181 00:09:25,955 --> 00:09:30,035 internally it is hard for us to, to, to correct ourselves after. 182 00:09:30,060 --> 00:09:32,790 Eight years of inertia of saying it the way that we have. 183 00:09:33,150 --> 00:09:36,209 Uh, so my two questions for you are, one, do you still find 184 00:09:36,209 --> 00:09:39,060 yourself referring to the company as GI Pod internally? 185 00:09:39,209 --> 00:09:43,680 And two, if I were to do a grep at a word count of the 186 00:09:43,680 --> 00:09:47,819 term GI pod in your code base, how many would I find? 187 00:09:49,800 --> 00:09:50,130 Okay. 188 00:09:50,130 --> 00:09:52,980 Do I still say get pot every once in a while? 189 00:09:54,480 --> 00:10:00,270 I do, but surprisingly little like I expected it to be a lot more The general 190 00:10:00,270 --> 00:10:05,910 save is GI pod now owner and then you carry on in terms of the word count. 191 00:10:05,940 --> 00:10:08,730 If we looked at the ratio of GI pod to owner, in our co 192 00:10:08,730 --> 00:10:12,660 base, it's orders of magnitudes, more GI pod than owner. 193 00:10:14,160 --> 00:10:15,090 We have a, yeah. 194 00:10:15,090 --> 00:10:18,300 Oh, we had an worldly working name of our product for the first three weeks. 195 00:10:18,300 --> 00:10:22,650 We were building it, and it is that, that legacy name is still in our code base 196 00:10:22,650 --> 00:10:28,290 because it, those, eh, fix that later naming decisions become load bearing. 197 00:10:28,410 --> 00:10:31,560 We don't think anything is gonna break if we just 198 00:10:31,590 --> 00:10:33,780 do a global find and replace at the same time. 199 00:10:35,175 --> 00:10:36,555 But it might. 200 00:10:36,555 --> 00:10:38,925 So that's a question of, okay, how, how much 201 00:10:38,925 --> 00:10:41,084 extra work do we wanna create for ourselves today? 202 00:10:41,265 --> 00:10:43,094 Mm. We're gonna keep kicking that can down the road. 203 00:10:43,094 --> 00:10:45,015 Surely this problem won't get worse with time. 204 00:10:45,435 --> 00:10:48,854 I mean, we, we have customers who obviously 205 00:10:48,854 --> 00:10:51,135 rely on our API and we're not gonna break them. 206 00:10:51,734 --> 00:10:55,454 You know, our API contracts, um, are wholly to us. 207 00:10:55,454 --> 00:10:56,444 We, we won't break them. 208 00:10:56,444 --> 00:11:00,135 So cl we'll, we will have GI Pot in our copays for all eternity. 209 00:11:01,094 --> 00:11:02,295 The ratio is gonna shift. 210 00:11:03,270 --> 00:11:03,510 Yeah. 211 00:11:03,990 --> 00:11:08,130 And it has to, uh, has the product itself changed significantly? 212 00:11:08,490 --> 00:11:11,610 That's, that's the other question because I find that shifting names 213 00:11:11,610 --> 00:11:15,690 is, if it's not an exactly an atomic operation, it's, it's pretty close. 214 00:11:15,690 --> 00:11:18,660 I mean, you only have one logo simultaneously in the upper left 215 00:11:18,660 --> 00:11:23,910 hand corner, but the product itself has to simultaneously serve 216 00:11:23,910 --> 00:11:28,110 the use case that it has been sold to solve for before, but also. 217 00:11:28,580 --> 00:11:30,710 Uh, pivoting to embrace new things. 218 00:11:30,920 --> 00:11:34,100 I, I will say I give you folks credit more so than I do. 219 00:11:34,100 --> 00:11:39,170 Most companies, uh, everyone now has slapped AI on the above the fold on 220 00:11:39,170 --> 00:11:43,490 their landing page and like, we are an AI company and have been for years. 221 00:11:43,580 --> 00:11:46,130 Funny 'cause I look back three years ago at your conference talks. 222 00:11:46,130 --> 00:11:48,440 I see no mention of it, but we'll let that slide. 223 00:11:48,680 --> 00:11:50,900 In your case, you, you have taken that deeper. 224 00:11:50,900 --> 00:11:53,570 You have renamed the company, you have. 225 00:11:53,630 --> 00:11:59,510 Made a public declaration that this is what we are about and whether it is the 226 00:11:59,510 --> 00:12:02,660 right path or the wrong path, no one can deny that you're committing to it. 227 00:12:03,890 --> 00:12:04,100 Yeah. 228 00:12:04,100 --> 00:12:06,560 The, you know, the thing that we've been building for. 229 00:12:07,305 --> 00:12:09,405 For a very long time now, it's essentially 230 00:12:10,064 --> 00:12:12,375 the automation of development environments. 231 00:12:12,375 --> 00:12:15,345 It's the ability to create a development environment with a click of a 232 00:12:15,345 --> 00:12:19,485 button, something that is incredibly useful for humans because it removes 233 00:12:19,485 --> 00:12:22,185 a lot of work and toil from setting up development environments and 234 00:12:22,185 --> 00:12:26,835 maintaining them five hours per week, uh, um, studies and, and data show. 235 00:12:27,765 --> 00:12:30,615 And that's very helpful for humans and it's existential for agents. 236 00:12:31,815 --> 00:12:36,495 If you want an agent to scale beyond your, beyond your machine, 237 00:12:36,495 --> 00:12:39,435 and you wanna run five of them in parallel, or even just avoid that 238 00:12:39,435 --> 00:12:44,115 agent accidentally sending an email to your boss having some unkind 239 00:12:44,115 --> 00:12:48,705 words or accessing production, because all of this happens to be 240 00:12:48,705 --> 00:12:52,125 aWeichel able on the same laptop you run your terminal agent in. 241 00:12:52,785 --> 00:12:54,825 If you want to avoid all that, you need to put them 242 00:12:54,825 --> 00:12:58,395 in isolated, readily set up development environments. 243 00:12:59,475 --> 00:13:00,314 You are not wrong. 244 00:13:00,344 --> 00:13:04,005 I, I have problems with cursor constantly because I have set up my 245 00:13:04,005 --> 00:13:09,375 ZSH prompt to reflect what I need as a human being editing the thing. 246 00:13:09,375 --> 00:13:12,495 It uses some power line nonsense and some other stuff as well, 247 00:13:12,495 --> 00:13:16,425 because I've had, you know, an afternoon to kill and I now in 248 00:13:16,425 --> 00:13:19,694 most, most, uh, terminal environments until it gets set up. 249 00:13:19,875 --> 00:13:22,005 It has glyphs that don't render properly. 250 00:13:22,185 --> 00:13:25,545 It has fonts that aren't present, and as a 251 00:13:25,545 --> 00:13:27,645 result, everything looks janky and broken. 252 00:13:28,175 --> 00:13:30,875 Most of these tools because I have, I have 253 00:13:30,875 --> 00:13:33,035 gotten my shell working for me as a human. 254 00:13:33,515 --> 00:13:35,075 Computers have not yet caught up to that. 255 00:13:35,795 --> 00:13:36,545 Absolutely. 256 00:13:36,875 --> 00:13:40,445 There's a reason why, you know, cloud calls, it dangerously skip permissions. 257 00:13:40,715 --> 00:13:42,935 If you wanted to give, if you want to give it 258 00:13:42,935 --> 00:13:45,665 a blanket check to do anything and everything. 259 00:13:46,485 --> 00:13:47,865 Yeah, I, I can't run that on my laptop. 260 00:13:47,865 --> 00:13:48,944 I have client data there. 261 00:13:48,944 --> 00:13:53,685 It is a hard stop, so I, I give it its own dedicated EC2 instance and 262 00:13:53,685 --> 00:13:58,545 for one side project in its own unbounded AWS account via instance role. 263 00:13:58,635 --> 00:14:00,854 So there's dangerous, and then there's whatever the hell 264 00:14:00,854 --> 00:14:04,545 this is, with basically an unbounded blank check to go 265 00:14:04,545 --> 00:14:07,305 ahead and spin up nat gateways to its heart's content. 266 00:14:07,574 --> 00:14:09,765 Uh, there's no way this will wind up being 267 00:14:09,765 --> 00:14:11,925 a hilariously expensive joke at my expense. 268 00:14:12,640 --> 00:14:14,260 Yeah, that's, that's a brave choice. 269 00:14:14,440 --> 00:14:19,210 There I say slightly more sensible choices to, um, have this 270 00:14:19,210 --> 00:14:24,370 in a controlled, guarded development environment set up. 271 00:14:24,370 --> 00:14:26,950 And that's where fundamentally what Oona is and what we built at, 272 00:14:26,950 --> 00:14:31,655 at Gipp put for a long time and now, um, extended for agents so. 273 00:14:32,460 --> 00:14:35,939 The heart of the product that is the environments remains. 274 00:14:36,240 --> 00:14:39,540 We now speak of that as ONA environments, and within these 275 00:14:39,540 --> 00:14:43,230 environments we run an agent, ONA agent that that does its 276 00:14:43,230 --> 00:14:47,580 work and it's subject to the same guardrails that previously 277 00:14:47,580 --> 00:14:50,610 existed for these environments, plus specific agent guardrails. 278 00:14:50,910 --> 00:14:55,200 So you can decide what it has access to if you want to, 279 00:14:55,230 --> 00:14:57,300 you can give it unbounded access to your AWS account. 280 00:14:57,300 --> 00:14:59,130 I would not recommend that by default. 281 00:14:59,130 --> 00:15:01,290 Obviously comes locked down, has same defaults, but. 282 00:15:02,430 --> 00:15:09,090 The key point here is we renamed the company because it signified 283 00:15:09,090 --> 00:15:11,460 the next step on this trajectory we've been on all along. 284 00:15:11,730 --> 00:15:12,750 You know, it's not a pivot. 285 00:15:12,750 --> 00:15:14,490 It's not a random edition offshoot. 286 00:15:14,490 --> 00:15:15,780 We gotta do something with ai. 287 00:15:16,170 --> 00:15:20,550 It's so naturally followed that these development environments 288 00:15:20,910 --> 00:15:25,920 that we built for humans also work very well for machines. 289 00:15:25,920 --> 00:15:27,660 In fact, we. 290 00:15:28,110 --> 00:15:31,290 When we architected the platform, we thought of machine use cases, 291 00:15:31,290 --> 00:15:35,610 not necessarily agents at the time, but it was clear that there'd be 292 00:15:35,610 --> 00:15:39,900 more, um, machine and machine use cases that become relevant, that also 293 00:15:39,900 --> 00:15:42,570 need development environments and that fit the bill so perfectly now. 294 00:15:44,205 --> 00:15:46,995 There's a lot to be said for the ability 295 00:15:46,995 --> 00:15:49,125 for systems to interface with each other. 296 00:15:49,155 --> 00:15:53,235 Well, I would argue that MCP is potentially a revolution in 297 00:15:53,235 --> 00:15:57,255 its infancy just because now you have a, it goes beyond APIs. 298 00:15:57,255 --> 00:16:01,435 These are things that self-describe in a. Parable way to each other, 299 00:16:01,435 --> 00:16:06,564 what the tool is, what this endpoint lets you do that has legs, uh, 300 00:16:06,564 --> 00:16:10,135 that, that extend far beyond a particular iteration of these things. 301 00:16:10,314 --> 00:16:10,555 Like what? 302 00:16:10,555 --> 00:16:13,285 It's effectively from my old person perspective, it's the 303 00:16:13,285 --> 00:16:16,405 sense of what if every time you connect to an endpoint. 304 00:16:16,675 --> 00:16:20,574 It would give you the equivalent of a man page that told you what it did, how 305 00:16:20,574 --> 00:16:23,875 it worked, what arguments it could take, and best results do the following. 306 00:16:24,324 --> 00:16:25,734 That is non-trivial. 307 00:16:25,734 --> 00:16:27,474 I'm sort of annoyed we didn't come up with 308 00:16:27,474 --> 00:16:30,834 that as a, as a standard long before now. 309 00:16:31,495 --> 00:16:34,464 I mean, you know, at least you didn't try to push the semantic web for decades. 310 00:16:34,464 --> 00:16:38,724 Like I'm pretty sure there's some people who, uh, you know, who are even 311 00:16:38,724 --> 00:16:42,120 more annoyed at the success of something as simple as MCP than you are. 312 00:16:42,810 --> 00:16:46,380 It's the, I think part of the problem and the reason we're seeing it work here 313 00:16:46,380 --> 00:16:50,250 is you cannot universally change the way that humans interact with something. 314 00:16:50,280 --> 00:16:53,010 Uh, source people will still be calling you GI Pod 315 00:16:53,010 --> 00:16:55,320 20 years from now in some corners of the world. 316 00:16:55,680 --> 00:17:01,020 The, but when you have a shift that's powered by LLMs, suddenly there is a, that 317 00:17:01,020 --> 00:17:05,069 sort of global context and Overton window that moves extraordinarily rapidly. 318 00:17:05,400 --> 00:17:08,224 I. In fact, that's one of the challenges I suspect you'll have is it's 319 00:17:08,224 --> 00:17:12,845 going to take some time for LLMs themselves to get word of the name change. 320 00:17:13,300 --> 00:17:17,224 I, I found that whenever I'm building something new and just vibe coding 321 00:17:17,224 --> 00:17:20,704 something, shit posty, it'll often park it on Versal for a front end. 322 00:17:20,795 --> 00:17:23,525 Now, I don't have strong opinions about front end. 323 00:17:23,525 --> 00:17:25,055 I just know I'm bad at it globally. 324 00:17:25,085 --> 00:17:29,045 But that's the one, the LLM picks and like I'm better correct the robot please. 325 00:17:30,360 --> 00:17:32,699 Absolutely like the, you know, the name GI Pod is 326 00:17:32,699 --> 00:17:35,610 essentially before the cutoff of most models right now. 327 00:17:35,879 --> 00:17:37,560 But then that too will change. 328 00:17:37,620 --> 00:17:38,760 Obviously there are new models. 329 00:17:38,760 --> 00:17:41,580 I mean, you know, only 4.5 just dropped, so. 330 00:17:43,020 --> 00:17:44,850 That, that too will change and the the models 331 00:17:44,850 --> 00:17:46,650 will, will adapt and, and learn a new thing. 332 00:17:47,280 --> 00:17:52,080 That said, I actually like the idea that we are so well known that 333 00:17:52,080 --> 00:17:54,930 even 20 years from now someone is gonna refer to us as gift pod 334 00:17:56,100 --> 00:17:56,340 I. 335 00:17:56,340 --> 00:17:58,560 The question is, is whether that is some, whether that's because 336 00:17:58,560 --> 00:18:02,040 people are actively using it then or someone is just so ornery and 337 00:18:02,040 --> 00:18:06,240 obstinate that they refuse to accept that anything after 2023 exists. 338 00:18:06,660 --> 00:18:09,420 I'm starting to see the joys of being a curmudgeon. 339 00:18:09,670 --> 00:18:13,450 So these days now, since people have to take a step back and ask 340 00:18:13,450 --> 00:18:15,760 the question a little bit differently since I, I imagine that 341 00:18:15,760 --> 00:18:19,570 the, the nuances of the answer are, are there, what does ONA do? 342 00:18:20,230 --> 00:18:24,040 Very fundamentally, the thinking goes, we now 343 00:18:24,040 --> 00:18:27,490 have these machines that can do work for us. 344 00:18:28,169 --> 00:18:29,879 You know that we can give a task and to 345 00:18:29,879 --> 00:18:33,090 varying degrees of autonomy can do work for us. 346 00:18:33,270 --> 00:18:36,810 A mental model that we found very helpful is time between disengagements. 347 00:18:37,320 --> 00:18:40,379 It's a mental model coming from self-driving cause, and it describes 348 00:18:40,379 --> 00:18:43,290 the time between the car disengaging and the human having to take over. 349 00:18:43,649 --> 00:18:46,770 It's a measure of autonomy and seconds is essentially lane 350 00:18:46,770 --> 00:18:51,389 assist and minutes to hours is backseat of a Waymo with. 351 00:18:51,854 --> 00:18:53,385 Agents we're seeing the same thing. 352 00:18:53,385 --> 00:18:55,905 You know where we're coming from, this tap, tap, auto complete 353 00:18:56,024 --> 00:19:00,824 lane assist, and we're moving to minutes, hours of sensible 354 00:19:00,824 --> 00:19:06,074 autonomous work called Code Codex on agent, all demonstrate that. 355 00:19:07,094 --> 00:19:10,060 Now, the question then is how do we turn 356 00:19:10,334 --> 00:19:12,554 this increasing autonomy into productivity? 357 00:19:13,064 --> 00:19:14,415 Because that's obviously what we're asked for. 358 00:19:14,415 --> 00:19:16,645 Fundamentally, software creation is an. 359 00:19:17,324 --> 00:19:18,165 Economic endeavor. 360 00:19:18,165 --> 00:19:19,605 So you know, it needs to be economical. 361 00:19:19,605 --> 00:19:21,824 How can we, how can we turn this into more productivity? 362 00:19:22,215 --> 00:19:25,125 And the only way we can really do that is by doing more things in parallel. 363 00:19:25,304 --> 00:19:27,675 If I now need to sit there and watch the agent do its 364 00:19:27,675 --> 00:19:31,185 thing, I didn't gain much because it's my time as a human. 365 00:19:31,185 --> 00:19:31,754 That's expensive. 366 00:19:31,784 --> 00:19:32,715 It's human attention. 367 00:19:32,715 --> 00:19:33,405 That's expensive. 368 00:19:34,215 --> 00:19:37,215 So how do we, how do we scale human attention fundamentally and. 369 00:19:38,460 --> 00:19:41,100 Again, the only way we can do this is by doing more things in parallel. 370 00:19:41,760 --> 00:19:45,120 Fundamentally, the interfaces that we as software 371 00:19:45,120 --> 00:19:47,370 engineers use today aren't built for this. 372 00:19:47,610 --> 00:19:52,590 They're built to do one thing very deeply at a time writing code. 373 00:19:52,590 --> 00:19:52,860 Right. 374 00:19:53,850 --> 00:19:56,040 Now our interfaces need to change, and the 375 00:19:56,040 --> 00:19:58,620 environments in which we do that work needs to change. 376 00:19:58,800 --> 00:20:00,990 My laptop is built to do one thing at a time. 377 00:20:01,500 --> 00:20:04,379 I mean, anyone who's tried to run different Python versions on 378 00:20:04,379 --> 00:20:08,610 one machine knows what I'm talking about, and so are my IDs. 379 00:20:08,610 --> 00:20:12,419 So my environments and my interfaces need to change. 380 00:20:13,064 --> 00:20:16,995 To get the productivity out of these agents, and that's very fundamentally 381 00:20:16,995 --> 00:20:21,794 what on does is it gives you ma, as many of these environments as 382 00:20:21,794 --> 00:20:25,304 you need perfectly set up for the task at hand and it gives you an 383 00:20:25,304 --> 00:20:30,345 interface that MA that helps you find flow and joy in doing more things. 384 00:20:30,975 --> 00:20:31,365 Parallel. 385 00:20:32,024 --> 00:20:35,895 This episode is brought to you by ona, formerly Git Pod. 386 00:20:35,985 --> 00:20:39,135 Are you tired of coding agents pushing your S3 bucket to the wrong 387 00:20:39,135 --> 00:20:42,764 AWS account or having them commit your entire downloads folder 388 00:20:42,764 --> 00:20:45,885 because they thought your tax documents were part of the code base? 389 00:20:46,004 --> 00:20:49,215 All while using privacy policies that basically say that your 390 00:20:49,215 --> 00:20:52,574 customer data deserves a nice vacation on random cloud servers. 391 00:20:52,905 --> 00:20:57,285 Introducing ona where your coding agents run in isolated sandboxes. 392 00:20:57,285 --> 00:20:59,265 Securely within your own VPC. 393 00:20:59,685 --> 00:21:03,675 ona lets you run agents at scale with workflows, bulk open 394 00:21:03,675 --> 00:21:06,705 requests that finally tackle that Java migration that you started 395 00:21:06,705 --> 00:21:10,395 in 2019 or automatically fix CDEs when your scans find them. 396 00:21:10,815 --> 00:21:13,035 ona also supports private AI models. 397 00:21:13,050 --> 00:21:16,980 Through Amazon Bedrock that your corporate overlords might even approve of. 398 00:21:17,340 --> 00:21:22,649 Head to Oona, that's o a.com and get $200 of free credits using 399 00:21:22,649 --> 00:21:26,580 the code screaming in the cloud because your laptop wasn't designed 400 00:21:26,580 --> 00:21:30,254 to babysit over caffeinated rogue coding agents with root access. 401 00:21:31,110 --> 00:21:35,760 At some level, I'm starting to feel that my A DHD in attentiveness and 402 00:21:35,760 --> 00:21:38,790 pivoting from thing to thing to thing has become something of an asset. 403 00:21:38,820 --> 00:21:42,960 When you have agent driven stuff, uh, I would like it a little bit more. 404 00:21:42,960 --> 00:21:46,470 If there were a healthy medium, somewhere between you have 405 00:21:46,470 --> 00:21:50,190 full access to everything, go ahead and never ask for feedback 406 00:21:50,430 --> 00:21:54,900 versus, oh, am I allowed to read this file that I just wrote? 407 00:21:55,170 --> 00:22:00,720 There's a, there is a different, there's a sliding scale of comfort with it and 408 00:22:01,650 --> 00:22:05,700 the things for which I wish to be interrupted and need to give human input on. 409 00:22:05,790 --> 00:22:08,280 And conversely, there are times I see it doing things where I have 410 00:22:08,280 --> 00:22:11,460 to see how fast I can hit control C because no, no, no, no, no. 411 00:22:11,460 --> 00:22:14,760 I happen to know that sort of thing very well and down that path lies madness. 412 00:22:15,600 --> 00:22:16,170 Absolutely. 413 00:22:16,170 --> 00:22:19,140 I think there, there are two key elements that, that you brought up here. 414 00:22:19,140 --> 00:22:21,780 One is globally, what is the thing allowed 415 00:22:21,780 --> 00:22:23,460 to do and what isn't it allowed to do? 416 00:22:24,525 --> 00:22:27,555 Right now, you know, we're as an, as an industry, we're working 417 00:22:27,555 --> 00:22:31,635 with these reasonably simplistic denialists, you know, where you 418 00:22:31,635 --> 00:22:34,305 tell an agent, Hey, you're not allowed to run AWS because I don't 419 00:22:34,305 --> 00:22:38,355 want you to drop my production RDS instance, but the agent is gonna 420 00:22:38,355 --> 00:22:41,265 get very, very clever and doesn't care about compliance at all. 421 00:22:41,265 --> 00:22:42,795 You know, agents don't care about getting fired, 422 00:22:42,795 --> 00:22:45,075 so it's gonna try and still make it happen. 423 00:22:45,615 --> 00:22:47,055 I've worked with people like that. 424 00:22:47,265 --> 00:22:48,405 Please continue. 425 00:22:49,215 --> 00:22:51,525 Yeah, it's not only agents, so. 426 00:22:52,545 --> 00:22:55,845 Just denying, Hey, you can't run the, AWS command isn't gonna do much good. 427 00:22:55,845 --> 00:22:56,985 It needs to go deeper than that. 428 00:22:56,985 --> 00:22:59,805 And that's something that we're very, uh, that we're exploring right now. 429 00:22:59,805 --> 00:23:01,515 Like, how can we bake that into the environment? 430 00:23:01,515 --> 00:23:04,785 How can we make these guardrails more sophisticated? 431 00:23:05,055 --> 00:23:05,475 That's one. 432 00:23:05,475 --> 00:23:08,355 The other is, if you're doing five things in 433 00:23:08,355 --> 00:23:10,635 parallel, you know, how do you steer this agent? 434 00:23:10,875 --> 00:23:12,555 How do you, how do you get good feedback? 435 00:23:12,645 --> 00:23:14,295 How do you give good feedback? 436 00:23:14,895 --> 00:23:16,155 And here we're. 437 00:23:16,965 --> 00:23:19,485 I think we've hit a very nice form factor that 438 00:23:19,485 --> 00:23:21,795 lets you guide the agent as it does its work. 439 00:23:22,155 --> 00:23:26,055 It's gonna pick up your messages when, when it think it is the right time. 440 00:23:26,295 --> 00:23:30,975 And we've worked hard on making sort of an emergency stop button, like you 441 00:23:30,975 --> 00:23:34,605 can hit escape and it's gonna stop dead in its tracks because it's really, 442 00:23:34,605 --> 00:23:37,815 really important for you to retain control over what the thing is doing 443 00:23:38,265 --> 00:23:38,535 there. 444 00:23:38,535 --> 00:23:41,565 There's also this idea that it, it is forcing 445 00:23:41,565 --> 00:23:45,165 in some ways, rigor that I am seeing people. 446 00:23:45,745 --> 00:23:48,535 Actually care about making things reproducible of, huh? 447 00:23:48,835 --> 00:23:51,745 I really will need a rollback strategy here 448 00:23:51,745 --> 00:23:53,485 instead of hand waving my way around it. 449 00:23:53,485 --> 00:23:55,525 Because sometimes it'll do disastrous things. 450 00:23:55,795 --> 00:23:58,255 And we've seen some public examples of it doing those 451 00:23:58,255 --> 00:24:02,545 sorts of things where it becomes really clear that people 452 00:24:02,545 --> 00:24:04,855 have paid insufficient attention to a lot of these. 453 00:24:04,885 --> 00:24:06,805 Like, Hey, I just deleted my entire database. 454 00:24:06,805 --> 00:24:08,185 What do I do about that? 455 00:24:08,185 --> 00:24:11,305 Like, well, ideally you make different slash better choices. 456 00:24:12,270 --> 00:24:14,489 Absolutely, like one interesting effect of this 457 00:24:14,489 --> 00:24:19,530 is I now raise PRS that I need to review myself. 458 00:24:20,429 --> 00:24:23,909 So I have an H invite code, create a draft pr, and then I 459 00:24:23,909 --> 00:24:26,729 review that draft PR as though it was written by someone else. 460 00:24:27,540 --> 00:24:31,139 So code that has my name on it, you know, now I need 461 00:24:31,139 --> 00:24:33,540 to make sure that it's worthy of having my name on it. 462 00:24:33,540 --> 00:24:35,310 Like it's still my reputation on the line here. 463 00:24:35,699 --> 00:24:40,110 And so there, there is, um, uh, there's an interesting change in dynamic. 464 00:24:41,294 --> 00:24:43,185 One other thing is it's. 465 00:24:44,145 --> 00:24:45,465 Actually incredibly addictive. 466 00:24:45,465 --> 00:24:48,255 Like for a long time I was really worried about how are we 467 00:24:48,255 --> 00:24:54,735 gonna find joy and flow in this multitasking A DHD feeding. 468 00:24:55,155 --> 00:24:58,515 What sounds like a nightmare, to be honest, like, you know, had 469 00:24:58,515 --> 00:25:01,485 someone told me two years ago that, hey, the thing you're really 470 00:25:01,485 --> 00:25:03,885 gonna do is you're gonna work on five things at the same time. 471 00:25:03,885 --> 00:25:07,755 I would've taught that person to insert expletive fear, you know? 472 00:25:08,115 --> 00:25:08,685 Yes. 473 00:25:09,165 --> 00:25:09,735 Now. 474 00:25:09,764 --> 00:25:11,655 So for me, this has been a really interesting 475 00:25:11,655 --> 00:25:14,024 question is how can we find flow and joy in this? 476 00:25:14,024 --> 00:25:17,355 And it turns out that one, it's an interface question, but 477 00:25:17,355 --> 00:25:22,185 also as software engineers, arguably we have a somewhat. 478 00:25:22,950 --> 00:25:27,510 Addictive addict, if that's the word, mindset to begin with. 479 00:25:27,510 --> 00:25:30,030 Because you know who is, ah, just one more 480 00:25:30,030 --> 00:25:31,560 change and then my tests are gonna pass. 481 00:25:31,560 --> 00:25:33,180 Just one more change and then it's gonna work. 482 00:25:33,270 --> 00:25:34,890 How many nights have we spent doing that? 483 00:25:34,890 --> 00:25:37,020 So arguably there's some addictive pattern here already. 484 00:25:37,470 --> 00:25:39,630 We're essentially playing a slot machine, you 485 00:25:39,630 --> 00:25:41,730 know, just one more change and it's gonna work. 486 00:25:42,360 --> 00:25:45,685 What agents have done is they've made it incredibly 487 00:25:45,735 --> 00:25:48,330 cheap to play five slot machines at the same time. 488 00:25:48,885 --> 00:25:50,745 Yeah, that, that's a good way of putting it. 489 00:25:51,105 --> 00:25:54,825 It's so addictive that I've contemplated adding parental controls for myself. 490 00:25:55,575 --> 00:25:59,835 Uh, I've seen git uh, work trees being used explicitly for this, where you 491 00:25:59,835 --> 00:26:02,295 can check out different branches to different directories and let these 492 00:26:02,295 --> 00:26:05,835 things run in parallel on either different issues or, alright, we're gonna 493 00:26:05,835 --> 00:26:07,590 have a bake off and see which one of you comes up with the best answer. 494 00:26:08,565 --> 00:26:10,004 What I'm waiting for is the agent now that 495 00:26:10,004 --> 00:26:12,405 supervises those things and makes those evaluations. 496 00:26:12,405 --> 00:26:14,685 Like I want some, I want like the project manager at this 497 00:26:14,685 --> 00:26:17,985 point, I something that can say, oh, this doesn't pass muster. 498 00:26:17,985 --> 00:26:20,865 Or, okay, here's a whole bunch of tasks, or I'm trying to one shot it. 499 00:26:20,865 --> 00:26:23,504 We're gonna break it down and pass it out to each of you in sequence. 500 00:26:24,524 --> 00:26:26,235 I think this is, this is a very interesting space. 501 00:26:26,235 --> 00:26:28,905 So like the, the sort of multi-agent interaction, 502 00:26:29,205 --> 00:26:30,705 I don't think anyone's corrected that yet. 503 00:26:31,225 --> 00:26:33,265 Um, there are very interesting ideas out there. 504 00:26:33,955 --> 00:26:36,415 This is certainly something that will come. 505 00:26:36,415 --> 00:26:42,385 Also, what we see right now as a key skill now, is to really decompose 506 00:26:42,385 --> 00:26:45,235 and break down a problem into a chunk that works for an agent. 507 00:26:45,235 --> 00:26:48,625 Like agents are tools and so you need to learn how to use them, how to 508 00:26:48,625 --> 00:26:52,855 prompt them, how to use 'em well, what, what size a problem they can attack. 509 00:26:53,980 --> 00:26:57,580 You know, doing this decomposition is, is a, is a very valuable skill right 510 00:26:57,580 --> 00:27:01,629 now that we'd obviously all want to be able to outsource to yet another agent. 511 00:27:01,840 --> 00:27:04,149 That that is a constant problem we're all dealing with right now. 512 00:27:04,149 --> 00:27:07,389 It's a universal problem where I, we are pushing 513 00:27:07,420 --> 00:27:10,810 the frontier bounds here and seeing what's possible. 514 00:27:10,810 --> 00:27:12,100 I think if you've only played with this stuff 515 00:27:12,100 --> 00:27:14,050 a few months ago and like, eh, it was okay. 516 00:27:15,155 --> 00:27:16,565 It's time to reevaluate it. 517 00:27:16,595 --> 00:27:20,435 This is one of those rapidly advancing areas, and 518 00:27:20,435 --> 00:27:22,775 I generally want to call out hype when I see it. 519 00:27:22,775 --> 00:27:24,455 Yes, we are in a hype bubble here. 520 00:27:24,455 --> 00:27:27,575 I think that is not particularly controversial, but 521 00:27:27,575 --> 00:27:30,575 unlike the insane blockchain hype bubble, there's clearly 522 00:27:30,575 --> 00:27:33,965 something of value here that is, this is not problem. 523 00:27:33,965 --> 00:27:36,125 This is not solution in search of a problem in quite the same way. 524 00:27:36,215 --> 00:27:40,775 This is something that is transforming the way some things are being done. 525 00:27:40,985 --> 00:27:43,955 Now, maybe we're a little too eager to map those to everything else. 526 00:27:44,255 --> 00:27:48,155 But there is some kernel here of this has staying power. 527 00:27:48,785 --> 00:27:49,445 Absolutely. 528 00:27:50,045 --> 00:27:50,465 And. 529 00:27:51,389 --> 00:27:54,450 Is our agents gonna replace humans? 530 00:27:54,570 --> 00:27:55,800 I personally don't think so. 531 00:27:56,399 --> 00:27:58,020 You know, they're, they're gonna augment humans. 532 00:27:58,020 --> 00:28:01,800 They're gonna make people more effective, but they're not gonna replace them. 533 00:28:02,250 --> 00:28:04,080 Also, Javins Paradox is very real. 534 00:28:04,080 --> 00:28:06,180 The moment we make something cheaper, we do more of it. 535 00:28:06,720 --> 00:28:08,909 So we're now making software production cheaper. 536 00:28:08,909 --> 00:28:11,580 So we're gonna do more of it simply, we're gonna write more software, 537 00:28:12,360 --> 00:28:13,470 we're gonna write more software. 538 00:28:13,470 --> 00:28:16,290 That's historically, that has been the antipater. 539 00:28:16,665 --> 00:28:19,275 Think about this, where it used to be that, oh, 540 00:28:19,305 --> 00:28:21,255 we're gonna solve our own custom problem in house. 541 00:28:21,255 --> 00:28:22,635 We're going to write it ourselves. 542 00:28:22,965 --> 00:28:24,735 I've worked in too many environments where there's 543 00:28:24,735 --> 00:28:27,105 such a strong, not invented here syndrome that everyone 544 00:28:27,105 --> 00:28:29,715 builds custom stuff, but becomes a maintenance nightmare. 545 00:28:29,865 --> 00:28:32,775 So it turned into a point a lot of shops, my own included, 546 00:28:33,045 --> 00:28:36,615 where we historically have been down this path of we're 547 00:28:36,615 --> 00:28:39,165 gonna build our own custom tooling for my newsletter. 548 00:28:39,165 --> 00:28:41,535 It is a rat's nest nightmare of different things 549 00:28:41,535 --> 00:28:43,545 bolted together to build a production system. 550 00:28:43,545 --> 00:28:44,625 And when someone asks me why I didn't use. 551 00:28:45,170 --> 00:28:47,540 Curated, do co. My question was. 552 00:28:48,135 --> 00:28:48,675 Wait, why? 553 00:28:48,675 --> 00:28:49,514 I didn't use what? 554 00:28:49,814 --> 00:28:51,945 Because I didn't know it existed, or I would 555 00:28:51,945 --> 00:28:54,435 have, and it would've saved me so much effort. 556 00:28:54,675 --> 00:28:57,915 But we're seeing that invert now where there's a bunch of little 557 00:28:57,915 --> 00:29:00,975 things that I need to do throughout the course of my workday. 558 00:29:01,304 --> 00:29:04,425 I am not going to hire a developer to do these things, and I'm not going to 559 00:29:04,425 --> 00:29:07,905 sit around and build all of the, all these tools or pay for these things. 560 00:29:08,205 --> 00:29:13,034 But hey, every week I need to find my top 10, uh, most engaged 561 00:29:13,064 --> 00:29:15,554 posts on Blue Sky so I can put it in the hidden newsletter. 562 00:29:15,745 --> 00:29:17,514 Easter egg that's in every episode. 563 00:29:17,844 --> 00:29:21,774 I can write a dumb script that does that in, I, I tell an agent to do 564 00:29:21,774 --> 00:29:24,114 it, and I go get a cup of coffee, and it's done by the time I get back. 565 00:29:24,354 --> 00:29:27,625 Suddenly writing more software is the change for the first time. 566 00:29:27,625 --> 00:29:30,834 Nons sarcastically that'll fix it, because usually that's a 567 00:29:30,834 --> 00:29:33,804 sarcastic thing to say, oh, I'm gonna write the more software. 568 00:29:33,804 --> 00:29:34,135 Great. 569 00:29:34,135 --> 00:29:34,854 That'll fix it. 570 00:29:35,125 --> 00:29:38,425 This will fix it because it's the glue between things. 571 00:29:39,375 --> 00:29:40,125 Absolutely. 572 00:29:40,185 --> 00:29:44,025 Also, we no longer need to excessively generalize 573 00:29:44,055 --> 00:29:46,095 because the creation of software has become so cheap. 574 00:29:46,095 --> 00:29:49,635 I can solve this one specific problem, and I don't need to solve it for 575 00:29:49,635 --> 00:29:52,995 these other three instances because, you know, I, I can just ask an agent 576 00:29:52,995 --> 00:29:56,445 to solve it for this, for these other three specific instances specifically. 577 00:29:56,775 --> 00:29:59,805 And so software becomes more and simpler that way. 578 00:30:01,140 --> 00:30:05,370 It, it also changes the way that I think we view the cost of doing software. 579 00:30:05,400 --> 00:30:09,060 The, the pricing models for all these agent things are very strange. 580 00:30:09,060 --> 00:30:12,000 I've, I've seen the leaderboards for people who are using the $200 a month 581 00:30:12,000 --> 00:30:15,600 clause subscription and how much, uh, value they're getting out of it. 582 00:30:15,630 --> 00:30:17,160 If they were paying per inference, it's tens 583 00:30:17,160 --> 00:30:19,200 of thousands of dollars a month in some cases. 584 00:30:19,530 --> 00:30:22,110 It, it makes me worry that, okay, is this as, is 585 00:30:22,110 --> 00:30:24,990 this as economically sustainable as I want it to be? 586 00:30:25,020 --> 00:30:27,360 'cause I'm not going back to writing JavaScript by hand. 587 00:30:27,360 --> 00:30:27,630 I'm just. 588 00:30:27,680 --> 00:30:31,640 Not, so I, I'm very interested in getting local inference 589 00:30:31,640 --> 00:30:34,280 to a point where it can at least do the fancy tab, complete 590 00:30:34,280 --> 00:30:37,400 style thing, even if it's not as good as the frontier models. 591 00:30:37,400 --> 00:30:39,080 There are, there are many things I don't 592 00:30:39,080 --> 00:30:41,150 need it to reach out to the internet for. 593 00:30:41,150 --> 00:30:43,400 I don't need the very latest and greatest Claude 594 00:30:43,400 --> 00:30:47,930 Sonnet 4.5 to go ahead and indent my YAML properly. 595 00:30:47,965 --> 00:30:49,970 I, I feel like that's the sort of thing 596 00:30:49,970 --> 00:30:51,445 that a model from three years ago can do. 597 00:30:53,370 --> 00:30:57,030 And there's that, um, token, short squeeze article that was 598 00:30:57,030 --> 00:30:59,790 all the hype on, on the orange website not too long ago. 599 00:31:01,155 --> 00:31:04,125 The, the key premise of it is that, you know, tokens 600 00:31:04,125 --> 00:31:06,495 get, get ever more, get ever cheaper and cheaper. 601 00:31:06,495 --> 00:31:12,705 So if you just look at GPT-4 level, so Elemis ELO one 30, um, 602 00:31:12,735 --> 00:31:17,865 intelligence a year ago as compared to now, it dropped by a factor of 140. 603 00:31:18,675 --> 00:31:21,525 At the same time, we're using about 10 K more tokens. 604 00:31:21,795 --> 00:31:23,565 So we're using an order of mag, two orders of 605 00:31:23,565 --> 00:31:25,545 magnitudes, more tokens than the price dropped. 606 00:31:26,610 --> 00:31:30,930 Unnecessarily we'll need to see two things. 607 00:31:30,930 --> 00:31:36,510 One is, as you point out more precise models, you know, that 608 00:31:36,510 --> 00:31:40,500 make that cost intelligence straight off to a point where, 609 00:31:40,500 --> 00:31:43,530 where it works like this one size fits all isn't gonna scale. 610 00:31:44,130 --> 00:31:44,850 The other is. 611 00:31:45,705 --> 00:31:52,125 We'll need to recognize that AI doesn't make the creation of software free. 612 00:31:52,125 --> 00:31:53,805 It just changes the economics. 613 00:31:54,405 --> 00:31:58,575 So scaling a model is much easier than scaling humans, and this is 614 00:31:58,575 --> 00:32:01,125 why we can produce more software, but that doesn't make it free. 615 00:32:02,565 --> 00:32:09,495 And this time right now where we live in VC money subsidized 616 00:32:09,525 --> 00:32:13,725 token land will need to come to an end eventually. 617 00:32:14,475 --> 00:32:17,805 So I think we're gonna see a proliferation of different 618 00:32:17,805 --> 00:32:20,775 models that make that trade off better and we'll need to see. 619 00:32:20,775 --> 00:32:24,165 And we are seeing already like pricing models that are much more 620 00:32:24,165 --> 00:32:28,185 aligned with the value you're getting rather than a flat fee. 621 00:32:28,695 --> 00:32:30,495 Yes and no, because we're not seeing 622 00:32:30,495 --> 00:32:32,385 outcome-based pricing on any of these things. 623 00:32:32,385 --> 00:32:35,625 It's not like, okay, I'll only charge you if the code works like that. 624 00:32:35,925 --> 00:32:37,485 That would be an interesting gamble. 625 00:32:37,515 --> 00:32:39,440 But I don't know anyone who'd want to take the other side. 626 00:32:40,604 --> 00:32:41,804 That's a really tough one. 627 00:32:41,804 --> 00:32:44,985 Finding a way to make this one really work, I think is extremely interesting 628 00:32:44,985 --> 00:32:49,635 because it aligns incentives so, so well, the question is, what is the outcome? 629 00:32:50,294 --> 00:32:54,675 You know, like code working an agent can show you that the code works. 630 00:32:54,824 --> 00:32:56,145 Does it do the right thing? 631 00:32:56,385 --> 00:32:56,955 I don't know. 632 00:32:57,195 --> 00:32:58,695 Does it solve your business problem? 633 00:32:58,875 --> 00:32:59,475 No idea. 634 00:32:59,685 --> 00:33:03,044 So the, you know, what is the outcome you're optimizing for? 635 00:33:04,094 --> 00:33:06,764 Which is why I reckon most don't. 636 00:33:07,695 --> 00:33:11,415 Price this way yet because it's incredibly tough nut to crack. 637 00:33:12,255 --> 00:33:12,465 Yeah. 638 00:33:12,630 --> 00:33:16,005 I, I think that this is where some of the most interesting stuff is yet to come. 639 00:33:16,275 --> 00:33:20,715 So I've been doing a lot of weird work lately in random shit posting 640 00:33:20,715 --> 00:33:23,895 things, and it's great watching it just get done and wait for me. 641 00:33:23,895 --> 00:33:26,295 And in some cases it'll even ping me when it's ready 642 00:33:26,295 --> 00:33:28,905 by hook it into, uh, the right notification service. 643 00:33:29,265 --> 00:33:29,745 But. 644 00:33:30,165 --> 00:33:33,045 I've been doing it hanging out on an EC2 instance, and it's 645 00:33:33,045 --> 00:33:38,025 doing that in a Team Ox section, uh, too, ah, te ox session. 646 00:33:38,265 --> 00:33:38,895 There we go. 647 00:33:39,345 --> 00:33:43,815 And that's great, but it's a colossal pain of the butt to do that from blink. 648 00:33:44,115 --> 00:33:47,835 Uh, I can do it, but it's not pleasant and it makes me sad. 649 00:33:48,105 --> 00:33:50,865 Do you see a future where this gets easier to be done on 650 00:33:50,865 --> 00:33:53,835 mobile devices as we're out walking around, not staring at the 651 00:33:53,835 --> 00:33:56,025 big screen, instead looking at the smaller, happier screen. 652 00:33:57,524 --> 00:33:59,024 Actually, this is already a reality. 653 00:33:59,024 --> 00:34:02,895 So with on spinning up development environments that aren't bound to the 654 00:34:02,895 --> 00:34:07,544 machine you are using ONA from, you can absolutely use it from your phone. 655 00:34:07,544 --> 00:34:11,864 And in fact, we've optimized the web experience also for mobile. 656 00:34:12,194 --> 00:34:15,194 The way I speak about this now is like I'm three times more 657 00:34:15,194 --> 00:34:18,165 productive on my phone than I was six months ago on my laptop. 658 00:34:19,650 --> 00:34:21,300 Like, let me make this very concrete. 659 00:34:21,300 --> 00:34:26,760 I I, at this point, I have a four months old son and many evenings I'll sit 660 00:34:26,760 --> 00:34:30,510 with him on one arm as he is falling asleep, but I don't dare put him down. 661 00:34:30,870 --> 00:34:34,020 Oh, can't do that, that, that, that restarts the cycle. 662 00:34:34,800 --> 00:34:35,850 Exactly, exactly. 663 00:34:35,850 --> 00:34:39,150 Then I have to, you know, shush him and try and put him sleep again. 664 00:34:39,660 --> 00:34:42,630 So clearly I can't use my laptop, but I can use my phone. 665 00:34:43,350 --> 00:34:47,190 And so many ideas for prototypes or actual 666 00:34:47,190 --> 00:34:49,320 changes that before would've been mere. 667 00:34:49,380 --> 00:34:52,139 Notes now are actual prototypes. 668 00:34:52,560 --> 00:34:55,740 I put them into ONA and by the time I wake up the next morning, 669 00:34:56,130 --> 00:34:59,610 the conversations I've had turned into actionable code. 670 00:34:59,700 --> 00:35:01,680 And that's a very fundamentally different way of working. 671 00:35:01,740 --> 00:35:04,770 So being able to do this from mobile is already reality, 672 00:35:04,980 --> 00:35:07,890 and you don't have to use Team X or screen to do it. 673 00:35:08,900 --> 00:35:11,660 Yeah, with the weird control characters and custom keyboards and the rest. 674 00:35:11,810 --> 00:35:12,020 Okay. 675 00:35:12,025 --> 00:35:13,190 You, you convinced me to try it out. 676 00:35:13,280 --> 00:35:15,590 A question I have for you that I've encountered a fair 677 00:35:15,590 --> 00:35:19,520 bit here is the, the multimodal approach to these things. 678 00:35:19,760 --> 00:35:23,120 I can tell an agent to build a thing, I, it can go vibe code. 679 00:35:23,120 --> 00:35:23,690 It's hard out. 680 00:35:23,750 --> 00:35:24,320 Great. 681 00:35:24,620 --> 00:35:24,710 Uh. 682 00:35:25,444 --> 00:35:27,904 To the point where I'll even find myself stuck in that paradigm for 683 00:35:27,904 --> 00:35:30,785 things I really shouldn't be like, oh, go ahead and change this one 684 00:35:30,785 --> 00:35:33,875 string here because I want to change the capitalization of something. 685 00:35:34,240 --> 00:35:37,504 I, I should just be able to pop into vi or whatnot or edit that. 686 00:35:37,685 --> 00:35:40,205 It, it feels like I have to pick a paradigm and stick with 687 00:35:40,205 --> 00:35:42,634 it, maybe past the point where it makes logical sense. 688 00:35:42,680 --> 00:35:43,430 How do you see that? 689 00:35:44,270 --> 00:35:44,540 Yeah. 690 00:35:44,540 --> 00:35:49,129 A lot of agents really are built for a future that isn't here yet and 691 00:35:49,129 --> 00:35:53,029 maybe never will be, where the agent goes a hundred percent of the way. 692 00:35:53,029 --> 00:35:55,700 And I guess the, the set of problems for which this 693 00:35:55,700 --> 00:35:58,430 is true is increasing as agents get more capable. 694 00:35:59,210 --> 00:36:02,330 But there are some things that lm simply aren't good at, or 695 00:36:02,330 --> 00:36:06,770 when, where source code just is the better way of specifying it. 696 00:36:06,770 --> 00:36:09,799 So if I want that color to be green instead of red, it's 697 00:36:09,799 --> 00:36:13,860 m. Much more likely that, you know, changing the hex 698 00:36:13,920 --> 00:36:16,650 myself is faster than trying to describe that to an agent. 699 00:36:17,190 --> 00:36:20,220 ONA is very much built around that idea where 700 00:36:20,220 --> 00:36:23,010 you can engage with code at the right level. 701 00:36:23,010 --> 00:36:24,870 You can choose to not engage with it at all 702 00:36:24,870 --> 00:36:27,120 directly and simply be in the conversation. 703 00:36:27,504 --> 00:36:30,535 Or you can fold open a side panel and there's VS code 704 00:36:30,685 --> 00:36:32,935 right there on the web on the exact same environment. 705 00:36:33,504 --> 00:36:37,944 And if that's not enough, you can open a classic IDE emax if you 706 00:36:37,944 --> 00:36:44,185 have 12 fingers or VI or uh, vs code and inter cusa if you want. 707 00:36:44,185 --> 00:36:47,004 And interact with that with that code more deeply. 708 00:36:47,004 --> 00:36:50,544 So in the same environment, I very much believe that 709 00:36:50,814 --> 00:36:53,814 agents get you very far and they'll go further and further. 710 00:36:54,480 --> 00:36:58,890 But there needs to be a way to engage with the code at that level. 711 00:36:59,340 --> 00:36:59,520 Yeah. 712 00:36:59,550 --> 00:37:01,800 Right now it just feels like that's the expensive context, which 713 00:37:01,800 --> 00:37:04,290 almost as much as switching between entire projects, which I've 714 00:37:04,290 --> 00:37:07,860 gotten used to, but the Ooh, different tool now it, it feels 715 00:37:07,860 --> 00:37:10,680 like even the key bindings feel different and I don't like it. 716 00:37:11,490 --> 00:37:12,210 Absolutely. 717 00:37:12,210 --> 00:37:14,400 You also want that conversation to be there. 718 00:37:14,400 --> 00:37:18,450 You know, what you don't want is to now go into, like, say you open. 719 00:37:19,050 --> 00:37:22,170 An editor and all of a sudden all your conversation, 720 00:37:22,170 --> 00:37:24,509 all that context, no pun intended, is gone. 721 00:37:24,960 --> 00:37:26,880 You really want that continuity between 722 00:37:26,880 --> 00:37:28,860 these different levels of, of engagement. 723 00:37:29,595 --> 00:37:29,835 Yeah. 724 00:37:29,985 --> 00:37:31,995 And then there's the other problem too of, alright, when 725 00:37:31,995 --> 00:37:34,215 do I want to get rid of that context and start fresh on 726 00:37:34,215 --> 00:37:35,985 this code base and have it take a different approach. 727 00:37:36,345 --> 00:37:37,755 There's no right answer yet. 728 00:37:38,685 --> 00:37:39,315 Absolutely. 729 00:37:39,315 --> 00:37:42,495 I think this is really where it comes back to learning how to use 730 00:37:42,495 --> 00:37:46,315 this tool and the tool making it easy for you to work with it. 731 00:37:46,510 --> 00:37:50,770 So for example, we essentially copied Claude Code's Clear Command. 732 00:37:50,955 --> 00:37:52,485 So in Owner also, it can just go slash 733 00:37:52,485 --> 00:37:54,255 clear and it's gonna reset the conversation. 734 00:37:55,545 --> 00:37:56,115 It's. 735 00:37:57,585 --> 00:38:01,815 Features like that, but also behavior like that, that I think 736 00:38:01,815 --> 00:38:04,995 will change over time as agents become more capable and as we 737 00:38:04,995 --> 00:38:09,255 all learn what the right ergonomics are for, uh, for these tools, 738 00:38:10,125 --> 00:38:11,325 it is still an evolving space. 739 00:38:11,685 --> 00:38:14,355 So my guess, my, my closing question for you is, in that 740 00:38:14,355 --> 00:38:17,415 future, as we see this evolving, what place does On Stand in 741 00:38:18,075 --> 00:38:22,214 Owner very fundamentally is the mission control, 742 00:38:22,214 --> 00:38:25,904 the platform for humans and agents writing software. 743 00:38:27,135 --> 00:38:28,215 And that's where we stand. 744 00:38:28,215 --> 00:38:35,145 And 99% of the software isn't written on on weekends, but it's written in 745 00:38:35,145 --> 00:38:40,125 enterprises, it's written in large organizations, and that's who we serve. 746 00:38:40,125 --> 00:38:44,895 Like we want to, we want to be able to bring these 747 00:38:44,895 --> 00:38:48,615 technologies in this way of working to everyone. 748 00:38:48,675 --> 00:38:51,435 And if you are a weekend warrior, please go try Ona. 749 00:38:52,140 --> 00:38:54,720 Go to owner.com, sign up, try it, use it. 750 00:38:55,050 --> 00:38:59,640 It works well for you if you work at an enterprise use owner. 751 00:38:59,670 --> 00:39:02,580 And this is the thing that I find really exciting that we can say this. 752 00:39:03,510 --> 00:39:06,030 What we're really looking to do is to bring. 753 00:39:06,960 --> 00:39:11,640 Environments agents to folks in regulated industries and large 754 00:39:11,640 --> 00:39:16,710 organizations who right now really struggle to get these tools in house. 755 00:39:16,830 --> 00:39:19,950 You know, as an engineer, of course, I want the latest tools, of course I do. 756 00:39:20,339 --> 00:39:24,420 My CISO might not be so happy with me putting my company's source or 757 00:39:24,420 --> 00:39:30,600 this company's source code into some arbitrary cloud or untractable, LLM. 758 00:39:31,300 --> 00:39:34,000 Where ONA stands is bringing these tools and 759 00:39:34,000 --> 00:39:39,580 capabilities to large organizations and individuals like 760 00:39:40,510 --> 00:39:41,170 I like that. 761 00:39:41,170 --> 00:39:44,680 I am curious to see how this story continues to evolve. 762 00:39:44,830 --> 00:39:47,170 I really wanna thank you for taking the time to speak with me. 763 00:39:47,440 --> 00:39:51,220 If people want to learn more, where is the best place for them to find you? 764 00:39:51,970 --> 00:39:54,700 The best place is to head over to owner.com. 765 00:39:55,575 --> 00:39:56,805 Check out the product right there. 766 00:39:56,805 --> 00:39:59,085 And then, um, also of course, Twitter, 767 00:39:59,085 --> 00:40:01,370 LinkedIn, the usual places to, to reach out. 768 00:40:02,940 --> 00:40:04,440 And, uh, thank you so much for having me. 769 00:40:05,430 --> 00:40:09,210 No, and thank you Chris Hel, co-founder and CTO at ona. 770 00:40:09,480 --> 00:40:12,600 I'm cloud economist Corey Quinn, and this is Screaming in the Cloud. 771 00:40:12,839 --> 00:40:14,640 If you've enjoyed this podcast, please, we have a 772 00:40:14,640 --> 00:40:16,920 five star review on your podcast platform of choice. 773 00:40:16,950 --> 00:40:19,589 Whereas if you've hated this podcast, please leave a five 774 00:40:19,589 --> 00:40:22,200 star review on your podcast platform of choice along with an 775 00:40:22,200 --> 00:40:24,690 angry, insulting comment that you don't even have to write. 776 00:40:24,779 --> 00:40:26,674 We'll let the LLM do it for you and don't worry. 777 00:40:27,005 --> 00:40:28,205 It'll probably turn out fine.