1 00:00:00,000 --> 00:00:04,950 Good morning. I'm Gergely and thank you for the introduction. 2 00:00:05,000 --> 00:00:09,719 I now spend most of my time writing the pragmatic engineer hosting the podcast which 3 00:00:09,720 --> 00:00:16,039 is software engineering a big tech from the inside and today I want to bring you into 4 00:00:16,040 --> 00:00:23,199 a journey show show you some stories some information that might not be as readily available 5 00:00:23,200 --> 00:00:25,790 inside AI labs big tech and startups. 6 00:00:25,840 --> 00:00:31,519 In the last few months I went inside the HQ of open AI and Tropic cursor ramp. 7 00:00:31,520 --> 00:00:35,990 I also talked with folks at Uber extensively at linear and other startups. 8 00:00:36,040 --> 00:00:42,680 Here's some of the photos where I went especially ramp with the office guy. 9 00:00:42,800 --> 00:00:49,119 And today I want to cover first what changed then what did not change things that have 10 00:00:49,120 --> 00:00:53,110 broken across the industry and what come next. 11 00:00:53,160 --> 00:00:56,600 Let's start with what's changed and a lot of things have changed. 12 00:00:57,200 --> 00:01:02,559 First thing nobody writes by code by hand anymore except for a a few of you folks but 13 00:01:02,560 --> 00:01:05,238 in in general it's it's just more more and more rare to see. 14 00:01:05,239 --> 00:01:11,399 Of course we we still edit some code by hand but it's crazy how much we are writing with 15 00:01:11,400 --> 00:01:16,839 AI and this was kind of obvious to see in January like we we all woke up in January after 16 00:01:16,840 --> 00:01:21,439 the winter break and we just saw that these models starting from Opus 4.5 and GPT 5.2 17 00:01:21,440 --> 00:01:26,159 were just really really good with the harnesses and from there on people switched and 18 00:01:26,160 --> 00:01:30,639 in most startups what they're tracking these numbers it's going to close to 100% AI generated 19 00:01:30,640 --> 00:01:33,800 code that of course engineers prompt. 20 00:01:34,000 --> 00:01:38,190 Another thing that's changed working with several agents it's increasingly common. 21 00:01:38,240 --> 00:01:43,279 In March I talked with Boris Cherny who told me about how he has five terminal tabs five 22 00:01:43,280 --> 00:01:46,759 cloth codes on his local machine plus five to ten cloths on the web. 23 00:01:46,760 --> 00:01:50,919 Now this is Boris Cherny this is an tropic obviously they will be you know cutting edge 24 00:01:50,920 --> 00:01:55,919 and and they will just honestly use AI sometimes maybe even for the sake of AI but I 25 00:01:55,920 --> 00:01:59,919 talked with Dima Zats who is an an old colleague of mine at an Uber. 26 00:01:59,920 --> 00:02:01,319 He now works at Linear. 27 00:02:01,320 --> 00:02:02,599 And he put it really nicely. 28 00:02:02,600 --> 00:02:06,519 He said, "It used to be a simple world, one mouse, one keyboard, one screen. 29 00:02:06,520 --> 00:02:09,118 You physically couldn't work on more things than one. 30 00:02:09,119 --> 00:02:11,119 But no, we no longer have that limitation." 31 00:02:11,120 --> 00:02:13,719 So, he's doing five to 10 work trees. 32 00:02:13,720 --> 00:02:16,830 He's doing five to 10 agents at the same time. 33 00:02:16,880 --> 00:02:20,958 And just the day before yesterday, I talked with Peter Mattis, 34 00:02:20,959 --> 00:02:22,759 co-founder and CEO of Cockroach Labs. 35 00:02:22,760 --> 00:02:27,799 He's the person who built Gmail's first back end storage and a bunch of really cool storage 36 00:02:27,800 --> 00:02:29,550 stuff inside of Google. 37 00:02:29,600 --> 00:02:30,958 And he told me the same thing. 38 00:02:30,959 --> 00:02:32,909 Often times he's doing things in parallel. 39 00:02:32,959 --> 00:02:36,750 My cognitive overhead of five to 10 agent sessions concurrently. 40 00:02:36,800 --> 00:02:38,319 And he just often does that as well. 41 00:02:38,320 --> 00:02:42,519 So, again, so many people, so many kind of hardcore engineers, if you will, 42 00:02:42,520 --> 00:02:45,639 are have kind of moved to this mode of just parallel agents. 43 00:02:45,640 --> 00:02:47,200 It's pretty common. 44 00:02:47,560 --> 00:02:50,350 IDEs are just fading away. 45 00:02:50,400 --> 00:02:53,719 This was really unexpected for me to see cuz I I love my IDE, 46 00:02:53,720 --> 00:02:57,479 but the trend of this heavyweight IDE usage is trending down. 47 00:02:57,480 --> 00:03:00,070 People are using them just not as much. 48 00:03:00,120 --> 00:03:06,039 Antigravity released in November, the Windsurf kind of rebrand, if you will, 49 00:03:06,040 --> 00:03:09,839 was the last major VS Code fork that was forked. 50 00:03:09,840 --> 00:03:15,199 And I talked with the OpenAI team just in June and they told me that in January they 51 00:03:15,200 --> 00:03:18,238 were considering forking VS Code for the next version of Codex, 52 00:03:18,239 --> 00:03:22,479 but they were kind of torn in between, should we do it, should we not? 53 00:03:22,480 --> 00:03:29,710 And for a little bit they thought it was a mistake to build a Codex without an IDE, 54 00:03:29,760 --> 00:03:33,079 but they quickly realized actually they they did it at the right time and and it was 55 00:03:33,080 --> 00:03:38,359 a win. JetBrains, one of the most beloved IDE brands, they're pivoting to Air, 56 00:03:38,360 --> 00:03:41,279 which is also more of an agentic harness, if you will. 57 00:03:41,280 --> 00:03:43,600 You can check out it looks similar to Codex. 58 00:03:44,400 --> 00:03:51,599 Cursor, the biggest VS Code fork, has pivoted in April to just an agentic interface. 59 00:03:51,600 --> 00:03:53,319 Again, looks very similar to to Codex. 60 00:03:53,320 --> 00:03:57,399 So, all of these tools and when I talk with the Cursor team inside of Cursor, 61 00:03:57,400 --> 00:04:01,119 they told me specifically that the IDE is now just a legacy product. 62 00:04:01,120 --> 00:04:04,520 They need it for enterprise, but it's not growing and it's shrinking. 63 00:04:05,280 --> 00:04:10,159 And I was talking with Kent Beck about this yesterday and he texted me this. 64 00:04:10,160 --> 00:04:12,119 He said like, "Hey, on the on the topic of IDE, 65 00:04:12,120 --> 00:04:16,390 it's not that we don't need more perspective and context for human-based decisions, 66 00:04:16,440 --> 00:04:18,078 it's that the context has changed. 67 00:04:18,079 --> 00:04:20,600 So, maybe that's why it's going away." 68 00:04:22,480 --> 00:04:26,879 Everyone and then some are building their own harness and agent platform. 69 00:04:26,880 --> 00:04:31,790 And by everyone I'm I'm I'm talking the startups, the big tech, the the mid-size companies. 70 00:04:31,840 --> 00:04:34,479 I I put this kind of tweet half-jokingly saying, 71 00:04:34,480 --> 00:04:38,310 "If you're not built your AI coding harness, are you even a serious tech company?" 72 00:04:38,360 --> 00:04:39,879 And here are companies that built it. 73 00:04:39,880 --> 00:04:43,639 Ramp, Stripe, Uber, Block, Shopify, Google, Meta, Amazon, Dropbox, Shopify, 74 00:04:43,640 --> 00:04:48,399 DoorDash, Grab, WorkOS, Monzo, HubSpot, Sierra, Harvey, Bowery, so many more. 75 00:04:48,400 --> 00:04:50,159 Usually, they have cool internal hacks. 76 00:04:50,160 --> 00:04:53,639 By the way, it's really easy to do it as well and they often do it so they can plug in 77 00:04:53,640 --> 00:04:57,799 their their data sources that they might not want to plug into Codex or or whatnot. 78 00:04:57,800 --> 00:05:01,799 And some of them are I I are building on top of Codex, Claude Code, Open Code, 79 00:05:01,800 --> 00:05:03,360 and so on. 80 00:05:04,720 --> 00:05:08,310 Developer work now often starts in Slack. 81 00:05:08,360 --> 00:05:12,759 Uh so, inside OpenAI, they just tag Codex implement this. 82 00:05:12,760 --> 00:05:15,110 Codex is plugged into all of the OpenAI systems. 83 00:05:15,160 --> 00:05:16,759 Anthropic, same thing with Claude, right? 84 00:05:16,760 --> 00:05:18,359 They launched it as a product as well. 85 00:05:18,360 --> 00:05:23,190 Startups increasingly build their own Slack agent and AI bot integrations. 86 00:05:23,240 --> 00:05:27,280 Linear does the same with their their Linear bot and so on. 87 00:05:27,760 --> 00:05:29,840 The agentic software factory. 88 00:05:30,720 --> 00:05:34,279 This used to be a buzzword that I used to be like, "Agentic software what? 89 00:05:34,280 --> 00:05:36,039 Like, you know, like like cars? 90 00:05:36,040 --> 00:05:39,910 Sure, one day." But no, it's actually uh happening. 91 00:05:39,960 --> 00:05:45,599 And one of the best examples is OpenAI's agentic software factory, 92 00:05:45,600 --> 00:05:49,959 which a deep dive is going out in a few hours in the pragmatic engineer so you'll be 93 00:05:49,960 --> 00:05:55,519 able to see but in this diagram that describes how their agentic software factory works 94 00:05:55,520 --> 00:06:01,839 those dash lines those dotted lines are where agents act on typically agent input and 95 00:06:01,840 --> 00:06:04,550 they go back to code x to generate a pull request. 96 00:06:04,600 --> 00:06:11,439 A human in the end often reviews a pull request at for critical ones for large vast radius 97 00:06:11,440 --> 00:06:16,199 but open AI has this really cool concept of a perf factory run by agents which monitor 98 00:06:16,200 --> 00:06:20,710 production and raises PRs to improve production. 99 00:06:20,760 --> 00:06:22,799 And they're just keep iterating on it. 100 00:06:22,800 --> 00:06:26,199 It's not fully automated and they're they'll have to go but it's happening and this is 101 00:06:26,200 --> 00:06:31,239 a big change. Migration is no longer take years if you were paying attention to where 102 00:06:31,240 --> 00:06:37,399 it's not as surprising but some examples Anthropic migrated famously Zig to Rust with 103 00:06:37,400 --> 00:06:43,599 one engineer who wrote most of Bun by the way so this was really fast and it it probably 104 00:06:43,600 --> 00:06:46,959 would have taken years and it probably just wouldn't have been done just because how 105 00:06:46,960 --> 00:06:49,920 when you migrate you need to maintain another thing. 106 00:06:49,960 --> 00:06:51,520 Open AI 107 00:06:51,840 --> 00:06:55,719 as I'll describe in a deep dive I I this is new information that I I got from them. 108 00:06:55,720 --> 00:07:01,510 They are migrating their API that hits all of their interfaces from Python to Rust. 109 00:07:01,560 --> 00:07:06,720 They are about 90% done which was kind of what everyone says by the way so 110 00:07:07,840 --> 00:07:12,799 but they're saying that it took about like four to five months to get here under production 111 00:07:12,800 --> 00:07:16,559 load nothing seen from the outside and they're expecting to to finish in a couple more 112 00:07:16,560 --> 00:07:20,279 months and usually we know these things would have taken years so but this is a great 113 00:07:20,280 --> 00:07:24,279 example cuz I I don't just want to show the kind of days which is a little bit like fairy 114 00:07:24,280 --> 00:07:31,559 tale stories. Airbnb moved they discontinued their enzyme um uh 115 00:07:31,560 --> 00:07:35,070 UI testing library and and and went to React testing library. 116 00:07:35,120 --> 00:07:41,510 They migrated in six weeks and did a really nice write up about how they did it AI assisted. 117 00:07:41,560 --> 00:07:45,399 Asana did the same thing in about 2 weeks. 118 00:07:45,400 --> 00:07:48,919 There There's some nuance there because they they they spent a few years before doing 119 00:07:48,920 --> 00:07:53,919 it. And again, these are these are practically full rewrites of test suites. 120 00:07:53,920 --> 00:07:56,039 It's It's not a replacing a library. 121 00:07:56,040 --> 00:07:58,199 It's There's some interesting details there. 122 00:07:58,200 --> 00:08:04,039 Uber moved from JUnit 4 to 5 in 4 months, uh which was millions of lines of code. 123 00:08:04,040 --> 00:08:09,559 And again, it it it literally would have taken them just years clog-slogging away beforehand. 124 00:08:09,560 --> 00:08:12,800 So, these things are just cheaper and better and easier. 125 00:08:13,120 --> 00:08:16,670 Agentic generated issues are on the the rise. 126 00:08:16,720 --> 00:08:19,869 Fresh data from Linear shows that on their platform, 127 00:08:19,919 --> 00:08:23,559 the white line is human-created issues tickets, 128 00:08:23,560 --> 00:08:26,519 and the blue line is agentic or MCB-created tickets. 129 00:08:26,520 --> 00:08:28,030 And for the first time, 130 00:08:28,080 --> 00:08:32,119 more agentic-created tickets are happening on their platform than human-created ones. 131 00:08:32,120 --> 00:08:36,359 That's really And you can see the sharp rise in the last few months. 132 00:08:36,640 --> 00:08:43,759 Um we're al- also seeing This is about uh agentic pull 133 00:08:43,760 --> 00:08:45,959 request agent only pull requests are on the rise. 134 00:08:45,960 --> 00:08:52,789 So, just in July and August, agentic pull requests are are spiking. 135 00:08:52,839 --> 00:08:56,559 Uh Sorry, agent-only reviewed pull requests with no human reviews, 136 00:08:56,560 --> 00:08:59,120 which is also very interesting trend. 137 00:08:59,640 --> 00:09:07,199 Uh This is data source from GitHub on how agent-generated PRs are just exploding. 138 00:09:07,200 --> 00:09:09,159 In January, there were 7.7 million. 139 00:09:09,160 --> 00:09:11,639 The The number is not as important than the the trend. 140 00:09:11,640 --> 00:09:13,879 Nearly 10x increase by August. 141 00:09:13,880 --> 00:09:17,560 Agent only So, this is the agentic software factory at work. 142 00:09:17,760 --> 00:09:22,279 And skills usage is also going more mainstream. 143 00:09:22,280 --> 00:09:22,959 More people are doing it. 144 00:09:22,960 --> 00:09:25,070 This is data from Factory AI, 145 00:09:25,120 --> 00:09:31,079 uh where they're seeing about 80-something percent of their users use AI skills. 146 00:09:31,080 --> 00:09:34,919 So, it's just again going kind of mainstream from compared to February when it was 30-something 147 00:09:34,920 --> 00:09:39,279 percent. Cost is is a big concern for a lot of you. 148 00:09:39,280 --> 00:09:41,190 Like you you will know this. 149 00:09:41,240 --> 00:09:43,839 Uh but back in back in May, 150 00:09:43,840 --> 00:09:47,919 uh there was already a widespread trend about companies worried about cost. 151 00:09:47,920 --> 00:09:52,079 This was around when it was made it news that Uber blew blew through its its annual budget 152 00:09:52,080 --> 00:09:56,999 for AI and Claude code was was uh becoming more popular. 153 00:09:57,000 --> 00:09:59,040 AI was getting more expensive. 154 00:09:59,800 --> 00:10:01,830 Now by September, 155 00:10:01,880 --> 00:10:07,119 a lot of the large companies have managed to take control of uh their cost curve and 156 00:10:07,120 --> 00:10:13,119 they're cutting back cost by took per per token cost about 50% keeping cost flat after 157 00:10:13,120 --> 00:10:16,800 investing in open models, smarter routing, and so on. 158 00:10:18,000 --> 00:10:21,840 Another interesting trend is this the disappearance of engineering specializations. 159 00:10:22,280 --> 00:10:27,119 Uh inside OpenAI talk with with with Solomon uh Chachere, 160 00:10:27,120 --> 00:10:28,239 a ChatGPT's head of engineering, 161 00:10:28,240 --> 00:10:32,359 and he said that although specializations in engineering would he suspected it would 162 00:10:32,360 --> 00:10:36,159 go away, he still surprised how quickly it has happened. 163 00:10:36,160 --> 00:10:40,190 Last years they were still looking for iOS and Android specialized engineers, 164 00:10:40,240 --> 00:10:46,390 this year they're not because with Codex in in in their case or LLMs, 165 00:10:46,440 --> 00:10:49,599 they feel that anyone can do the work at a good enough level, 166 00:10:49,600 --> 00:10:50,799 which is a very very interesting one. 167 00:10:50,800 --> 00:10:53,830 And and I've seen this a lot for front-end, mobile, 168 00:10:53,880 --> 00:10:57,359 I mean most engineers have been familiar with back-end, but the same thing. 169 00:10:57,360 --> 00:11:00,240 A front-end engineer can do more back-end now. 170 00:11:02,360 --> 00:11:05,710 Projects are increasingly being done by one or two engineers. 171 00:11:05,760 --> 00:11:08,599 So, inside Anthropic I talked with Catherine Less, 172 00:11:08,600 --> 00:11:13,279 head of engineering at for Claude platform, and she she just put it out so simple. 173 00:11:13,280 --> 00:11:14,479 She said, "On an individual project, 174 00:11:14,480 --> 00:11:17,599 you cannot often have more than one or two engineers working on it. 175 00:11:17,600 --> 00:11:20,599 This is because these engineers are already running several agents. 176 00:11:20,600 --> 00:11:23,838 And so as an engineer, you're already fighting against your agents, 177 00:11:23,839 --> 00:11:25,759 which those agents are stepping on each other's toes. 178 00:11:25,760 --> 00:11:27,399 So, there's only, you know, 179 00:11:27,400 --> 00:11:31,439 maximum one other engineer you can do you can take who will also bring their agents cuz 180 00:11:31,440 --> 00:11:33,079 now you're looking at you know, 181 00:11:33,080 --> 00:11:38,479 a lot of individuals doing it and one person projects are becoming so much more common 182 00:11:38,480 --> 00:11:42,040 in inside of startups who I'm talking with by the way. 183 00:11:43,200 --> 00:11:47,479 Some other other other other changes uh teams are getting smaller. 184 00:11:47,480 --> 00:11:48,479 We're seeing this everywhere. 185 00:11:48,480 --> 00:11:52,870 This kind of comes logically from the the points before. 186 00:11:52,920 --> 00:11:56,479 There's less junior hiring happening across the industry. 187 00:11:56,480 --> 00:11:59,599 It's it's falling not dramatically, but it just keeps going slower. 188 00:11:59,600 --> 00:12:03,159 We're seeing data up inside of pragmatic engineering elsewhere about intern hiring just 189 00:12:03,160 --> 00:12:08,840 keeping going and lower and lower and agentic infra is finally its own discipline. 190 00:12:09,040 --> 00:12:10,799 So, this was a lot of changes. 191 00:12:10,800 --> 00:12:12,799 I I mean we went through 16 different things. 192 00:12:12,800 --> 00:12:14,119 I I was just throwing things at you. 193 00:12:14,120 --> 00:12:16,679 I I was trying to keep it concise, 194 00:12:16,680 --> 00:12:21,999 but I just so many different things are all changing and these changes I want to stress 195 00:12:22,000 --> 00:12:24,070 out. These are the last 12 months. 196 00:12:24,120 --> 00:12:29,710 Before none of these were really true maybe except for the junior hiring one. 197 00:12:29,760 --> 00:12:31,879 So, what did not change because you know, 198 00:12:31,880 --> 00:12:35,799 the industry is is not like we're we're not pulling a meta where we're we're trying to 199 00:12:35,800 --> 00:12:39,999 disrupt completely ourselves, but it is happening. 200 00:12:40,000 --> 00:12:42,030 Some things have not changed though. 201 00:12:42,080 --> 00:12:46,639 Number one is the teams is remained as the unit of work. 202 00:12:46,680 --> 00:12:47,838 Inside of Entropic, 203 00:12:47,839 --> 00:12:53,159 I was pretty pleased to hear that teams are a big thing and again Caitlin told me this. 204 00:12:53,160 --> 00:12:55,319 She said one thing I've heard from some people is oh, 205 00:12:55,320 --> 00:12:59,719 we have two humans and a bunch of agents and then she says like this is not where at. 206 00:12:59,720 --> 00:13:03,190 I still have teams whose job is to own a piece of software, 207 00:13:03,240 --> 00:13:08,838 own on call and so on and while each of these humans is supercharged by AI, 208 00:13:08,839 --> 00:13:10,959 the size and the shape is still similar. 209 00:13:10,960 --> 00:13:13,190 They still have two pizza teams. 210 00:13:13,240 --> 00:13:16,039 Like before. Same thing by the way with OpenAI. 211 00:13:16,040 --> 00:13:18,599 Same thing and when you look at any successful agentic product, 212 00:13:18,600 --> 00:13:19,799 it's it's with a team, right? 213 00:13:19,800 --> 00:13:21,350 Like people go on vacation. 214 00:13:21,400 --> 00:13:24,039 They they create energy with each other. 215 00:13:24,040 --> 00:13:25,239 They inspire each other. 216 00:13:25,240 --> 00:13:26,279 They hold a high bar. 217 00:13:26,280 --> 00:13:27,599 They keep each other accountable. 218 00:13:27,600 --> 00:13:28,679 This is not going anywhere. 219 00:13:28,680 --> 00:13:32,399 Don't let anyone tell you that the you know two people and a bunch of agents, 220 00:13:32,400 --> 00:13:33,679 that's not going to be a team. 221 00:13:33,680 --> 00:13:36,680 That will be a one one off project maybe. 222 00:13:37,080 --> 00:13:39,670 Planning has remained for complex work. 223 00:13:39,720 --> 00:13:41,919 Uh inside of Anthropic for example, 224 00:13:41,920 --> 00:13:46,319 they Kathleen told me that for their more complex infrastructure related products, 225 00:13:46,320 --> 00:13:51,159 they planning hasn't changed much from for example like you know years back when she 226 00:13:51,160 --> 00:13:53,470 worked at Stripe and so on. 227 00:13:53,520 --> 00:13:57,680 And building the wrong thing is still a waste even when building is faster. 228 00:13:57,920 --> 00:14:00,999 Planning is especially important for infra complexity aligning teams communicating with 229 00:14:01,000 --> 00:14:03,240 customers and I think this is not changing. 230 00:14:03,360 --> 00:14:06,160 Testing validation remains important. 231 00:14:06,280 --> 00:14:11,799 Uh Jared Summer on the the we know this but this is for example Jared Summer on on the 232 00:14:11,800 --> 00:14:15,470 bun team who rewrote bun inside of Anthropic. 233 00:14:15,520 --> 00:14:19,559 He said that the AI writes pretty much all code but we have all write the AI write pretty 234 00:14:19,560 --> 00:14:22,919 much all tests. Before AI for production ready software, 235 00:14:22,920 --> 00:14:27,079 we spent about the same time writing tests as we did on writing code and this has not 236 00:14:27,080 --> 00:14:29,360 changed. The the ratio has not changed. 237 00:14:29,400 --> 00:14:34,159 And he is stressing that you need to have a way to trust your code and tests are he says 238 00:14:34,160 --> 00:14:35,679 are probably the best way to do so. 239 00:14:35,680 --> 00:14:40,359 There are of course different layers of testing kind of formal verification is also starting 240 00:14:40,360 --> 00:14:44,240 to become more popular but but we need to validate this this code that we're producing. 241 00:14:45,040 --> 00:14:45,919 This is an interesting one. 242 00:14:45,920 --> 00:14:50,119 I keep hearing about how non-engineers are now able to ship production code. 243 00:14:50,120 --> 00:14:55,030 And I've been asking around the past week everyone if this is happening. 244 00:14:55,080 --> 00:14:56,510 No, it's it's not. 245 00:14:56,560 --> 00:14:58,479 Uh they're not shipping to prod. 246 00:14:58,480 --> 00:14:59,950 They are building websites. 247 00:15:00,000 --> 00:15:04,838 They are tagging in Slack saying oh this is a bug can you look into it and sometimes 248 00:15:04,839 --> 00:15:09,239 that Slack bot might create a pull request but those people typically don't even know 249 00:15:09,240 --> 00:15:11,879 what a pull request is and they are not shipping production code. 250 00:15:11,880 --> 00:15:16,838 They are alerting the engineering team to bugs but engineers every company every startup 251 00:15:16,839 --> 00:15:19,359 from Linear to Cursor to OpenAI I've seen, 252 00:15:19,360 --> 00:15:24,279 you don't have non-engineers push production ready code in any shape or form. 253 00:15:24,280 --> 00:15:28,520 So, this is still not happening and I'm I'm not sure if it will, by the way. 254 00:15:28,720 --> 00:15:30,439 Uh and then this is interesting. 255 00:15:30,440 --> 00:15:32,439 Old patterns are being rediscovered again. 256 00:15:32,440 --> 00:15:34,639 I talked with Matt Pocock, 257 00:15:34,640 --> 00:15:38,270 uh the creator of the GrilMe scale uh and a well-known educator, 258 00:15:38,320 --> 00:15:42,959 and he told me that after he discovered the term tracer bullets from the Pragmatic Programmer 259 00:15:42,960 --> 00:15:45,910 book, he started to dive into different books. 260 00:15:45,960 --> 00:15:50,359 He was looking for keywords for leading words where the agent does better work, 261 00:15:50,360 --> 00:15:51,959 and he started to read the old books. 262 00:15:51,960 --> 00:15:55,719 He came across the philosophy of software design where he got into like deep modules 263 00:15:55,720 --> 00:15:59,239 and shallow modules, and he figured that when he says these words to the AI, 264 00:15:59,240 --> 00:16:01,110 it writes better code. 265 00:16:01,160 --> 00:16:04,719 Who would have thought, like on one end it was trained on it, 266 00:16:04,720 --> 00:16:10,590 and on the other hand those methods, I guess, work even with guiding the AI. 267 00:16:10,640 --> 00:16:11,840 Fascinating. 268 00:16:13,280 --> 00:16:14,480 So, 269 00:16:14,920 --> 00:16:16,799 we went over what changed, what did not change. 270 00:16:16,800 --> 00:16:18,680 I'll talk about what broke. 271 00:16:19,320 --> 00:16:25,559 Code code quantity and and our assumptions about it have just gone out the the wall. 272 00:16:25,560 --> 00:16:26,919 We see the GitHub outages, 273 00:16:26,920 --> 00:16:30,680 and they are publishing to their credit data on on what's happening. 274 00:16:30,920 --> 00:16:36,190 This speed of how many pull requests are are are pushed per month, the commits, 275 00:16:36,240 --> 00:16:39,639 the new repos, in our 3-year timeline, they're up about 5x, 276 00:16:39,640 --> 00:16:44,630 but in just in the last like 6 months, they're they're up by so much more. 277 00:16:44,680 --> 00:16:51,199 Uh data that GitHub gave me actually showed that they're Yeah, 278 00:16:51,200 --> 00:16:54,799 so they're seeing again that 10x increase in agentic-only pull requests, 279 00:16:54,800 --> 00:16:57,479 but but in last 2 months, pull requests have gone up 2x, 280 00:16:57,480 --> 00:16:58,719 that includes everything as well. 281 00:16:58,720 --> 00:17:00,960 It's it's just speeding up really quickly. 282 00:17:01,160 --> 00:17:04,318 Code reviews are are just done for. 283 00:17:04,319 --> 00:17:08,750 I I talked with an engineer who asked not to put his name on on here, 284 00:17:08,800 --> 00:17:10,599 but he said what a lot of people are saying. 285 00:17:10,600 --> 00:17:13,119 Everyone is playing the theater of doing reviews, 286 00:17:13,120 --> 00:17:15,479 but with the volume of changes get thrown your way, 287 00:17:15,480 --> 00:17:18,399 he observed that people just find the path of least resistance. 288 00:17:18,400 --> 00:17:19,958 Some teammates just give up on this too. 289 00:17:19,959 --> 00:17:21,990 Looks good to me on all of them. 290 00:17:22,040 --> 00:17:26,239 Inside this sort of there gradually starting to phase out code reviews themselves because 291 00:17:26,240 --> 00:17:26,958 it just doesn't work. 292 00:17:26,959 --> 00:17:29,399 Like I'm I'm start I just call these zombie code reviews. 293 00:17:29,400 --> 00:17:30,199 This is happening everywhere. 294 00:17:30,200 --> 00:17:34,270 If if your team is still doing code reviews, you're probably not doing code reviews. 295 00:17:34,320 --> 00:17:38,519 If you're unless you've you've cut cut out on you're not doing code reviews on 80% of 296 00:17:38,520 --> 00:17:39,639 them or you're not mandating them. 297 00:17:39,640 --> 00:17:43,359 Inside of Ramp, uh they have started to say like, "Look, 298 00:17:43,360 --> 00:17:47,190 we're not going to we're going to trust you to not do code reviews for certain non-critical 299 00:17:47,240 --> 00:17:50,000 parts of code." And I think more companies will do that. 300 00:17:50,760 --> 00:17:53,839 Quality and reliability is you can see this on every software across. 301 00:17:53,840 --> 00:17:55,190 It's just breaking. 302 00:17:55,240 --> 00:17:56,079 Everything is broken. 303 00:17:56,080 --> 00:17:57,759 Some some people can read about talk about it. 304 00:17:57,760 --> 00:17:58,919 You you look at different sites. 305 00:17:58,920 --> 00:18:05,479 I see it all the time in just the sites I use uh from from Spotify to Substack to a lot 306 00:18:05,480 --> 00:18:08,390 of a lot lot lot lot of paper cuts. 307 00:18:08,440 --> 00:18:12,039 Clearly people are are shipping more with agents or agents are shipping more and it's 308 00:18:12,040 --> 00:18:15,030 it's you can see it everywhere right now. 309 00:18:15,080 --> 00:18:15,599 This is interesting. 310 00:18:15,600 --> 00:18:17,040 Infra capacity. 311 00:18:17,960 --> 00:18:19,839 We know there's a GPU shortage across the industry, 312 00:18:19,840 --> 00:18:22,319 but unless you're an AI lab or an inference company, 313 00:18:22,320 --> 00:18:24,719 you're not going to worry too much about it. 314 00:18:24,720 --> 00:18:27,559 Or an AI startup that wants to get GPUs, then you will be worried. 315 00:18:27,560 --> 00:18:30,359 There's a memory shortage, which means memory is more expensive, but again, 316 00:18:30,360 --> 00:18:32,879 most of us will not see too much of that. 317 00:18:32,880 --> 00:18:35,520 What I was interested to it's it 318 00:18:35,920 --> 00:18:40,919 very surprised to hear there's CPU shortage as well for cloud companies that are running 319 00:18:40,920 --> 00:18:42,159 just you know, normal workloads. 320 00:18:42,160 --> 00:18:45,279 It's starting to get harder and harder to secure your CPUs. 321 00:18:45,280 --> 00:18:49,079 I I covered this in the Pragmatic Engineer just very recently with a bunch of examples 322 00:18:49,080 --> 00:18:52,000 including from Turbo Buffer from Anthropic. 323 00:18:52,560 --> 00:18:57,719 The ordering a CPU or a server used to take 1 to 2 weeks of of back ordering time. 324 00:18:57,720 --> 00:18:59,399 It's now up to 6 months. 325 00:18:59,400 --> 00:19:02,879 Inside of cloud providers, you cannot if you're a midsize customer, 326 00:19:02,880 --> 00:19:07,719 you cannot reserve CPUs in certain regions or in some regions then no more orders can 327 00:19:07,720 --> 00:19:09,070 be taken at all. 328 00:19:09,120 --> 00:19:14,599 And some companies are paying for CPU capacity that is not arrived is not will not arrive 329 00:19:14,600 --> 00:19:17,679 until December and they're paying for it cuz that's the only way they can get it. 330 00:19:17,680 --> 00:19:19,280 It's it's manic. 331 00:19:21,120 --> 00:19:24,000 Another thing that broke is our focus and our productivity. 332 00:19:24,720 --> 00:19:28,270 I just took one quote from Dima also at Linear. 333 00:19:28,320 --> 00:19:32,239 Um and I had a really good conversation with he he said like look Dima is an amazing 334 00:19:32,240 --> 00:19:34,439 engineer from from before AI as well. 335 00:19:34,440 --> 00:19:36,439 And he told me there's just a lot of context switching. 336 00:19:36,440 --> 00:19:39,270 There's always another agent that waits your response. 337 00:19:39,320 --> 00:19:40,879 And also it's a lot more work. 338 00:19:40,880 --> 00:19:41,399 He was really honest. 339 00:19:41,400 --> 00:19:45,639 He said like look when AI was ramping up I felt more productive than my peers cuz I did 340 00:19:45,640 --> 00:19:48,710 it in an hour or two what it took them a day and it was awesome. 341 00:19:48,760 --> 00:19:53,350 However, now this has become the baseline expectation that everyone does parallel things. 342 00:19:53,400 --> 00:19:57,560 And so he ends up feeling as having more work compared to pre-AI. 343 00:19:57,800 --> 00:20:00,319 It's worth reflecting if if you're feeling like that cuz a lot of people are feeling 344 00:20:00,320 --> 00:20:01,680 like this. 345 00:20:02,960 --> 00:20:05,400 Finally, engineering leadership. 346 00:20:06,240 --> 00:20:09,959 Yeah. Um there's something going on with with with engineering leadership. 347 00:20:09,960 --> 00:20:13,119 A lot of you are here so I'm I'm glad that you're you've not walked out the door but 348 00:20:13,120 --> 00:20:14,799 a lot of people have walked out the door. 349 00:20:14,800 --> 00:20:15,719 I did a deep dive on this. 350 00:20:15,720 --> 00:20:20,679 I talked with about like 20 20 25 engineering leaders who are either taking a career 351 00:20:20,680 --> 00:20:22,879 break or will be taking a career break. 352 00:20:22,880 --> 00:20:26,759 And there's a lot of reasons for this but the the main one I'll just list it briefly. 353 00:20:26,760 --> 00:20:30,630 The job of an engineering leader just got a lot worse a lot harder. 354 00:20:30,680 --> 00:20:34,599 Some of the some of the because of founder mode, expectations with AI. 355 00:20:34,600 --> 00:20:37,479 Some founders are pushing you to fire half the team. 356 00:20:37,480 --> 00:20:42,159 The other are pushing you to show revenue going 2x when AI is going to go 2x and so on. 357 00:20:42,160 --> 00:20:43,439 It's unrealistic expectations. 358 00:20:43,440 --> 00:20:45,430 It's just harder to get job. 359 00:20:45,480 --> 00:20:50,119 Startups many startups are losing and leaders see that they're losing. 360 00:20:50,120 --> 00:20:54,679 Your equity is getting worthless or the startup is on path to bankruptcy so you just 361 00:20:54,680 --> 00:20:57,710 you try to turn around you cannot you check out. 362 00:20:57,760 --> 00:21:01,470 A lot of startups are not adopting AI workflows fast enough. 363 00:21:01,520 --> 00:21:06,599 And if you are in this room you will know that career stability is also important. 364 00:21:06,600 --> 00:21:11,719 You're working at a company and you might work here in two years or three years or four 365 00:21:11,720 --> 00:21:14,310 years time, but there's a good chance you won't. 366 00:21:14,360 --> 00:21:19,839 What you want to make sure is if this doesn't work out or it's it's one of the reasons 367 00:21:19,840 --> 00:21:22,880 above, you can go and work at another place. 368 00:21:22,920 --> 00:21:24,559 Carry on being an engineer leader. 369 00:21:24,560 --> 00:21:28,279 To do that, your current company should adopt AI practice. 370 00:21:28,280 --> 00:21:31,880 You should be changing how the world is changing and when it's not, 371 00:21:31,920 --> 00:21:33,479 that's what I see a lot of engineer leaders quit. 372 00:21:33,480 --> 00:21:38,239 This is slow-moving organizations where you are stuck working in the same way as in 2022 373 00:21:38,240 --> 00:21:43,440 for the next two or three years, you will go in the career kind of hole. 374 00:21:44,320 --> 00:21:45,279 Another one, smaller teams. 375 00:21:45,280 --> 00:21:47,639 With smaller teams, you need fewer leaders. 376 00:21:47,640 --> 00:21:50,079 A lot of smaller teams come with middle management being cut as well. 377 00:21:50,080 --> 00:21:52,720 Sometimes there are just a lot less need for this. 378 00:21:53,760 --> 00:21:56,679 Fractional CTO work is becoming surprisingly popular, 379 00:21:56,680 --> 00:22:00,599 especially in metro hubs like New York, San Francisco, maybe even Seattle. 380 00:22:00,600 --> 00:22:03,639 So, some of these engineer leaders who quit voluntarily, 381 00:22:03,640 --> 00:22:06,239 they just don't want to go back right now cuz it's too messy, 382 00:22:06,240 --> 00:22:08,079 but they will do fractional CTO work. 383 00:22:08,080 --> 00:22:10,039 And all these fractional CTOs want to be hired full-time, 384 00:22:10,040 --> 00:22:11,319 but they don't want to go back full-time. 385 00:22:11,320 --> 00:22:13,480 It's a very interesting trend. 386 00:22:13,840 --> 00:22:15,719 Also, I mean, elephant in the room, 387 00:22:15,720 --> 00:22:19,479 at the AI startups will pay way more total compensation, 388 00:22:19,480 --> 00:22:24,079 sometimes even cash compensation, than non-AI startups play play execs. 389 00:22:24,080 --> 00:22:29,239 You see how Anthropic is getting the CTOs of every single major publicly traded company 390 00:22:29,240 --> 00:22:31,879 to join as an IC or someone with a small team. 391 00:22:31,880 --> 00:22:32,959 They're also paying more. 392 00:22:32,960 --> 00:22:34,920 That's also one of the open secrets. 393 00:22:35,200 --> 00:22:38,119 A lot of leaders like you are realize your hands-on. 394 00:22:38,120 --> 00:22:39,999 If you want to do a startup, now is a perfect time. 395 00:22:40,000 --> 00:22:43,559 You can probably also reach VC funding if you if you have the background and you have 396 00:22:43,560 --> 00:22:47,510 the pedigree. And finally, some people are just burnt out. 397 00:22:47,560 --> 00:22:49,879 Feeling it's a good time to take a break, see things settle, 398 00:22:49,880 --> 00:22:51,800 and then come back a bit later. 399 00:22:53,240 --> 00:22:57,760 So, what comes next after all of this? 400 00:22:58,400 --> 00:23:02,839 Here are a couple of things that I'm pretty pretty certain that are happening and will 401 00:23:02,840 --> 00:23:06,680 happen. First one is just cloud agents and harnesses. 402 00:23:07,560 --> 00:23:13,399 Everyone with some engine resources are doing it for many reasons, but with ramp, 403 00:23:13,400 --> 00:23:16,079 ramp is doing one of the most advanced right now, 404 00:23:16,080 --> 00:23:19,919 one of the most advanced agentic harnesses that works really well for them, 405 00:23:19,920 --> 00:23:22,159 way better than Codex or Cloud Code or any of them. 406 00:23:22,160 --> 00:23:25,639 And I I talked with them in detail on why they're doing it and you they do it for three 407 00:23:25,640 --> 00:23:28,999 reasons. One, local machines are a lot more limited than the cloud. 408 00:23:29,000 --> 00:23:32,039 Two, they can actually just they they're huge on front-end. 409 00:23:32,040 --> 00:23:35,399 They build better front-end tooling in the cloud than they have on local machines. 410 00:23:35,400 --> 00:23:40,919 And three, they wanted to have remote developer environments anyway and this just kind 411 00:23:40,920 --> 00:23:43,519 of like it kills two birds with one stone. 412 00:23:43,520 --> 00:23:45,879 The only problem with this approach is when you have a CPU shortage, 413 00:23:45,880 --> 00:23:49,400 it might be a little bit trickier to secure, but again, it's a small detail. 414 00:23:49,440 --> 00:23:52,999 So many other companies are doing it and I so many will follow. 415 00:23:53,000 --> 00:23:55,199 Uh this is again Dima from Linear. 416 00:23:55,200 --> 00:23:59,999 I asked Dima like, "Hey Dima, do you what do you think about cloud versus local?" 417 00:24:00,000 --> 00:24:00,879 And he said like, 418 00:24:00,880 --> 00:24:05,439 "He was such a big fan of a local development and the industry is still lagging behind 419 00:24:05,440 --> 00:24:09,359 on on dev pods, that was something we did at Uber, and tooling for cloud development." 420 00:24:09,360 --> 00:24:13,519 However, he changed his mind about the future because he now sees cloud as the future 421 00:24:13,520 --> 00:24:17,599 cuz whenever he kicks off stuff running in the cloud, it doesn't hog his machine. 422 00:24:17,600 --> 00:24:19,279 It's faster, easier, etc. 423 00:24:19,280 --> 00:24:21,280 This is this is definitely future. 424 00:24:21,520 --> 00:24:24,430 We are we will stop reading the code. 425 00:24:24,480 --> 00:24:25,559 This doesn't just come from me, 426 00:24:25,560 --> 00:24:30,470 it comes from Charity Majors who is such a big skeptic and I love her. 427 00:24:30,520 --> 00:24:32,799 But we we on the podcast we're talking about like, 428 00:24:32,800 --> 00:24:36,519 "What would it take for you to be comfortable shipping code without reading and understanding 429 00:24:36,520 --> 00:24:39,319 it?" Because this is what ops have been doing. 430 00:24:39,320 --> 00:24:44,070 This is what QA have been doing and this is it's not an it's not an if, it's a when. 431 00:24:44,120 --> 00:24:45,360 It's coming. 432 00:24:46,600 --> 00:24:50,559 We're all building we will be building out and are building out new type of internal 433 00:24:50,560 --> 00:24:54,479 infra. This means eval's will be part of CICD if it's not already. 434 00:24:54,480 --> 00:24:57,439 We will have these agentic factories that we keep iterating on. 435 00:24:57,440 --> 00:25:00,999 Agents will be part of deployments, observability, incident management, and so on. 436 00:25:01,000 --> 00:25:05,350 And again, it's it's it's already happening and will just keep happening even more. 437 00:25:05,400 --> 00:25:07,799 We will have a refactoring wave of migrations, rewriting. 438 00:25:07,800 --> 00:25:14,599 I mean, look, when it's when it's so easy to or cheaper to to refactor, why not do it? 439 00:25:14,600 --> 00:25:19,880 Plus, it's great exercise to like use risk-free some some of these agentic skills. 440 00:25:21,040 --> 00:25:25,799 AI fluency and positivity will be a baseline for hiring and to be hired. 441 00:25:25,800 --> 00:25:30,599 I talked with a engineering director at Series A startup who said that they have started 442 00:25:30,600 --> 00:25:33,199 to screen for AI positivity in their in in their hiring process. 443 00:25:33,200 --> 00:25:35,199 He said, "We just don't want people to join. 444 00:25:35,200 --> 00:25:39,119 We we want people to join who wants us help to push the limits of what we can build with 445 00:25:39,120 --> 00:25:40,879 AI." As a leader, just you know, 446 00:25:40,880 --> 00:25:44,239 just know this is the case and that this is true when you're hiring and also it's true 447 00:25:44,240 --> 00:25:45,920 when you are interviewing. 448 00:25:46,880 --> 00:25:51,959 And getting to AI maturity is something that so many companies will be doing. 449 00:25:51,960 --> 00:25:53,519 A startup might think they're already there. 450 00:25:53,520 --> 00:25:54,679 A larger company will be pushing for it. 451 00:25:54,680 --> 00:25:58,150 This is Lara Hogan, who's at the Pragmatic Summit in February. 452 00:25:58,200 --> 00:26:01,279 She's saying that orgs who are moving faster and maturing do something different. 453 00:26:01,280 --> 00:26:04,119 They begin with a business outcome and work backwards from there, 454 00:26:04,120 --> 00:26:07,999 and they build an agentic system that reduces the handoffs, increases learning, 455 00:26:08,000 --> 00:26:09,279 and removes friction. 456 00:26:09,280 --> 00:26:14,630 Now, all of this is a little bit high-level, but it is the the right direction. 457 00:26:14,680 --> 00:26:20,919 And she's saying that the wrong way to do things is one of the easy way to do is is do 458 00:26:20,920 --> 00:26:23,079 an individual-level simple automations, 459 00:26:23,080 --> 00:26:28,599 but what's hard to do but what all us leaders should be doing is doing team-level agentic 460 00:26:28,600 --> 00:26:30,479 systems or even company-level agentic systems. 461 00:26:30,480 --> 00:26:34,990 That's why you'll see all of that that AI infra happening, cuz that's how you do it. 462 00:26:35,040 --> 00:26:40,550 Finally, engineers with deep domain knowledge will be so much more in demand. 463 00:26:40,600 --> 00:26:44,639 Titus Winters, yesterday, author of Software Engineering at Google, 464 00:26:44,640 --> 00:26:46,879 told me something such interesting. 465 00:26:46,880 --> 00:26:48,359 I I I love Titus. 466 00:26:48,360 --> 00:26:51,430 He said, "What you need to succeed anywhere? 467 00:26:51,480 --> 00:26:53,670 Intelligence, wisdom, and charisma. 468 00:26:53,720 --> 00:26:57,119 You need Intelligence meaning you know how to do it. 469 00:26:57,120 --> 00:27:01,599 Wisdom means you know what to do and charisma means you know you convince others to do 470 00:27:01,600 --> 00:27:06,479 it. Now when intelligence becomes commonplace or you know with AI, 471 00:27:06,480 --> 00:27:10,070 the how to do it becomes just kind of very cheap, so widespread, 472 00:27:10,120 --> 00:27:13,600 wisdom and charisma become much more important. 473 00:27:13,800 --> 00:27:16,880 As leaders especially remember this one. 474 00:27:17,360 --> 00:27:21,600 And I'm not saying you know don't focus on intelligence at all, but he has a point here. 475 00:27:22,320 --> 00:27:26,600 So in this new world, how do you thrive as an engineering leader? 476 00:27:26,720 --> 00:27:29,320 My advice is future-proof your career. 477 00:27:30,200 --> 00:27:32,679 Get hands-on if you're not already, but most of you are. 478 00:27:32,680 --> 00:27:34,959 Stay hands-on building software shipping to production. 479 00:27:34,960 --> 00:27:35,879 You will hear stories. 480 00:27:35,880 --> 00:27:36,959 Will Larson in the other room, 481 00:27:36,960 --> 00:27:41,279 he's chief product and technology officer at a 50-person engineering organization. 482 00:27:41,280 --> 00:27:46,559 He has shipped more uh code into production the past 15 months than in the past like 483 00:27:46,560 --> 00:27:48,910 5 or 10 years combined. 484 00:27:48,960 --> 00:27:51,959 And if he's doing it, you'll you can do that as well. 485 00:27:51,960 --> 00:27:55,750 Build out AI infra systems that help your team and your company succeed. 486 00:27:55,800 --> 00:27:57,399 And you'll also be hands-on with this. 487 00:27:57,400 --> 00:28:00,159 Use AI to remove friction from your own work, from your team's work, 488 00:28:00,160 --> 00:28:01,479 and for your company's work. 489 00:28:01,480 --> 00:28:02,839 And don't outsource your learning. 490 00:28:02,840 --> 00:28:04,879 Use AI to get better. 491 00:28:04,880 --> 00:28:06,239 Don't for for the important stuff, 492 00:28:06,240 --> 00:28:10,199 don't kind of hand it off and you know like you're you're not going to get ahead and 493 00:28:10,200 --> 00:28:12,999 and but your peers will that do it. 494 00:28:13,000 --> 00:28:15,999 And finally, accept that you will probably do just do less people management for now. 495 00:28:16,000 --> 00:28:17,559 This is not really the time. 496 00:28:17,560 --> 00:28:18,919 ICs don't expect it as much. 497 00:28:18,920 --> 00:28:20,599 We don't even know what that career looks like. 498 00:28:20,600 --> 00:28:21,359 We will figure it out. 499 00:28:21,360 --> 00:28:22,679 We will get back to it. 500 00:28:22,680 --> 00:28:27,519 But also this is not your main expectation from from above uh and for from founders and 501 00:28:27,520 --> 00:28:32,990 so on. Finally, just lean in, learn more before, and you'll be fine. 502 00:28:33,040 --> 00:28:35,359 With Peter Mattis, I had such an inspiring conversation. 503 00:28:35,360 --> 00:28:37,839 Again, this person built so much incredible software. 504 00:28:37,840 --> 00:28:44,599 He is a CTO of a I think a of 600-ish person organization uh 100 and something engineers 505 00:28:44,600 --> 00:28:46,199 if if I recall correctly. 506 00:28:46,200 --> 00:28:49,519 At some point he went on to being a hands-off engineering leader and now he's back to 507 00:28:49,520 --> 00:28:52,919 hands-on and he told me you just got to get in there. 508 00:28:52,920 --> 00:28:54,719 You need to be using these AI tools all the time. 509 00:28:54,720 --> 00:28:56,479 You need to push yourself to learn. 510 00:28:56,480 --> 00:29:00,679 He says that I feel I've learned more in the past year than in the previous 5 years combined 511 00:29:00,680 --> 00:29:04,439 which is weird cuz he's a wicked experienced software engineer and he's built some amazing 512 00:29:04,440 --> 00:29:10,199 software. And finally, just ask yourself, why did I get into tech when you started? 513 00:29:10,200 --> 00:29:11,910 What why did you? 514 00:29:11,960 --> 00:29:13,279 Peter told me his reason. 515 00:29:13,280 --> 00:29:16,719 He said, I got into software engineer because I like building stuff and now I can build 516 00:29:16,720 --> 00:29:20,239 faster and without some of the compromises I had before. 517 00:29:20,240 --> 00:29:23,680 It's a little bit exhausting but it's also very very exciting. 518 00:29:23,960 --> 00:29:28,359 So with this, just keep on keeping on and check out the Pragmatic Engineer and I'll see 519 00:29:28,360 --> 00:29:30,110 you at the Entisys booth. 520 00:29:30,160 --> 00:29:31,560 Thank you.