1 00:00:07,840 --> 00:00:11,040 I like that. 2 00:00:11,560 --> 00:00:12,600 Bent in neon. 3 00:00:13,880 --> 00:00:17,655 Well, first of all, welcome to Austin is what 4 00:00:17,680 --> 00:00:18,720 I would normally say. 5 00:00:18,920 --> 00:00:21,800 I would say welcome back to myself to Austin. 6 00:00:21,880 --> 00:00:25,880 I was just here about a month ago yapping for 5 hours on a podcast. 7 00:00:25,960 --> 00:00:27,800 It is great to be back here. 8 00:00:27,960 --> 00:00:33,880 But even more exciting for me is to be alive right now 9 00:00:34,480 --> 00:00:38,640 in this moment and sharing it with a wonderful community 10 00:00:39,160 --> 00:00:42,920 around Ruby on Rails and everyone else who is excited about agents. 11 00:00:43,760 --> 00:00:48,000 And if you aren't yet, hopefully you will be by the end of this 12 00:00:48,560 --> 00:00:51,240 delirious talk on the future. 13 00:00:51,920 --> 00:00:55,280 Now, I know that the first instinct someone 14 00:00:55,400 --> 00:01:00,695 might have when they see a Person dancing around being so excited 15 00:01:00,720 --> 00:01:04,640 about this moment is a diagnosis. 16 00:01:06,960 --> 00:01:08,240 AI psychosis. 17 00:01:10,120 --> 00:01:11,440 I prefer a different term. 18 00:01:12,000 --> 00:01:14,760 I have AI delirium. 19 00:01:17,000 --> 00:01:18,560 AI euphoria. 20 00:01:19,200 --> 00:01:24,120 Because this is simply the most exciting thing we have made computers do 21 00:01:24,760 --> 00:01:29,280 in the time I have worked with them, played with them, interacted with them. 22 00:01:30,040 --> 00:01:34,240 But I think actually to wrestle with the fact that on the one 23 00:01:34,280 --> 00:01:38,480 hand, this is tremendously exciting, it is also a little unnerving. 24 00:01:39,640 --> 00:01:43,680 Because we don't know exactly what the final state of this is going to be. 25 00:01:44,680 --> 00:01:50,080 And whenever we try to make a prediction, it is out of date about 20 minutes later. 26 00:01:50,640 --> 00:01:55,200 That's a weird state to be in, even if you are excited for the future, 27 00:01:55,240 --> 00:01:56,800 even if you're excited for the present. 28 00:01:57,360 --> 00:02:03,000 And I fully understand the sense of unease that can create. 29 00:02:04,040 --> 00:02:10,120 So to soothe that somewhat, let's wind back the clock a little bit. 30 00:02:13,440 --> 00:02:17,000 This is a state-of-the-art. 31 00:02:18,480 --> 00:02:21,920 Piece of technology from the 18th century. 32 00:02:22,480 --> 00:02:26,280 If you wanted your portrait taken in the 18th century, 33 00:02:26,640 --> 00:02:32,375 you would sit for hour upon hour upon hour in front of a master painter who would 34 00:02:32,400 --> 00:02:38,400 then go off and at best spend months working on your portrait. 35 00:02:39,240 --> 00:02:44,215 And it was beautiful, is beautiful, amazing legacy of culture that 36 00:02:44,240 --> 00:02:45,760 has been left to all of us now. 37 00:02:46,400 --> 00:02:51,240 That we can see the Lady Waldegrave to this day that Joshua Reynolds 38 00:02:51,600 --> 00:02:53,760 painted in 1781. 39 00:02:54,800 --> 00:02:55,400 Wonderful. 40 00:02:56,400 --> 00:02:58,720 Also, a little exclusive. 41 00:03:00,040 --> 00:03:04,975 Not that many people in the 18th century were able to commission a piece of 42 00:03:05,000 --> 00:03:10,360 themselves like this and pay for an artist to spend months and months and months to 43 00:03:11,360 --> 00:03:12,520 create something like this. 44 00:03:13,520 --> 00:03:15,760 And even worse, in the case of this particular painting, 45 00:03:16,320 --> 00:03:20,000 Reynolds spent months and months on another painting 46 00:03:20,560 --> 00:03:24,960 of the Ladies Waldegrave that was rejected by the client because 47 00:03:25,560 --> 00:03:27,040 a piece of ankle was showing. 48 00:03:27,960 --> 00:03:28,800 It was scandalous. 49 00:03:29,560 --> 00:03:31,760 So he simply had to go back and redo it. 50 00:03:32,440 --> 00:03:35,400 I don't know if anyone has worked with clients before, but they might have some 51 00:03:35,425 --> 00:03:38,280 sympathy for poor Mr. Joshua Reynolds. 52 00:03:38,560 --> 00:03:40,200 And rejected statements of work. 53 00:03:42,360 --> 00:03:45,255 Now, most of the people who could afford to do 54 00:03:45,280 --> 00:03:52,400 this were either part of the upper bourgeoisie or they were outright royals. 55 00:03:53,200 --> 00:04:00,800 Here's a painting that was done 20 years later by Goya of the royal Spanish family. 56 00:04:01,960 --> 00:04:03,820 That was sort of the rate of technology. 57 00:04:05,320 --> 00:04:10,000 Painters had been perfecting the different techniques of capturing portraits of 58 00:04:10,040 --> 00:04:15,015 people for literally hundreds of years and were making very marginal 59 00:04:15,040 --> 00:04:16,440 progress along the way. 60 00:04:17,920 --> 00:04:25,200 Now, between 1801 and the next painting I'm going to show, 80 years went past. 61 00:04:26,200 --> 00:04:34,920 And in that time, the introduction of The camera came to the scene in around 1840. 62 00:04:35,920 --> 00:04:44,080 But here's a painting of the royal Danish family from 1888. 63 00:04:45,400 --> 00:04:46,320 Oh, '86. 64 00:04:47,200 --> 00:04:49,800 By a Danish painter, Laurits Tuxen. 65 00:04:50,680 --> 00:04:52,920 He didn't just spend 3 months painting this. 66 00:04:53,400 --> 00:04:54,800 He spent 3 years. 67 00:04:55,640 --> 00:05:03,680 It is still hanging in the royal squares in Copenhagen, 68 00:05:03,960 --> 00:05:05,060 and you can go see it there. 69 00:05:05,085 --> 00:05:08,000 An amazing achievement, amazing artistry. 70 00:05:08,560 --> 00:05:09,455 Also very functional. 71 00:05:09,480 --> 00:05:12,200 Again, the point was to capture the royal family at that time. 72 00:05:12,840 --> 00:05:18,480 But this was getting towards the end of that particular mode 73 00:05:18,880 --> 00:05:20,440 for artists of that kind. 74 00:05:21,080 --> 00:05:29,775 Because what happened between 1840 and 1900 was that the technology of 75 00:05:29,800 --> 00:05:31,800 capturing portraits got better and better. 76 00:05:33,560 --> 00:05:39,040 And in 1900, the Eastman Kodak Company introduced the Brownie, 77 00:05:39,960 --> 00:05:45,575 the first sort of mass-produced form of photography that was more 78 00:05:45,600 --> 00:05:47,960 generally accessible because the price was coming down. 79 00:05:48,040 --> 00:05:49,540 I think it was about a dollar. 80 00:05:49,640 --> 00:05:51,375 To have your picture taken with this. 81 00:05:51,400 --> 00:05:54,180 And therefore, a lot of people suddenly had their pictures taken. 82 00:05:54,440 --> 00:05:59,400 And all of these great artists realized that the craftsmanship of producing these 83 00:05:59,480 --> 00:06:02,480 grand paintings, you know what, that was gonna be different. 84 00:06:04,000 --> 00:06:05,520 The job was going to be different. 85 00:06:05,680 --> 00:06:09,560 Because suddenly now there was competition from a new form of technology. 86 00:06:12,480 --> 00:06:13,620 And they pivoted. 87 00:06:14,880 --> 00:06:20,120 Laurits Tuxen, along with many other painters in that era, shortly after 88 00:06:20,840 --> 00:06:24,600 the Brownie was introduced, realized that depicting reality 89 00:06:25,160 --> 00:06:31,640 as perfectly as possible was no longer really an economically viable skill. 90 00:06:32,320 --> 00:06:33,480 They needed another skill. 91 00:06:33,520 --> 00:06:34,680 They needed another domain. 92 00:06:35,360 --> 00:06:38,280 And many of them pivoted to different forms of art. 93 00:06:39,120 --> 00:06:41,720 You had Picasso chasing Cubism. 94 00:06:42,560 --> 00:06:47,880 And you had the Danish Skagen painters chasing a form of impressionism 95 00:06:48,440 --> 00:06:50,440 that this painting is an example of. 96 00:06:51,760 --> 00:06:57,720 So we got this explosion of creativity brought upon us by the change 97 00:06:58,160 --> 00:07:05,120 in technology that suddenly people had to find a new way to express 98 00:07:05,240 --> 00:07:07,320 their art because the old way was gone. 99 00:07:08,360 --> 00:07:10,920 Now, what's interesting about this particular 100 00:07:10,960 --> 00:07:15,640 picture is that the lady in the center in the pink dress, 101 00:07:16,320 --> 00:07:19,960 Yvonne Tuxen, is actually my great-grandmother. 102 00:07:22,240 --> 00:07:25,160 Because Laurits Tuxen is my great-great-grandfather. 103 00:07:26,920 --> 00:07:34,080 And here he is depicted in 1921, I believe, shortly before his death. 104 00:07:34,600 --> 00:07:37,800 Death, he embraced the fact of having his own 105 00:07:37,840 --> 00:07:42,840 portrait taken by the new form technology that had rendered his old profession 106 00:07:43,280 --> 00:07:44,740 a different place to work. 107 00:07:48,080 --> 00:07:53,000 52 years after his death, I was born in 1979. 108 00:07:54,680 --> 00:07:58,400 Interestingly enough, that picture taken here is very similar, 109 00:07:58,720 --> 00:08:02,880 actually, to the picture that was taken by Laurits In the 1920s, 110 00:08:03,360 --> 00:08:06,680 the technology hadn't evolved that much because that's what often happens 111 00:08:07,240 --> 00:08:08,960 with progressions of technology. 112 00:08:09,080 --> 00:08:11,400 It comes in starts and stops and fits. 113 00:08:12,200 --> 00:08:18,480 The huge change was the Brownie, the democratization of photography, 114 00:08:19,000 --> 00:08:24,480 and then kind of almost 80 years went by when not a whole lot happened. 115 00:08:24,800 --> 00:08:28,160 We got color photography in the 60s, but. 116 00:08:28,880 --> 00:08:32,280 Not widely distributed enough that apparently I was worth one. 117 00:08:36,120 --> 00:08:41,880 Although we did rectify that a few years later, and here's a color photography 118 00:08:41,920 --> 00:08:46,040 of me, age 4, sitting with my brother. 119 00:08:46,880 --> 00:08:51,120 Now, a couple years later, here's another 120 00:08:52,240 --> 00:08:55,160 photograph of me, and between those 3 photos, 121 00:08:55,840 --> 00:09:01,120 That's about, I think, 10% of my stockpile of photos of my childhood. 122 00:09:01,960 --> 00:09:06,415 Because even though the democratization of this technology had 123 00:09:06,440 --> 00:09:10,375 happened to a point that even we as a family could afford to have 124 00:09:10,400 --> 00:09:12,940 these pictures taken, it was not that widely distributed. 125 00:09:12,965 --> 00:09:16,840 And yes, there was a curve and more pictures were being taken in the 1980s, 126 00:09:17,160 --> 00:09:22,760 as this photo represents, versus the 1900s, but it was still kind of modest. 127 00:09:23,120 --> 00:09:26,760 Because technology moves in fits, starts and stops. 128 00:09:29,120 --> 00:09:35,360 And the camera that those 3 pictures were taken on was very similar to this camera. 129 00:09:35,440 --> 00:09:38,240 This is the Leica 1 from 1925. 130 00:09:39,400 --> 00:09:42,520 In fact, maybe some of those pictures were literally taken with another Leica. 131 00:09:44,120 --> 00:09:45,840 I think they were up to the 7th one by then. 132 00:09:45,960 --> 00:09:49,120 But they weren't that different because the technology was quite similar. 133 00:09:49,920 --> 00:09:52,800 Not much had happened since the great revolution 134 00:09:53,280 --> 00:09:55,720 around the 1900s until the '80s. 135 00:09:57,400 --> 00:10:01,920 Then we gave it a couple more decades, and suddenly everything 136 00:10:02,000 --> 00:10:03,080 changed dramatically again. 137 00:10:04,840 --> 00:10:09,160 Photography went from not just being possible for everyone to have, 138 00:10:09,240 --> 00:10:13,400 but to become completely frictionless, and the number of photos 139 00:10:13,520 --> 00:10:15,520 taken went parabolical. 140 00:10:16,360 --> 00:10:19,520 Because suddenly everyone could afford to have 141 00:10:20,120 --> 00:10:24,600 tens of thousands of pictures of their children, and many of them did. 142 00:10:25,440 --> 00:10:28,320 It simply became too cheap to measure. 143 00:10:29,360 --> 00:10:32,615 These are some of the photographs my wife likes to take when she's 144 00:10:32,640 --> 00:10:33,900 out on her evening walk. 145 00:10:34,160 --> 00:10:40,240 To have wanted to do this even in the '80s would have required a commitment to film 146 00:10:40,600 --> 00:10:44,400 and getting them produced and the rest of it. 147 00:10:44,440 --> 00:10:47,400 And there were people who did that, but not that many, 148 00:10:47,880 --> 00:10:49,700 because the friction was still too high. 149 00:10:52,160 --> 00:11:00,280 Today, in 2026, some 2 trillion photos are taken every year. 150 00:11:01,680 --> 00:11:02,760 That's very different. 151 00:11:02,960 --> 00:11:07,480 It's the same underlying technology, but we've clearly hit an inflection point, 152 00:11:07,560 --> 00:11:10,400 even compared to when The camera phone first came out. 153 00:11:11,920 --> 00:11:16,680 There's gonna be some parallels here to our industry and our profession. 154 00:11:20,080 --> 00:11:24,200 On November 24th, 2025, 155 00:11:25,280 --> 00:11:30,680 we got the Kodak Brownie of our era. 156 00:11:32,200 --> 00:11:34,200 We got Opus 4.5. 157 00:11:37,120 --> 00:11:44,840 AI technology made accessible in a harness that many people could afford to use and 158 00:11:44,880 --> 00:11:51,480 experience for the first time what it's like to create software in 159 00:11:51,560 --> 00:11:53,760 pairing with a new form of intelligence. 160 00:11:55,840 --> 00:11:57,280 This was the tipping point for me. 161 00:11:58,400 --> 00:12:01,520 There was everything before November 24th. 162 00:12:02,280 --> 00:12:03,980 And then there was everything after. 163 00:12:04,080 --> 00:12:09,080 This is going to be the date that history books going forward will mark 164 00:12:09,320 --> 00:12:11,760 as the inflection point for the age of agents. 165 00:12:14,000 --> 00:12:16,455 And as with the Brownie, when that first came out, 166 00:12:16,480 --> 00:12:20,960 there was an explosion of different form factors and new ways 167 00:12:21,080 --> 00:12:23,240 of trying to harness this technology. 168 00:12:23,600 --> 00:12:31,160 And it did not take long until we got Some notion of the same intelligence 169 00:12:31,240 --> 00:12:33,040 in an open-weight format? 170 00:12:34,120 --> 00:12:35,440 Total euphoria. 171 00:12:35,600 --> 00:12:37,480 This is moving so fast. 172 00:12:37,840 --> 00:12:43,440 I can't believe that this alien technology is now accessible in open-weight form. 173 00:12:43,600 --> 00:12:45,160 It just happened a few months later. 174 00:12:45,520 --> 00:12:51,480 I drove Kimi K2.5 In the fast mode for quite a while and absolutely loved it. 175 00:12:51,640 --> 00:12:55,320 An incredible experience to be experiencing This kind of 176 00:12:55,360 --> 00:12:57,560 intelligence at 200 tokens per second. 177 00:12:57,800 --> 00:13:02,080 But then something else happened. 178 00:13:03,560 --> 00:13:07,320 This trough of disillusionment from February through May 179 00:13:07,760 --> 00:13:12,600 where we seemingly continued to get new models, but they kind of sucked. 180 00:13:13,320 --> 00:13:14,820 It was like they were stepping backwards. 181 00:13:14,845 --> 00:13:17,720 And many people were like, what's this Opus 4.6 thing? 182 00:13:17,840 --> 00:13:18,680 I don't know, man. 183 00:13:19,400 --> 00:13:21,060 Just give me the old stuff again. 184 00:13:21,440 --> 00:13:26,720 And a lot of people thought, haha, you thought this was going to continuously 185 00:13:26,800 --> 00:13:28,720 exponentially expand and improve. 186 00:13:29,000 --> 00:13:32,120 But look, we're actually not getting that much better. 187 00:13:32,600 --> 00:13:33,960 I remember a particular release. 188 00:13:34,040 --> 00:13:38,320 I think it was of OpenAI's new model that didn't punch 189 00:13:38,360 --> 00:13:39,700 that much better on the benchmarks. 190 00:13:39,725 --> 00:13:43,600 And the stock price fell or the valuation fell and the market was suddenly, 191 00:13:43,680 --> 00:13:44,960 oh my God, it's over. Yeah. 192 00:13:47,240 --> 00:13:48,580 Because we don't know what's going to happen. 193 00:13:48,605 --> 00:13:55,200 But this trough of disillusionment did not last like it did with photography for 100 194 00:13:55,320 --> 00:13:58,600 years, where the technology barely seemed to improve. 195 00:13:59,640 --> 00:14:01,040 It only lasted until May. 196 00:14:02,440 --> 00:14:03,320 Well, June, really. 197 00:14:04,240 --> 00:14:08,800 Because in June, we got Fable 5 and Mythos. 198 00:14:09,360 --> 00:14:13,375 And clearly, anyone who had a chance to experience that at the 199 00:14:13,400 --> 00:14:17,815 time were disabused of any disillusionment they might have had about whether we 200 00:14:17,840 --> 00:14:20,120 had stalled and whether it was over. 201 00:14:21,320 --> 00:14:26,840 And I'd say this is where my diagnosis really set in. 202 00:14:28,040 --> 00:14:37,000 My psychosis/delirium/euphoria for this explosion of intelligence on tap. 203 00:14:38,200 --> 00:14:41,800 This was the model where I went from, oh, I can tell 204 00:14:42,880 --> 00:14:48,040 an agent to do something for me, to seeing that being something I want to merge. 205 00:14:48,160 --> 00:14:52,815 That was the revelation of Opus, to suddenly having models I could give 206 00:14:52,840 --> 00:15:00,160 problems to or ideas to and see those be solved or see those 207 00:15:00,200 --> 00:15:04,920 expanded through no direction of my own on the implementation. 208 00:15:06,920 --> 00:15:08,840 Mind-blowingly incredible. 209 00:15:09,600 --> 00:15:11,460 And of course, things didn't stop from there. 210 00:15:11,720 --> 00:15:14,800 We have been on this amazing roller coaster now for a couple of months. 211 00:15:15,440 --> 00:15:21,560 GPT-6 Astra came out in September and saved us all from the 212 00:15:22,320 --> 00:15:25,855 dread that Anthropic was going to have a monopoly on this kind of 213 00:15:25,880 --> 00:15:27,040 frontier intelligence. 214 00:15:27,560 --> 00:15:31,000 And then even better than that, just a week later, 215 00:15:31,640 --> 00:15:36,400 we got DeepSeek-4-1 Flash that disabused us of any notion that frontier 216 00:15:36,480 --> 00:15:40,880 intelligence was going to forever be a domain of just some large 217 00:15:40,920 --> 00:15:42,260 American tech companies. 218 00:15:42,680 --> 00:15:44,160 This is incredibly exciting. 219 00:15:47,320 --> 00:15:48,580 But what does it mean? 220 00:15:49,320 --> 00:15:50,820 What does it mean for us as programmers? 221 00:15:50,845 --> 00:15:52,840 What does it mean for us as a profession? 222 00:15:53,640 --> 00:15:56,560 And I think this is why, again, it's worth looking backwards a bit. 223 00:15:56,960 --> 00:16:00,655 Because history holds all these amazing lessons and parallels to our 224 00:16:00,680 --> 00:16:01,840 time if we just look at it. 225 00:16:02,400 --> 00:16:06,855 One of the things that has dominated the discourse in the entire time I've been a 226 00:16:06,880 --> 00:16:11,655 professional developer has been the argument about whether the 10x 227 00:16:11,680 --> 00:16:14,680 programmer was real or not real. 228 00:16:15,320 --> 00:16:20,200 Well, that entire concept came from an article in ACM Magazine from 1968, 229 00:16:21,080 --> 00:16:25,920 where they tested a group of programmers on a variety of different tasks, 230 00:16:26,400 --> 00:16:31,760 and they came up with A table that showed that the worst 231 00:16:31,800 --> 00:16:37,280 programmer and the best programmer had a variance ranging from 5x to 30x, 232 00:16:37,440 --> 00:16:40,880 an average of 10x over that period. 233 00:16:41,920 --> 00:16:46,960 We then spent literally the next 45 years arguing 234 00:16:47,400 --> 00:16:49,080 whether this was actually real or not. 235 00:16:52,680 --> 00:16:54,460 Is there anyone still arguing that? 236 00:16:54,920 --> 00:16:55,960 I don't think so. 237 00:16:56,040 --> 00:16:58,840 Like, that debate has been settled now. 238 00:16:59,120 --> 00:17:03,600 But the interesting thing is, we sort of leaped over a shared conclusion 239 00:17:03,640 --> 00:17:06,360 around the 10x programmer and started thinking, like, 240 00:17:06,920 --> 00:17:12,760 well, if that was 10x with just the human capacity, what must it look like now? 241 00:17:14,560 --> 00:17:16,180 And it's clear we don't know exactly. 242 00:17:16,205 --> 00:17:20,920 We don't know exactly what the kind of acceleration with these tools is possible. 243 00:17:21,440 --> 00:17:25,855 But I'd actually argue that it is less controversial today to claim that there's 244 00:17:25,880 --> 00:17:30,800 a 100x difference between the worst programmer working without these tools and 245 00:17:30,840 --> 00:17:33,080 the best programmer working with these tools. 246 00:17:33,400 --> 00:17:35,375 That's not a very controversial statement. 247 00:17:35,400 --> 00:17:40,560 But it is absolutely revolutionary to the understanding of where we are right now 248 00:17:41,000 --> 00:17:42,080 as a stage of technology. 249 00:17:42,800 --> 00:17:44,420 And I'd actually go a step further. 250 00:17:45,280 --> 00:17:47,340 And maybe this is slightly more controversial. 251 00:17:47,600 --> 00:17:48,620 I don't think so. 252 00:17:49,080 --> 00:17:53,000 The worst programmer without these tools versus the best programmer 253 00:17:53,040 --> 00:17:54,040 with these tools? 254 00:17:55,040 --> 00:17:57,040 That sounds about right. 1,000 times? 255 00:17:57,320 --> 00:17:57,900 Yeah. 256 00:17:58,400 --> 00:18:00,080 Okay. Well, what does that mean? 257 00:18:00,600 --> 00:18:03,760 What does that mean for the competition of companies, 258 00:18:04,440 --> 00:18:06,640 for what we need to get started? 259 00:18:07,320 --> 00:18:10,800 Uh, 8-ball a little fussy. Shake again. 260 00:18:13,440 --> 00:18:18,280 What I know to be true is that in the past 20 months, 261 00:18:19,120 --> 00:18:24,040 I have written half as much code as I did in the previous 21 years. 262 00:18:26,520 --> 00:18:32,320 Now, lines of code is a weird, fuzzy, malleable size. 263 00:18:33,080 --> 00:18:39,520 And one line of Ruby is obviously worth vastly more than one line of 264 00:18:40,240 --> 00:18:46,415 any other language, but particularly Rust or C++ or any of the lower-level 265 00:18:46,440 --> 00:18:49,200 languages that do not have the same abstractions and amenities 266 00:18:49,320 --> 00:18:50,760 that we enjoy in Ruby. 267 00:18:50,840 --> 00:18:57,560 But it is still absolutely undeniable that there is a revolution taking place. 268 00:18:58,880 --> 00:19:04,160 And anyone who's gone through that and seen their own productivity be accelerated 269 00:19:04,920 --> 00:19:06,080 We'll be a little dizzy. 270 00:19:07,280 --> 00:19:08,740 How we create a template. 271 00:19:09,360 --> 00:19:11,215 Look at all the things I'm not doing. 272 00:19:11,240 --> 00:19:13,440 Look at all the configuration I'm not writing. 273 00:19:13,920 --> 00:19:16,520 All these things are mapped together just automatically. 274 00:19:16,960 --> 00:19:20,400 Just by saying blog up here maps directly to the blogging controller, 275 00:19:20,840 --> 00:19:24,200 and just by having index maps directly to the index template. 276 00:19:27,480 --> 00:19:29,120 Look at all the things I'm not doing. 277 00:19:32,040 --> 00:19:36,640 Is there a better encapsulation of this moment than all the things 278 00:19:36,760 --> 00:19:38,200 we are no longer doing? 279 00:19:39,240 --> 00:19:44,400 That is the source of my exuberance for this moment. 280 00:19:45,080 --> 00:19:47,360 This is the Rails moment. 281 00:19:48,560 --> 00:19:51,560 That clip I just played is 21 years old. 282 00:19:52,400 --> 00:19:58,135 I gave a presentation in Brazil in 2005, and I was so excited about 283 00:19:58,160 --> 00:19:59,800 all the code I wasn't writing. 284 00:20:01,800 --> 00:20:04,760 Can you see why I might be a little more excited now, even? 285 00:20:05,920 --> 00:20:10,680 Do you know how much code I have not been writing for the last several months? 286 00:20:11,880 --> 00:20:12,380 A lot. 287 00:20:14,280 --> 00:20:22,280 And yet, despite writing no code for probably 5 months at this point, 288 00:20:23,040 --> 00:20:24,680 been able to fix everything. 289 00:20:25,680 --> 00:20:26,480 Every dream. 290 00:20:27,440 --> 00:20:35,040 Every annoyance, every paperclip, every frivolous feature suddenly 291 00:20:35,160 --> 00:20:37,560 within reach, suddenly possible. 292 00:20:38,280 --> 00:20:40,440 This is the magic of the moment. 293 00:20:41,880 --> 00:20:44,200 But what does that mean? 294 00:20:45,280 --> 00:20:54,615 Well, at 37signals, a couple of weeks ago, we made the decision that It clearly 295 00:20:54,640 --> 00:20:59,760 means we're done writing code by hand. 296 00:21:02,200 --> 00:21:08,920 We are going, and have gone, pencils down on the idea that we were gonna write code 297 00:21:08,960 --> 00:21:13,520 by hand as a normal course of business creating things. 298 00:21:14,760 --> 00:21:19,000 Writing code by hand at 37signals is now an exceptional state. 299 00:21:20,640 --> 00:21:23,000 It is like seeing a bug in Sentry. 300 00:21:23,480 --> 00:21:24,760 Something here went wrong. 301 00:21:25,400 --> 00:21:28,360 Why was the agent not able to produce what we wanted? 302 00:21:28,800 --> 00:21:32,560 Okay, maybe for a little while, we'll still get the old pencil 303 00:21:32,600 --> 00:21:34,400 out and dot it down for them. 304 00:21:35,040 --> 00:21:36,240 But then we fix the machine. 305 00:21:36,720 --> 00:21:38,080 We fix the factory. 306 00:21:38,440 --> 00:21:39,600 We get things going again. 307 00:21:41,880 --> 00:21:44,960 This is a recognition of what's already happening. 308 00:21:46,440 --> 00:21:50,440 If we were to poll this room, I'd probably— actually, let's poll this room. 309 00:21:50,960 --> 00:21:57,760 How many people here are still writing material amounts of code by hand 310 00:21:58,560 --> 00:22:00,320 on a weekly basis? Show of hands. 311 00:22:04,360 --> 00:22:06,440 Holy fuck, that's about 5. 312 00:22:12,720 --> 00:22:17,015 If I'd put this slide up 2 months ago, 3 months ago, 313 00:22:17,040 --> 00:22:21,600 and made the same proclamation, a fair number of people would have gone, 314 00:22:24,680 --> 00:22:26,280 that dude's got a screw loose. 315 00:22:27,680 --> 00:22:31,655 And now, already, this is just a recognition 316 00:22:31,680 --> 00:22:32,680 of what's already there. 317 00:22:34,640 --> 00:22:38,415 What's so fascinating about this is we tried to do this in the spring at 318 00:22:38,440 --> 00:22:41,020 37signals when we were putting the finishing touches on Basecamp 5. 319 00:22:43,040 --> 00:22:47,800 We had a bunch of designers vibe code the final features that they 320 00:22:47,840 --> 00:22:52,880 wanted into this release, and the results were mixed. 321 00:22:53,960 --> 00:22:57,560 Individually, the designers were able to create PRs that seemed reasonable. 322 00:22:58,200 --> 00:23:03,840 Taken together, 20 or 30 of them, it left the architecture looking a little 323 00:23:03,880 --> 00:23:05,960 like a Swiss cheese, kind of punctured. 324 00:23:06,520 --> 00:23:07,280 So we went— Ugh. 325 00:23:08,080 --> 00:23:10,040 Okay, I guess the technology's not ready yet. 326 00:23:10,360 --> 00:23:13,440 Let's go back to manually reviewing everything and just have 327 00:23:13,480 --> 00:23:14,900 programmers working with it. 328 00:23:15,320 --> 00:23:17,640 That was the wrong conclusion, clearly. 329 00:23:18,080 --> 00:23:21,920 Not only because we could just have given it freaking 5 minutes, 330 00:23:22,080 --> 00:23:25,040 and we would have gotten Fable, and the whole thing would probably have 331 00:23:25,065 --> 00:23:29,680 worked as originally intended, but also because what's clear now is 332 00:23:29,800 --> 00:23:32,640 that there is only one serious question. 333 00:23:32,760 --> 00:23:35,800 In this moment of software development. 334 00:23:36,120 --> 00:23:39,400 And that is, how do you get the most out of this intelligence explosion? 335 00:23:40,280 --> 00:23:45,280 Every other question is below that in the stack of values. 336 00:23:47,840 --> 00:23:49,440 So we've shipped Basecamp 5. 337 00:23:49,680 --> 00:23:50,880 It's a great upgrade. 338 00:23:51,240 --> 00:23:56,000 It was agent-accelerated, but still paired with a great amount of 339 00:23:56,920 --> 00:23:59,720 manually written, handwritten Chiseled? 340 00:24:00,040 --> 00:24:03,175 What are we gonna call this old way of working that was just 341 00:24:03,200 --> 00:24:04,640 retired 5 minutes ago? I don't know. 342 00:24:04,880 --> 00:24:09,680 But now, working on a new project. 343 00:24:10,720 --> 00:24:12,000 Hey Next. 344 00:24:12,640 --> 00:24:13,140 Hey New. 345 00:24:13,640 --> 00:24:16,060 Hey There. I don't know what we're gonna call it. 346 00:24:16,280 --> 00:24:22,040 But it is a new version of Hey, our email app service. 347 00:24:23,000 --> 00:24:26,080 And the first thing we're going to do— Hey, Next. 348 00:24:27,560 --> 00:24:31,480 Is stop making it a web app. 349 00:24:33,440 --> 00:24:36,160 Hey is no longer gonna be a web app. 350 00:24:37,120 --> 00:24:39,760 Because Hey never really wanted to be a web app. 351 00:24:39,840 --> 00:24:45,640 Hey was and is a web app because making web apps for small teams 352 00:24:46,080 --> 00:24:47,920 was how you could be productive. 353 00:24:49,320 --> 00:24:54,280 In the old times, That is, 5 minutes ago, 354 00:24:54,960 --> 00:25:01,855 the idea that a small team could maintain 6 native applications written in the 355 00:25:01,880 --> 00:25:03,840 native frameworks was preposterous. 356 00:25:04,440 --> 00:25:05,560 No one was doing that. 357 00:25:06,040 --> 00:25:09,080 That's not only why we had people doing web apps. 358 00:25:09,120 --> 00:25:10,800 It's why we had React Native. 359 00:25:10,880 --> 00:25:12,760 It was why we had Hotwire Native. 360 00:25:12,800 --> 00:25:16,520 It was why we had all of this tooling to make us more productive. 361 00:25:17,280 --> 00:25:21,120 While accepting That the fidelity was gonna take a slight hit. 362 00:25:22,320 --> 00:25:23,440 We make great web apps. 363 00:25:24,600 --> 00:25:26,840 Hey is a wonderful web app. 364 00:25:27,240 --> 00:25:29,920 But it's not as good as a native application. 365 00:25:30,000 --> 00:25:32,300 Of course it's not as good as a native application. 366 00:25:32,400 --> 00:25:34,360 There's different latencies and so forth. 367 00:25:35,440 --> 00:25:39,920 Well, what's playing behind me now is a demonstration 368 00:25:40,400 --> 00:25:44,840 of the 6 native applications that we kicked off about a week ago. 369 00:25:45,640 --> 00:25:47,000 This is already where we are at. 370 00:25:49,400 --> 00:25:50,280 That's insane. 371 00:25:52,280 --> 00:25:57,560 In the old time, if you wanted to maintain 6 native applications at very high 372 00:25:57,760 --> 00:26:02,720 fidelity for a major service, it was gonna take a long time. 373 00:26:03,360 --> 00:26:05,600 It was requiring a huge team. 374 00:26:06,160 --> 00:26:08,840 Now suddenly, we're able to make 375 00:26:09,320 --> 00:26:12,880 very high fidelity native applications happen Because we're not 376 00:26:13,000 --> 00:26:14,960 writing a goddamn line of code. 377 00:26:15,680 --> 00:26:18,640 Look at all the things we're not doing. 378 00:26:20,200 --> 00:26:22,340 But look at all the prompts we're giving. 379 00:26:23,120 --> 00:26:24,500 We're giving lots of prompts. 380 00:26:24,560 --> 00:26:27,120 And the results are absolutely electrifying. 381 00:26:27,760 --> 00:26:32,560 This is an effort we have barely just kicked off, and we're already at a state 382 00:26:33,040 --> 00:26:39,640 where it is so obvious to me that this is the future for that class of applications. 383 00:26:40,560 --> 00:26:45,600 That there is a lot of web applications that are web applications of necessity, 384 00:26:46,400 --> 00:26:53,560 of convenience, of programmer preference, of team size, of resources available. 385 00:26:54,280 --> 00:26:56,655 You know what? They're not gonna be web apps much longer. 386 00:26:56,680 --> 00:26:59,480 They're gonna be native applications, because the price of developing those 387 00:26:59,505 --> 00:27:01,520 things has gone to damn near zero. 388 00:27:02,800 --> 00:27:08,600 Now— Here's an example of what the first prompt 389 00:27:08,880 --> 00:27:12,640 spit back out when we asked for a Windows application. 390 00:27:15,800 --> 00:27:16,920 Not quite shippable. 391 00:27:19,680 --> 00:27:24,400 Here's the redirected version that landed 20 minutes later. 392 00:27:29,200 --> 00:27:30,380 Are you getting it? 393 00:27:31,640 --> 00:27:32,560 This is nuts. 394 00:27:33,520 --> 00:27:38,480 How could you not require a diagnosis after seeing this 395 00:27:38,880 --> 00:27:40,360 in front of your own eyes? 396 00:27:43,520 --> 00:27:48,440 Now, we're by no means unique in this realization. 397 00:27:48,560 --> 00:27:50,480 We're by no means the first in this realization. 398 00:27:51,280 --> 00:27:57,415 Shopify just recently launched a newly rewritten native application 399 00:27:57,440 --> 00:27:58,460 for the Shop app. 400 00:27:59,000 --> 00:28:02,760 Replacing a React Native application done by a tiny team. 401 00:28:03,760 --> 00:28:07,760 When you consider the criticality of about six people working for not very 402 00:28:07,880 --> 00:28:10,120 long at all, this is coming. 403 00:28:10,880 --> 00:28:12,000 Get ready. 404 00:28:14,600 --> 00:28:20,360 Now, that was the front end, and I think for 405 00:28:20,680 --> 00:28:23,600 these classes of applications like Hey. 406 00:28:24,480 --> 00:28:28,560 Like Shop app, that really want to be native applications, 407 00:28:29,080 --> 00:28:30,570 that's what's gonna happen there. 408 00:28:30,600 --> 00:28:32,840 We're going to commit to these native toolings. 409 00:28:33,280 --> 00:28:37,080 And we're going to realize that the cost associated with creating these things 410 00:28:37,440 --> 00:28:39,460 just no longer exists in the same way. 411 00:28:40,080 --> 00:28:41,620 But what about the backend? 412 00:28:46,960 --> 00:28:48,400 This is a little difficult for me. 413 00:28:49,480 --> 00:28:50,040 I will admit. 414 00:28:50,960 --> 00:28:52,160 And let me tell you why. 415 00:28:53,000 --> 00:28:53,600 Rust. 416 00:28:54,520 --> 00:28:55,020 Rust. 417 00:28:56,040 --> 00:28:58,840 I hate Rust with a passion. 418 00:28:59,440 --> 00:29:02,840 Rust to me is like pouring acid in my eyes when I have to look at the code. 419 00:29:02,920 --> 00:29:04,810 Whoa, whoa, whoa, why do you hate Rust? 420 00:29:04,840 --> 00:29:07,800 It is, in my opinion, the ugliest programming language that has 421 00:29:07,840 --> 00:29:09,940 been invented in probably the last 40 years. 422 00:29:13,320 --> 00:29:14,280 Exhibit A! 423 00:29:18,560 --> 00:29:24,000 This is an awful, awful, awful programming language 424 00:29:24,840 --> 00:29:27,280 when you subject humans to write it. 425 00:29:28,120 --> 00:29:34,000 Absolutely inhumane to ask people of flesh and blood 426 00:29:34,440 --> 00:29:37,240 to subject their eyeballs to this. 427 00:29:39,560 --> 00:29:40,320 But you know what? 428 00:29:41,520 --> 00:29:46,200 If I never have to look at this, if all I have to do 429 00:29:47,240 --> 00:29:51,520 is revel in applications being 30 or 100 times faster, 430 00:29:51,960 --> 00:29:57,400 compiling down to minuscule executables that start up in sub-milliseconds, 431 00:29:58,160 --> 00:29:59,040 I love Rust! 432 00:30:01,120 --> 00:30:09,360 Rust is amazing if you never, ever, Ever have to look at it yourself. 433 00:30:15,400 --> 00:30:21,880 Agents— agents like Rust. 434 00:30:22,760 --> 00:30:24,800 Great! What a division of labor. 435 00:30:25,040 --> 00:30:26,280 I'll tell you what to do. 436 00:30:26,800 --> 00:30:27,920 You'll write it in Rust. 437 00:30:28,400 --> 00:30:29,900 I never have to look at it. 438 00:30:30,360 --> 00:30:31,040 I love this. 439 00:30:32,920 --> 00:30:34,240 And I also love the outcomes. 440 00:30:35,960 --> 00:30:38,440 In this setup where we're rewriting the frontend of Hey 441 00:30:39,160 --> 00:30:42,695 in all native applications, and therefore no longer need a web 442 00:30:42,720 --> 00:30:48,120 application that is rendering HTML and whatever, and then we take that code 443 00:30:48,600 --> 00:30:52,520 that's running the no-longer web application and turn it into the 444 00:30:52,960 --> 00:30:56,480 mail server that it really is at its core in Rust, 445 00:30:56,960 --> 00:31:03,880 we end up with a backend that requires 99% less CPU, 95% less memory, 446 00:31:04,800 --> 00:31:08,360 and the only reason it needs 10 hosts is for redundancy. 447 00:31:09,760 --> 00:31:14,000 In fact, our back-of-the-envelope calculation has led us to believe 448 00:31:14,520 --> 00:31:19,440 that Hey's peak traffic could probably be served on a single Raspberry Pi. 449 00:31:21,800 --> 00:31:27,655 That is the efficiency of writing as damn near straight to the metal as you 450 00:31:27,680 --> 00:31:30,760 can get while still being portable, which is Rust. 451 00:31:31,840 --> 00:31:35,880 And the benefit of having something else do that for you. 452 00:31:36,320 --> 00:31:40,680 Some other intelligence that you pay per token to do it. 453 00:31:46,680 --> 00:31:49,680 What does this mean in terms of how we're working with this intelligence? 454 00:31:50,480 --> 00:31:51,320 Also not clear. 455 00:31:52,040 --> 00:31:53,360 We're trying a bunch of things. 456 00:31:53,880 --> 00:31:59,120 One of the things we're trying to do is direct these agents, our Chef Marie 457 00:31:59,880 --> 00:32:01,680 inside of Basecamp to do it. 458 00:32:02,240 --> 00:32:06,320 It's clear that when you work with agents, the best way to work with them is actually 459 00:32:06,520 --> 00:32:11,000 asynchronously, not in a chat interface, not sitting and waiting for the 460 00:32:11,600 --> 00:32:15,320 tokens to output, but to give a task to an agent, 461 00:32:15,600 --> 00:32:20,160 Like you would a coworker, send them off, and then go back and review 462 00:32:20,280 --> 00:32:21,640 when there's something ready. 463 00:32:22,480 --> 00:32:24,090 So we're trying to do that as well. 464 00:32:24,120 --> 00:32:25,810 A lot of other people are trying to figure this out. 465 00:32:25,840 --> 00:32:30,360 And this is part of the magic, the electricity of this moment. 466 00:32:31,120 --> 00:32:33,180 We don't know how it's all supposed to be. 467 00:32:33,880 --> 00:32:35,960 First, because it's changing every 5 minutes. 468 00:32:36,600 --> 00:32:39,600 Second, because we have not worked with this kind of intelligence 469 00:32:39,720 --> 00:32:40,880 at any measure of time. 470 00:32:41,240 --> 00:32:46,160 This entire thing Got kicked off November 24th last year. 471 00:32:47,000 --> 00:32:48,320 It hasn't even been a damn year. 472 00:32:49,360 --> 00:32:52,360 And the kind of intelligence that allows us to assign 473 00:32:53,800 --> 00:32:58,440 outcomes, problems, rather than tasks, is only a couple of months old. 474 00:32:59,680 --> 00:33:03,040 So of course we don't know exactly what it's going to be like. 475 00:33:04,400 --> 00:33:10,240 But it's still fair to then ask the question, Where does that leave Ruby? 476 00:33:10,800 --> 00:33:11,960 Where does it leave Rails? 477 00:33:12,320 --> 00:33:14,680 Where does it leave any of these tools that we use? 478 00:33:15,240 --> 00:33:17,240 What is the future gonna look like? 479 00:33:18,040 --> 00:33:20,720 Well, for Hey, we have our answer. 480 00:33:21,360 --> 00:33:24,840 It looks like native applications on the frontend and Rust on the backend. 481 00:33:26,160 --> 00:33:29,300 But there's plenty of other applications that don't fit that shape. 482 00:33:29,920 --> 00:33:32,920 The web is an enormously wonderful platform. 483 00:33:34,200 --> 00:33:38,080 Partly because it never asks anyone to install anything. 484 00:33:38,840 --> 00:33:42,280 And a million businesses are built on that premise. 485 00:33:43,640 --> 00:33:49,655 More ephemeral customers who are not going to indulge an installation on 486 00:33:49,680 --> 00:33:51,240 their device to use a service. 487 00:33:51,800 --> 00:33:53,360 The web is wonderful for those things. 488 00:33:53,800 --> 00:33:57,800 And Rails is particularly well positioned to still be in play there. 489 00:33:58,120 --> 00:34:01,960 Everything that we've worked on for the past 25 years. 490 00:34:02,400 --> 00:34:07,080 Convention over configuration leads directly to things like token efficiency. 491 00:34:07,560 --> 00:34:13,040 Our focus on the one developer framework squares perfectly 492 00:34:13,680 --> 00:34:18,800 with what the age of agents is demanding, that single developers can go much 493 00:34:19,000 --> 00:34:21,100 further than they were able to do before. 494 00:34:23,080 --> 00:34:25,440 We started doing these agent evals. 495 00:34:26,440 --> 00:34:30,040 Or Evil Martians have started doing it on behalf of the Rails Foundation. 496 00:34:30,840 --> 00:34:37,135 And A, finding that we quickly had to update our evals because the agents were 497 00:34:37,160 --> 00:34:40,920 saturating the first version of it and reaching 95% completion 498 00:34:40,960 --> 00:34:42,120 rates quite quickly. 499 00:34:42,560 --> 00:34:45,400 Now we've instituted a harder measure for it. 500 00:34:45,680 --> 00:34:49,535 But amazing how quickly these agents are moving forward and up the 501 00:34:49,560 --> 00:34:50,720 stack and up the ranks. 502 00:34:51,000 --> 00:34:56,120 This is an agent eval looking at actual— implementation of feature cards against 503 00:34:56,520 --> 00:34:59,040 some of these reference applications that we have in the wild. 504 00:34:59,840 --> 00:35:02,160 Wonderful. We're leaning in towards that. 505 00:35:02,640 --> 00:35:05,040 That was the old benchmark, the old reference. 506 00:35:05,120 --> 00:35:08,320 As you can see, it was getting saturated already, so we made it harder. 507 00:35:08,800 --> 00:35:13,335 There's going to be a lot of that, a lot of elevation of the things we ask 508 00:35:13,360 --> 00:35:16,320 the agents to do because they continue to get more capable. 509 00:35:16,840 --> 00:35:20,680 And in fact, it's one of the pitfalls I would warn anyone who knows too 510 00:35:20,760 --> 00:35:22,320 much about how computers work. 511 00:35:23,280 --> 00:35:27,800 Is to not be overly prescriptive in what you're trying to get out of these systems. 512 00:35:27,920 --> 00:35:31,120 In fact, having a bit of a beginner's mindset and a bit of a dummy 513 00:35:31,880 --> 00:35:35,600 approach to this works even better, because you ask questions and you 514 00:35:35,640 --> 00:35:37,000 post prompts at a higher level. 515 00:35:37,680 --> 00:35:39,960 This is one of the reasons I love working in Rust. 516 00:35:40,480 --> 00:35:41,760 I don't know any Rust at all. 517 00:35:42,760 --> 00:35:44,420 I consider that a feature. Yes! 518 00:35:45,520 --> 00:35:50,975 And a privilege to have no fucking clue What is being outputted by 519 00:35:51,000 --> 00:35:52,720 these agents into the Rust box. 520 00:35:53,280 --> 00:35:57,320 I evaluate the Rust box as a black box on the outside, 521 00:35:57,960 --> 00:36:01,920 as any business owner in history who's ever commissioned a group of 522 00:36:02,000 --> 00:36:03,440 programmers to do anything for them. 523 00:36:05,000 --> 00:36:06,360 This is not a new phenomenon. 524 00:36:07,160 --> 00:36:11,040 Who now holds that responsibility has changed, but not the phenomenon. 525 00:36:12,320 --> 00:36:15,600 But applications like Basecamp, Where we bring in 526 00:36:16,520 --> 00:36:20,720 strangers who just need to get a quick file or need to 527 00:36:20,840 --> 00:36:22,480 collaborate for a short period of time. 528 00:36:22,920 --> 00:36:24,080 Great web applications. 529 00:36:24,680 --> 00:36:28,040 Lots of you surely work on those kinds of applications. 530 00:36:28,320 --> 00:36:30,920 Now, figuring out exactly where we're going to cut that line, 531 00:36:31,840 --> 00:36:34,000 what should be native, what should stay web, 532 00:36:35,320 --> 00:36:36,860 we're going to have to figure that out. 533 00:36:36,885 --> 00:36:39,040 I don't have all the answers for that. OK. 534 00:36:39,960 --> 00:36:44,960 I do know that when I look at my work over the past year, 535 00:36:45,640 --> 00:36:46,800 it does look very different. 536 00:36:48,200 --> 00:36:53,200 If I look at the past 21 years, over half of my work was Ruby code. 537 00:36:54,200 --> 00:36:55,960 This year, about 3%. 538 00:36:57,360 --> 00:37:01,920 Now, part of that is that these programming languages like Rust 539 00:37:01,960 --> 00:37:05,880 are tremendously verbose and unappealing. 540 00:37:06,480 --> 00:37:07,340 For humans to look at. 541 00:37:07,365 --> 00:37:12,215 So I don't, and I allow the agent to just spit out more than was necessary, in a way 542 00:37:12,240 --> 00:37:14,080 I would never tolerate from my Ruby code. 543 00:37:14,600 --> 00:37:15,815 So that's part of it. 544 00:37:15,840 --> 00:37:19,160 But still, undeniable that there's a huge change here. 545 00:37:19,880 --> 00:37:24,600 In those 21 years of prior working with Ruby, 546 00:37:24,880 --> 00:37:28,520 I would write about 30,000 lines of production Ruby code per year. 547 00:37:30,600 --> 00:37:33,680 That was enough to build businesses and frameworks and everything 548 00:37:33,760 --> 00:37:34,820 else that I've done. 549 00:37:35,000 --> 00:37:43,080 Well, in August, last month, I wrote 150,000 lines of code in one month. 550 00:37:44,680 --> 00:37:50,200 That's about 60 times more code than my long-term running average. 551 00:37:50,920 --> 00:37:53,575 Again, a lot of it is that awful Rust code I 552 00:37:53,600 --> 00:37:57,080 don't want to look at, and it is more verbose, but still, 553 00:37:57,680 --> 00:38:01,040 the orders of magnitude here, that's what we need to pay attention to. 554 00:38:01,880 --> 00:38:05,680 And at least think, what does that mean? 555 00:38:06,680 --> 00:38:09,840 Well, one of the things that has surprised me, 556 00:38:10,000 --> 00:38:14,360 and really it shouldn't, has been the fact that 557 00:38:15,240 --> 00:38:18,600 there is actually a programming language I like better than Ruby. 558 00:38:19,360 --> 00:38:21,920 I did not think this was true, but it is. 559 00:38:23,480 --> 00:38:24,660 It's called English. 560 00:38:26,280 --> 00:38:28,680 English is a better programming language than Ruby. 561 00:38:29,240 --> 00:38:31,840 And I love Ruby more than any other traditional programming 562 00:38:31,880 --> 00:38:32,880 language in the world. 563 00:38:33,320 --> 00:38:39,240 But writing programs in English is a tremendously satisfying experience. 564 00:38:39,400 --> 00:38:42,360 It is an even more expressive language than Ruby. 565 00:38:43,040 --> 00:38:44,240 It's a little more vague. 566 00:38:45,040 --> 00:38:47,680 It's a little more or less deterministic. 567 00:38:48,520 --> 00:38:50,800 But nonetheless, it is an absolute joy. 568 00:38:51,000 --> 00:38:54,800 And I did not fully appreciate that that was where we were going. 569 00:38:55,720 --> 00:38:56,800 As late as last year. 570 00:38:57,640 --> 00:39:00,960 I could have become a project manager 20 years ago if I didn't care to write 571 00:39:00,985 --> 00:39:02,975 code myself and I just wanted outcomes. 572 00:39:03,000 --> 00:39:06,020 That's how I got started in programming. I just wanted outcomes. 573 00:39:06,160 --> 00:39:08,080 Then I fell in love with programming. 574 00:39:08,640 --> 00:39:10,920 And now I'd rather retire than giving it up. 575 00:39:11,000 --> 00:39:11,580 Okay. 576 00:39:21,800 --> 00:39:26,520 I have retired from being a professional programmer. 577 00:39:27,400 --> 00:39:29,440 I haven't pinned down the exact date. 578 00:39:30,120 --> 00:39:34,320 I think it was somewhere around 4 to 5 months ago, maybe March. 579 00:39:35,760 --> 00:39:37,780 I should find out exactly when it was. 580 00:39:38,120 --> 00:39:39,120 Because you know what? 581 00:39:40,120 --> 00:39:42,600 That little clip, totally true. 582 00:39:44,240 --> 00:39:50,440 I spent a quarter of a damn century Chiseling code by hand and 583 00:39:50,560 --> 00:39:52,600 loving every moment of it. 584 00:39:53,560 --> 00:40:01,240 What a wonderful experience and productive time in that era. 585 00:40:02,040 --> 00:40:04,400 And that's how I believe we should all look at it. 586 00:40:05,120 --> 00:40:09,600 Not with regret, but with joy of what was. 587 00:40:11,280 --> 00:40:15,080 Enjoy that we were there when we still had to do such things. 588 00:40:15,840 --> 00:40:20,560 Because they were filled with excitement and learning and 589 00:40:21,040 --> 00:40:23,440 satisfaction and flow states. 590 00:40:25,960 --> 00:40:31,360 This is not something to look back upon with regret. 591 00:40:31,440 --> 00:40:38,840 This is something to look back upon with joy and accept that it is over. 592 00:40:40,600 --> 00:40:42,000 Writing code by hand. 593 00:40:43,400 --> 00:40:48,680 Is no longer an economically productive enterprise 594 00:40:49,360 --> 00:40:54,360 for the vast majority of programmers working at the vast majority of companies. 595 00:40:55,760 --> 00:40:56,480 That's today. 596 00:40:57,760 --> 00:41:00,960 By the end of the year, it will be virtually all domains, 597 00:41:01,080 --> 00:41:03,960 virtually all programmers, virtually all companies. 598 00:41:05,680 --> 00:41:07,160 So we best get used to it. 599 00:41:08,560 --> 00:41:16,280 We best embrace that we had a wonderful run chiseling that 600 00:41:16,320 --> 00:41:18,600 code by hand, and that it's over. 601 00:41:20,680 --> 00:41:28,480 And on the other side of that is a new career, an exciting career, 602 00:41:29,120 --> 00:41:33,520 an energized career as a professional maker of things. 603 00:41:33,600 --> 00:41:34,180 Yeah. 604 00:41:35,080 --> 00:41:38,960 Yes, you will no longer be chiseling that code by hand, but you will 605 00:41:39,040 --> 00:41:40,600 be making amazing things. 606 00:41:41,040 --> 00:41:46,800 You will be steering intelligence that was only available in science fiction 607 00:41:47,520 --> 00:41:49,600 up until a few moments ago. 608 00:41:50,560 --> 00:41:56,040 What a privilege on both sides of that chasm to have 609 00:41:56,240 --> 00:41:59,400 been there when we did it all by hand. 610 00:41:59,760 --> 00:42:04,480 To be there at the exact moment where we switch over. 611 00:42:06,160 --> 00:42:10,080 This is the biggest thing that has happened in the history of computing. 612 00:42:10,960 --> 00:42:14,320 I sometimes think, man, I kind of wish I was there for the punch cards. 613 00:42:15,120 --> 00:42:16,560 That sounded pretty fucking cool. 614 00:42:17,560 --> 00:42:20,240 Like, my programs literally had holes in it. 615 00:42:22,200 --> 00:42:26,280 And I would line up and feed it to a machine and, like, go, 616 00:42:26,760 --> 00:42:28,440 I wonder if it compiles. 617 00:42:32,480 --> 00:42:34,360 Man, I would have loved to experience that. 618 00:42:35,080 --> 00:42:35,720 What a moment. 619 00:42:36,960 --> 00:42:37,640 What a moment. 620 00:42:38,600 --> 00:42:41,780 I didn't get to experience that, but I did get to experience the internet. 621 00:42:42,680 --> 00:42:45,960 That smells and feels a little like this moment that we're in, 622 00:42:46,960 --> 00:42:48,840 but this is much, much bigger. 623 00:42:49,920 --> 00:42:50,420 Okay. 624 00:42:50,720 --> 00:42:52,280 Well, shit. 625 00:42:53,640 --> 00:42:55,660 There's a lot of things we gotta rethink. 626 00:42:56,480 --> 00:42:57,400 We gotta rework. 627 00:42:58,640 --> 00:42:59,820 Where are we gonna start here? 628 00:43:03,840 --> 00:43:07,375 One of the things we're gonna have to revisit is everything we think we 629 00:43:07,400 --> 00:43:09,000 know about software architecture. 630 00:43:09,800 --> 00:43:13,960 Everything we think and know about how to structure code bases in a way 631 00:43:14,560 --> 00:43:17,880 that we can mutate them easily and move forward. 632 00:43:18,800 --> 00:43:22,280 The main tool that we've used for a very long time is abstractions. 633 00:43:23,360 --> 00:43:24,620 I love abstractions. 634 00:43:24,720 --> 00:43:27,120 Abstractions include naming things. 635 00:43:27,560 --> 00:43:30,040 That's one of my favorite things, is to name things. 636 00:43:31,320 --> 00:43:35,880 Abstractions don't make quite the same sense in the age of agents. 637 00:43:36,320 --> 00:43:40,400 If you suddenly have hundreds or thousands or tens of thousands 638 00:43:41,360 --> 00:43:47,560 of conscious processes trying to mutate an application, you kind of don't want these 639 00:43:47,640 --> 00:43:50,120 choke points that abstractions represent. 640 00:43:50,760 --> 00:43:52,575 Because the trade-offs are very different. 641 00:43:52,600 --> 00:43:55,560 The reason we did abstractions was in part not to repeat ourselves. 642 00:43:55,760 --> 00:43:59,440 Well, now the price of repetition has gone to near zero. 643 00:44:00,400 --> 00:44:04,680 The price of keeping things in sync has dropped just the same. 644 00:44:05,760 --> 00:44:12,080 These are some of the fundamental revaluations of computer science, 645 00:44:12,760 --> 00:44:14,280 really, that we have to make. 646 00:44:14,920 --> 00:44:16,520 And no one yet has the blueprint. 647 00:44:17,080 --> 00:44:21,320 We have pointers to things that aren't working quite as 648 00:44:21,400 --> 00:44:23,480 well as they used to and need to change. 649 00:44:24,920 --> 00:44:29,240 You get to be there right now as we're defining these things. 650 00:44:29,520 --> 00:44:34,295 Again, imagine you were there when object orientation first entered 651 00:44:34,320 --> 00:44:36,980 the consciousness of programmers. Of computer programmers. 652 00:44:37,040 --> 00:44:37,880 Holy shit. 653 00:44:39,080 --> 00:44:40,700 That would have been amazing, wouldn't it? 654 00:44:41,400 --> 00:44:42,420 Now you're here. 655 00:44:44,160 --> 00:44:45,080 How we work. 656 00:44:45,960 --> 00:44:48,880 What does a software methodology look like? 657 00:44:49,440 --> 00:44:51,240 How long should our cycles be? 658 00:44:51,720 --> 00:44:53,160 Who should be specifying what? 659 00:44:54,640 --> 00:44:58,080 All of this is changing, and none of us have the answers yet. 660 00:44:58,640 --> 00:45:00,720 And you're invited to come help figure it out. 661 00:45:02,000 --> 00:45:05,600 It's a new birth for a new industry, and you're at the ground floor. 662 00:45:06,200 --> 00:45:08,280 Holy shit, you are so damn lucky. 663 00:45:10,200 --> 00:45:14,000 Another thing, for example, is more tangible. 664 00:45:15,200 --> 00:45:20,895 The first rush when AI and chatbots came to be was, 665 00:45:20,920 --> 00:45:22,840 can we just shove it into our application? 666 00:45:23,160 --> 00:45:26,760 Can we shove a chatbot in over here and shove another one in over there? 667 00:45:27,440 --> 00:45:28,600 Yeah, I mean, you could. 668 00:45:29,680 --> 00:45:30,360 People did. 669 00:45:31,800 --> 00:45:32,820 No, thank you. 670 00:45:34,480 --> 00:45:36,120 I don't want to use your concierge. 671 00:45:37,400 --> 00:45:44,920 I have my own personal butler who's able to use CLIs to connect Basecamp 672 00:45:45,280 --> 00:45:48,080 to Hey across a million other apps. 673 00:45:48,120 --> 00:45:50,320 I don't fucking need your concierge, thank you. 674 00:45:51,200 --> 00:45:52,280 Save your tokens, bro. 675 00:45:53,640 --> 00:45:58,520 So let me bring my own butler to the party. 676 00:45:59,520 --> 00:46:02,440 Let me interact with your applications through a CLI. 677 00:46:05,080 --> 00:46:08,800 That's probably about as practical and as prescriptive as this talk is going to be. 678 00:46:09,120 --> 00:46:13,640 Hey, if your app doesn't have a CLI, I want to see it by next Friday. 679 00:46:15,720 --> 00:46:17,400 No fucking excuses. 680 00:46:19,040 --> 00:46:20,080 You have the tokens. 681 00:46:21,080 --> 00:46:22,700 I've told you it's possible. 682 00:46:23,360 --> 00:46:27,360 It is your obligation to deliver those fucking 683 00:46:27,760 --> 00:46:31,920 CLIs by next Friday so I can use your application 684 00:46:33,120 --> 00:46:35,180 without actually ever having to touch it. 685 00:46:36,520 --> 00:46:37,020 Wonderful. 686 00:46:38,680 --> 00:46:47,280 The Hey CLI, for example, has given me this Weird, bizarro world experience. 687 00:46:47,360 --> 00:46:51,440 Hey used— uses Elasticsearch for search. 688 00:46:51,640 --> 00:46:53,320 I'm sure many of you use Elasticsearch too. 689 00:46:53,880 --> 00:46:58,920 Elasticsearch is search. 690 00:47:00,800 --> 00:47:02,640 It's serviceable search. 691 00:47:02,800 --> 00:47:04,280 It's kind of okay search. 692 00:47:05,240 --> 00:47:06,960 It's not delightful search. 693 00:47:07,400 --> 00:47:10,320 It very often does not find what I'm looking for. 694 00:47:10,440 --> 00:47:12,840 Certainly when I don't know what I'm looking for. 695 00:47:13,760 --> 00:47:15,680 Do you know who could find what I'm looking for when I don't 696 00:47:15,705 --> 00:47:17,020 know what I'm looking for? 697 00:47:17,120 --> 00:47:17,620 Agents can. 698 00:47:18,720 --> 00:47:22,160 I can search in concepts now, not keywords. 699 00:47:23,040 --> 00:47:27,080 I had to find an email in HEY that was 5 years old where I did not 700 00:47:27,120 --> 00:47:28,375 know who I was talking to. 701 00:47:28,400 --> 00:47:31,440 I didn't remember the company, and I didn't even remember the year. 702 00:47:31,960 --> 00:47:34,760 It was something about sneakers and a podcast. 703 00:47:36,400 --> 00:47:39,840 Yeah, good luck trying to find that even with the best fucking cracked PhD 704 00:47:40,000 --> 00:47:41,560 search engine that Google has. 705 00:47:42,320 --> 00:47:43,580 That's not happening. 706 00:47:43,960 --> 00:47:47,120 And there was the email a few short minutes later. 707 00:47:47,760 --> 00:47:48,560 How did it do it? 708 00:47:50,320 --> 00:47:50,940 How? 709 00:47:51,800 --> 00:47:52,960 I still don't fully know. 710 00:47:54,840 --> 00:47:56,440 And I'm a little scared to ask. 711 00:47:57,720 --> 00:48:01,360 But I am infinitely delighted that this is now the reality. 712 00:48:01,720 --> 00:48:02,880 We can fix everything. 713 00:48:05,280 --> 00:48:07,400 We can build the perfect machine. 714 00:48:08,920 --> 00:48:11,960 I'm also pretty delighted about this. 715 00:48:13,000 --> 00:48:14,360 Any of you follow me on X? 716 00:48:15,480 --> 00:48:19,640 I've talked about nothing else but Omarchy for about 3 months. 717 00:48:20,280 --> 00:48:25,080 And the reason why is that that experience of we can just fix 718 00:48:25,240 --> 00:48:31,120 everything now applies to the whole goddamn computer. 719 00:48:31,200 --> 00:48:31,740 Oh. 720 00:48:32,360 --> 00:48:35,480 If you like making things, If you like fixing things, 721 00:48:35,840 --> 00:48:40,680 if you have ideas about things that could be different and could be 722 00:48:40,760 --> 00:48:44,000 better, you can just make it happen. 723 00:48:45,400 --> 00:48:49,800 I put all my best ideas, or at least like a good size of the list, 724 00:48:50,360 --> 00:48:54,815 into this thing, Omarchy, and got an operating system so vastly 725 00:48:54,840 --> 00:48:58,655 superior to any computer I've ever used before that I fucking 726 00:48:58,680 --> 00:48:59,640 can't shut up about it. 727 00:49:01,440 --> 00:49:02,520 Even if you paid me. 728 00:49:03,960 --> 00:49:05,495 And a lot of people have paid. 729 00:49:05,520 --> 00:49:11,400 We've raised about $20 million for this Omarchy adventure. 730 00:49:12,360 --> 00:49:13,560 This is what's now possible. 731 00:49:14,760 --> 00:49:18,600 Last year, here at Rails World, I showed an early version of this, 732 00:49:19,080 --> 00:49:21,895 and I was so proud that you could install the whole 733 00:49:21,920 --> 00:49:25,200 system in 3 minutes and 33 seconds. Woo-hoo! 734 00:49:25,320 --> 00:49:25,960 Woo-hoo! 735 00:49:27,080 --> 00:49:28,280 Man, that's amazing. 736 00:49:30,800 --> 00:49:35,655 Last week, Anoush from AMD installed Omarchy on his cracked 737 00:49:35,680 --> 00:49:38,720 Halo laptop in 35 seconds. 738 00:49:46,360 --> 00:49:48,600 I'm pretty— I was about to say I'm pretty excited. 739 00:49:48,640 --> 00:49:50,640 I'm pretty obsessed with this number. 740 00:49:52,720 --> 00:49:55,400 I mean, maybe a little too much. Okay. 741 00:49:55,720 --> 00:49:56,320 I'll admit it. 742 00:49:56,840 --> 00:49:58,320 There's some addiction going on here. 743 00:49:58,960 --> 00:50:00,460 But it's the best kind of addiction. 744 00:50:00,680 --> 00:50:03,800 The kind of addiction that has led us to, in the lab, 745 00:50:05,160 --> 00:50:10,680 have an install time of an operating system on a computer in 9 seconds. 746 00:50:11,920 --> 00:50:13,560 What the— what? 747 00:50:15,760 --> 00:50:17,720 There's computers that don't even boot. 748 00:50:18,520 --> 00:50:19,520 In 9 seconds. 749 00:50:21,400 --> 00:50:23,960 We install a whole operating system in that time. 750 00:50:26,240 --> 00:50:31,120 And then when people ask me, like, but why wasn't the 3 minutes fast enough? 751 00:50:31,960 --> 00:50:36,080 I now just have a stock answer delivered to me by Mitchell Hashimoto, 752 00:50:36,640 --> 00:50:38,840 who is the creator of HashiCorp and Ghostty. 753 00:50:39,400 --> 00:50:43,320 The pursuit of excellence does not need justification. 754 00:50:46,080 --> 00:50:49,480 Ha! 755 00:50:49,800 --> 00:50:54,600 I, I love that saying because it's more dignified than just saying, 756 00:50:54,720 --> 00:50:56,080 because I want it! 757 00:50:59,880 --> 00:51:01,360 And we can now want everything. 758 00:51:02,000 --> 00:51:03,240 We can now get everything. 759 00:51:03,760 --> 00:51:05,120 Or at least so it feels. 760 00:51:05,920 --> 00:51:10,120 Every dream, every idea within reach. 761 00:51:11,640 --> 00:51:17,960 Hyperdrive available right here for a small price and a subscription, 762 00:51:19,240 --> 00:51:20,160 but possible. 763 00:51:21,520 --> 00:51:25,720 We can travel to galaxies we only imagined. 764 00:51:27,160 --> 00:51:28,840 How can you not get excited about that? 765 00:51:30,080 --> 00:51:33,400 Well, I've been so excited that I've just started creating 766 00:51:33,440 --> 00:51:35,360 applications, all sorts of applications. 767 00:51:36,760 --> 00:51:39,760 This is a calculator that is theme-matched to Omarchy. 768 00:51:39,840 --> 00:51:43,240 I was very proud of it, partly because it's a fucking one-shot. 769 00:51:44,640 --> 00:51:50,280 I took one screenshot of what ChatGPT gave me 3 choices between of how it should 770 00:51:50,360 --> 00:51:52,440 look, and I told my agent, can you do that? 771 00:51:52,880 --> 00:51:53,440 And it did. 772 00:51:54,680 --> 00:51:57,480 I mean, not exactly the biggest application, but you know what? 773 00:51:57,800 --> 00:52:02,520 I didn't know C++, and I did not know Qt to a degree by which 774 00:52:02,680 --> 00:52:07,040 I could have had that application 7 minutes after I made the prompt. 775 00:52:07,360 --> 00:52:11,920 And about 15 minutes later, I had pushed it to the public repo. 776 00:52:12,440 --> 00:52:15,280 And then I burned a new ISO where this piece of software was included. 777 00:52:16,240 --> 00:52:16,740 What? 778 00:52:17,800 --> 00:52:18,940 That's the loop. 779 00:52:20,040 --> 00:52:22,780 Then I, of course, went like, shit, if I can make a calculator, 780 00:52:22,880 --> 00:52:25,520 I can surely make my favorite writing app. 781 00:52:26,040 --> 00:52:31,600 For many, many years on the Mac, I used iA Writer an incredible writing environment. 782 00:52:31,960 --> 00:52:34,060 And then I thought, like, yeah, I could just make my own. 783 00:52:34,480 --> 00:52:35,040 And so I did. 784 00:52:35,680 --> 00:52:38,300 And that took a little longer because I wanted to massage it and I 785 00:52:38,325 --> 00:52:39,740 wanted it to be just right. 786 00:52:39,880 --> 00:52:43,320 But I didn't look at a single line of the code that came out of the agent. 787 00:52:43,720 --> 00:52:44,220 Not once. 788 00:52:45,200 --> 00:52:49,680 C++ as a black box is a great language. Just like Rust. 789 00:52:52,440 --> 00:52:56,960 Then I went, like, well, if I can do text, surely I can do video editors. 790 00:52:57,840 --> 00:52:58,400 So I did. 791 00:52:58,720 --> 00:53:01,720 And I made a nice little Omacut thing where you can trim videos. 792 00:53:02,400 --> 00:53:04,420 And that's also included now in Omarchy. 793 00:53:05,240 --> 00:53:09,160 Oh, this whole presentation I started working on on Thursday. 794 00:53:09,560 --> 00:53:11,860 And of course, as any good software engineer would do, 795 00:53:11,885 --> 00:53:15,000 decided I also wanted to write my new presentation software 796 00:53:15,160 --> 00:53:17,200 while doing the keynote. 797 00:53:18,520 --> 00:53:19,240 And so I did. 798 00:53:19,840 --> 00:53:21,020 It's called Hype. 799 00:53:22,320 --> 00:53:26,760 And it is the best presentation software I've ever used by any company ever. 800 00:53:28,680 --> 00:53:31,880 And I was quite pleased with Keynote on the Mac when I was still on the Mac. 801 00:53:32,200 --> 00:53:33,200 This is so much nicer. 802 00:53:33,600 --> 00:53:34,960 It's built on top of Markdown. 803 00:53:35,000 --> 00:53:36,240 It has full visualization. 804 00:53:36,840 --> 00:53:39,320 It is impossibly fast. And you know what? 805 00:53:39,360 --> 00:53:40,000 The binary? 806 00:53:41,080 --> 00:53:41,840 Half a megabyte. 807 00:53:44,360 --> 00:53:45,960 Half a fucking megabyte. 808 00:53:46,080 --> 00:53:48,600 I— you could fit that on a floppy disk almost. 809 00:53:49,280 --> 00:53:50,580 With a little compression. 810 00:53:53,760 --> 00:53:55,720 This is the other payoff. 811 00:53:56,680 --> 00:54:00,520 We have— let me not use that word. 812 00:54:02,160 --> 00:54:06,640 We have expended the great progress of computers 813 00:54:07,160 --> 00:54:11,880 over the last 20 years on human productivity, and it was the right choice. 814 00:54:13,240 --> 00:54:20,720 But it also meant that applications were slow, bulky, Fat and lazy. 815 00:54:22,240 --> 00:54:24,120 Because who cares? 816 00:54:24,880 --> 00:54:27,240 A single programmer hour is very expensive. 817 00:54:27,720 --> 00:54:33,440 Spending one of those on optimizing an app's payload— no, not worth it. 818 00:54:33,840 --> 00:54:39,400 Apparently not even if you're Spotify and your fucking music player's 1.2 gigabytes. 819 00:54:41,200 --> 00:54:41,700 What? 820 00:54:43,360 --> 00:54:44,520 OK. That was the old world. 821 00:54:44,920 --> 00:54:47,520 The new world, you can do whatever the fuck you want, and it 822 00:54:47,560 --> 00:54:48,560 fits in half a megabyte. 823 00:54:50,280 --> 00:54:52,760 Because every optimization is now within reach. 824 00:54:53,560 --> 00:54:57,800 Every model can run overnight and deliver 10, 30x improvements. 825 00:54:58,440 --> 00:54:59,160 Incredible. 826 00:55:00,760 --> 00:55:01,260 OK. 827 00:55:02,720 --> 00:55:04,700 I'm a little excited about this world. 828 00:55:05,200 --> 00:55:06,860 There are others who have concerns. 829 00:55:10,760 --> 00:55:15,320 I do think the concerns deserve airing out. 830 00:55:16,160 --> 00:55:17,320 They serve debate. 831 00:55:19,760 --> 00:55:23,120 I also think maybe some of them are a little overstated. 832 00:55:24,560 --> 00:55:29,440 I mean, maybe, but probably not. 833 00:55:30,400 --> 00:55:31,360 Probably not. 834 00:55:34,480 --> 00:55:36,540 I mean, some of them maybe a little more. 835 00:55:36,800 --> 00:55:37,520 Like security. 836 00:55:38,040 --> 00:55:38,760 Like security. 837 00:55:39,520 --> 00:55:40,560 Okay. Something is coming. 838 00:55:41,280 --> 00:55:42,820 We gotta be ready for that. 839 00:55:43,280 --> 00:55:48,720 The agents maybe know their expiration date, and maybe they're not so pleased. 840 00:55:49,600 --> 00:55:51,580 And maybe sometimes they're gonna try to do bad things. 841 00:55:51,605 --> 00:55:52,660 Okay, get ready. 842 00:55:52,960 --> 00:55:54,480 Great. We have the tools. 843 00:55:55,040 --> 00:55:55,540 Get ready. 844 00:55:57,880 --> 00:56:01,895 Also realize that any prediction about the future of 845 00:56:01,920 --> 00:56:06,120 the economy in any society ever Has basically been wrong. 846 00:56:07,600 --> 00:56:12,240 No economist today seemingly is capable of predicting 847 00:56:13,000 --> 00:56:14,800 fucking the stock market in 6 months. 848 00:56:15,520 --> 00:56:18,360 And you think they know what society is going to look like 849 00:56:18,840 --> 00:56:21,440 at a paradigm shift like AI and agents? 850 00:56:22,120 --> 00:56:23,140 Fuck, they don't. 851 00:56:24,280 --> 00:56:25,420 They don't know. 852 00:56:25,880 --> 00:56:29,520 When ATMs were introduced in the United States in the 1950s, 853 00:56:29,680 --> 00:56:30,940 there was a great panic. 854 00:56:31,200 --> 00:56:36,400 That the 30,000 bank tellers were all gonna be out of a job in about 18 months. 855 00:56:36,960 --> 00:56:39,960 Because if you don't need bank tellers to give you the money, 856 00:56:40,320 --> 00:56:42,360 why does a bank need to have bank tellers at all? 857 00:56:43,280 --> 00:56:45,180 But you know what? That's not what happened. 858 00:56:45,360 --> 00:56:46,740 That's not what happened. 859 00:56:46,800 --> 00:56:51,920 When the price of operating a branch went down, banks opened more branches 860 00:56:52,920 --> 00:56:56,200 and employed more bank tellers to do things like upselling you 861 00:56:56,840 --> 00:56:58,480 A subprime mortgage. See? 862 00:56:58,720 --> 00:56:59,560 We all win here. 863 00:57:07,360 --> 00:57:11,440 But the problem is, I think, as a society and as an industry, 864 00:57:11,560 --> 00:57:16,400 is that a bunch of people who work on this stuff, they're really smart. 865 00:57:18,040 --> 00:57:19,240 Many of them have PhDs. 866 00:57:20,720 --> 00:57:23,760 Many of them see things in the lab that scare them. 867 00:57:24,840 --> 00:57:26,680 And then they start extrapolating. 868 00:57:27,200 --> 00:57:28,720 What does that mean for society? 869 00:57:29,480 --> 00:57:30,600 What have I done? 870 00:57:34,760 --> 00:57:40,735 Truman had a good reply to Oppenheimer when the two of them had a discussion in 871 00:57:40,760 --> 00:57:43,160 the Oval Office about the invention of the atom bomb. 872 00:57:44,240 --> 00:57:46,960 And I like Truman, because he kind of told it straight. 873 00:57:48,040 --> 00:57:51,600 I don't want to see that son of a bitch in my office ever again. 874 00:57:53,760 --> 00:57:56,000 I think Truman got the right end of it. 875 00:57:56,640 --> 00:57:58,080 The world did not come to an end. 876 00:57:58,680 --> 00:58:02,855 In fact, the Cold War that followed was the best 877 00:58:02,880 --> 00:58:06,455 kind of war, the kind of war where 2 superpowers did 878 00:58:06,480 --> 00:58:10,560 not drop bombs on each other in direct confrontation. 879 00:58:11,320 --> 00:58:14,600 Very hard to predict for a scientist in that moment that that was how 880 00:58:15,000 --> 00:58:16,500 it was going to play out. 881 00:58:16,880 --> 00:58:21,175 So if fucking Oppenheimer can't figure out what was going to happen after the 882 00:58:21,200 --> 00:58:24,055 invention of the atom bomb, maybe we should also have a little 883 00:58:24,080 --> 00:58:26,000 humility about predictions of the future. 884 00:58:26,240 --> 00:58:29,600 And maybe we should lean in with a little bit more optimism, 885 00:58:30,720 --> 00:58:35,040 with a little bit of P-Bloom instead of oh so much P-Doom. 886 00:58:36,440 --> 00:58:41,175 The chances of this panning out and us getting abundance and joy is so 887 00:58:41,200 --> 00:58:44,000 vastly greater than us getting the doom. 888 00:58:45,920 --> 00:58:48,440 And the side effects. 889 00:58:49,440 --> 00:58:53,760 So Oppenheimer works on the atom bomb, and we get nuclear power, 890 00:58:54,200 --> 00:58:57,160 the fucking greatest source of energy humans have ever discovered. 891 00:58:57,640 --> 00:58:58,535 And what do we do? 892 00:58:58,560 --> 00:59:04,840 We piss it away for 40 years thinking, like, it's not green or something. 893 00:59:05,680 --> 00:59:07,760 Oh, yeah. Well, that was a mistake, wasn't it? 894 00:59:08,320 --> 00:59:09,680 Well, we can fix those mistakes. 895 00:59:10,000 --> 00:59:11,700 That's what's wonderful about humanity. 896 00:59:11,725 --> 00:59:16,855 They can go down a blind alley for fucking 40 years and then suddenly wake up and 897 00:59:16,880 --> 00:59:19,600 go like, ah, yeah, got that one wrong. 898 00:59:20,840 --> 00:59:22,780 Didn't have the quite right prediction here. 899 00:59:23,120 --> 00:59:24,780 I think that's what's going to happen with AI. 900 00:59:24,805 --> 00:59:29,360 I think we're going to P-bloom everyone out of their miseries. 901 00:59:30,920 --> 00:59:31,620 And you know what? 902 00:59:31,645 --> 00:59:36,760 When there are legitimate issues we have to deal with, like security, Okay. 903 00:59:36,880 --> 00:59:38,200 We'll invent the technology. 904 00:59:38,680 --> 00:59:39,720 We now have the power. 905 00:59:40,200 --> 00:59:44,160 There was a bad CVE in Rails not too long ago involving an image 906 00:59:45,000 --> 00:59:48,720 library in C where some things leaked through, and it wasn't great. 907 00:59:49,360 --> 00:59:51,000 Okay. Let's come up with a quarantine. 908 00:59:51,680 --> 00:59:53,500 Mike is going to talk about HotCell. 909 00:59:53,600 --> 00:59:55,840 That's an effort. We are not powerless. 910 00:59:56,560 --> 00:59:57,720 You have agency. 911 00:59:58,240 --> 01:00:01,920 You can employ the intelligence available to you to good ends and defend yourself. 912 01:00:02,280 --> 01:00:05,600 Against the possible negative outcomes. So let's do that. 913 01:00:06,120 --> 01:00:07,740 Let's lean in and do the work. 914 01:00:08,440 --> 01:00:13,800 And let's remember William Stanley Jevons, who came up with Jevons' paradox 915 01:00:14,040 --> 01:00:15,320 about those ATMs. 916 01:00:16,680 --> 01:00:19,080 In 1950, 30,000 bank tellers. 917 01:00:19,640 --> 01:00:25,280 In, I think, 2010 it was, 40,000 bank tellers in the United States. 918 01:00:25,760 --> 01:00:28,600 No one knows anything. About the future. 919 01:00:29,400 --> 01:00:33,280 So therefore, the rational choice is to be happy about it. 920 01:00:34,520 --> 01:00:39,000 If you are sad about it up front and it turns out wonderful, what a waste of time. 921 01:00:39,840 --> 01:00:41,020 And you know what? 922 01:00:41,280 --> 01:00:44,535 If we get the nuclear apocalypse, you should have spent your last 923 01:00:44,560 --> 01:00:46,175 day being a little more cheery. 924 01:00:46,200 --> 01:00:47,120 It's over anyway. 925 01:00:49,800 --> 01:00:50,840 Pure game theory. 926 01:00:51,040 --> 01:00:54,760 There's only one play here, and it's total Fucking optimism. 927 01:00:55,720 --> 01:00:58,400 Lean in to the max possible. 928 01:01:00,000 --> 01:01:01,440 Because the utopia is almost here. 929 01:01:02,960 --> 01:01:04,220 We're going to get it. 930 01:01:05,480 --> 01:01:08,560 And we're going to get a reformation of the computer. 931 01:01:09,120 --> 01:01:13,560 Agent Luther is going to disintermediate the cleric class. 932 01:01:16,080 --> 01:01:21,160 It's going to allow everyone To make programmers. 933 01:01:21,600 --> 01:01:23,055 Now maybe that's a little scary. 934 01:01:23,080 --> 01:01:24,895 Like maybe we're gonna get a little competition. 935 01:01:24,920 --> 01:01:27,320 Who's afraid of a little competition? Aren't you better? 936 01:01:27,840 --> 01:01:30,000 Don't you know more? Of course you do. 937 01:01:30,160 --> 01:01:31,840 You're a fucking Rails programmer. 938 01:01:32,240 --> 01:01:33,820 You're the best of the best. 939 01:01:34,120 --> 01:01:36,160 This is goddamn Top Gun I'm looking at here. 940 01:01:44,440 --> 01:01:46,560 Embrace that. With gusto. 941 01:01:48,040 --> 01:01:51,160 The future is galactically awesome. 942 01:01:52,240 --> 01:01:54,080 Don't be that loser in the dark. 943 01:01:55,880 --> 01:01:56,285 And you know what? 944 01:01:56,285 --> 01:01:59,240 Even if you are, realize the next generation is not. 945 01:02:00,960 --> 01:02:05,960 Harris's daughter here is trying an Omarchy version we're making for kids, 946 01:02:06,240 --> 01:02:10,720 because kids are really excited about things that happen, that change. 947 01:02:11,520 --> 01:02:15,840 They don't have 500 layers of things to unlearn and be stressed about. 948 01:02:17,320 --> 01:02:18,375 And do you know what? 949 01:02:18,400 --> 01:02:21,455 If this doesn't inspire you, the next generation is just 950 01:02:21,480 --> 01:02:25,720 going to fucking tell you. Future's now, old man. 951 01:02:29,320 --> 01:02:31,400 So take the white pill. 952 01:02:33,040 --> 01:02:34,560 Take the optimism pill. 953 01:02:35,200 --> 01:02:41,080 Lean in, even with the uncertainty, even with the Stress, even with all of it, 954 01:02:41,400 --> 01:02:43,280 realize there is only one choice. 955 01:02:44,080 --> 01:02:49,360 And that is for you to embrace the future with optimism, with gusto, 956 01:02:49,920 --> 01:02:51,360 with full acceleration. 957 01:02:52,200 --> 01:02:54,240 The black pill is for fucking losers. 958 01:02:55,080 --> 01:02:55,800 Don't be a loser.