1 00:00:00,120 --> 00:00:05,080 Speaker 1: If artificial general Intelligence, I mean real AGI truly lands 2 00:00:05,480 --> 00:00:08,679 this year here in twenty twenty six, the world isn't 3 00:00:08,679 --> 00:00:11,000 going to slowly adapt. No, it's not going to have 4 00:00:11,039 --> 00:00:14,119 a series of you know, town hall meetings or some 5 00:00:14,279 --> 00:00:17,000 gradual rollout period where we all learn the new buttons. No, 6 00:00:18,000 --> 00:00:20,760 the analysis we're looking at today says the world is 7 00:00:20,839 --> 00:00:21,960 going to flinch. 8 00:00:22,760 --> 00:00:25,120 Speaker 2: That is such a visceral image. Isn't it a flinch? 9 00:00:25,160 --> 00:00:26,399 It's a physical reaction? 10 00:00:26,719 --> 00:00:28,719 Speaker 1: It totally is it stuck with me? I was reading 11 00:00:28,760 --> 00:00:31,519 through this, this incredible analysis on the arrival of AGI, 12 00:00:31,879 --> 00:00:34,359 and it's focusing specifically on the timelines for this year, 13 00:00:34,920 --> 00:00:38,759 and that one word just hung in the air. Flinch. 14 00:00:39,039 --> 00:00:39,320 Speaker 2: Yeah. 15 00:00:39,359 --> 00:00:41,560 Speaker 1: You know, usually when we talk about big tech changes, 16 00:00:41,799 --> 00:00:44,759 we use these really soft, kind of corporate comfortable words. 17 00:00:44,759 --> 00:00:47,000 Speaker 2: Oh yeah, synergy integration exactly. 18 00:00:47,039 --> 00:00:49,560 Speaker 1: We talk about adoption curves, we talk about gradual integration. 19 00:00:49,799 --> 00:00:51,840 But the source where Unpacking the Day just says, forget 20 00:00:51,880 --> 00:00:55,039 all that. If AGI arrives with the speed, the sheer 21 00:00:55,119 --> 00:00:58,240 velocity that the data suggests it might, and it's pointing 22 00:00:58,320 --> 00:01:01,600 right at twenty twenty six as the tipping point, society 23 00:01:01,640 --> 00:01:02,399 doesn't have time to. 24 00:01:02,359 --> 00:01:03,759 Speaker 2: Adapt It just reacts. 25 00:01:03,840 --> 00:01:06,319 Speaker 1: It reacts physically, It flinches, and. 26 00:01:06,239 --> 00:01:08,599 Speaker 2: We need to be really, really specific about what that 27 00:01:08,640 --> 00:01:11,879 reaction means, because when you flinch, it isn't a calculated 28 00:01:11,920 --> 00:01:14,439 decision at all. You don't sit there and weigh the 29 00:01:14,480 --> 00:01:17,959 pros and cons field in your face. It's involuntary. It's 30 00:01:18,000 --> 00:01:21,040 a shock to the nervous system. And the argument here 31 00:01:21,159 --> 00:01:24,920 isn't that we flinch because machines suddenly, you know, wake 32 00:01:25,000 --> 00:01:27,560 up and start demanding voting rights. The sci fi stup, 33 00:01:27,640 --> 00:01:31,200 that's the Hollywood version. The source is arguing we flinch 34 00:01:31,319 --> 00:01:35,760 because of speed, just pure unadulterated velocity, exactly. 35 00:01:35,879 --> 00:01:38,640 Speaker 1: It's the physics of the situation. It's about momentum, isn't it. 36 00:01:38,640 --> 00:01:42,400 Speaker 2: It is our systems. I'm talking about our laws, our 37 00:01:42,400 --> 00:01:46,319 corporate workflows, even just our day to day human expectations. 38 00:01:46,840 --> 00:01:48,640 They all have a speed limit. They have a kind 39 00:01:48,640 --> 00:01:52,840 of metabolic rate and agi it just doesn't. 40 00:01:52,560 --> 00:01:55,920 Speaker 1: Welcome to thrilling threads. It is Sunday, February first, twenty 41 00:01:55,959 --> 00:01:58,920 twenty six, and we are sitting right here in the 42 00:01:58,920 --> 00:02:01,879 thick of it. The source material we're unpacking today isn't 43 00:02:01,879 --> 00:02:05,359 looking at some distant you know, far off future. It's 44 00:02:05,400 --> 00:02:08,039 looking at the calendar right now. It is arguing that 45 00:02:08,120 --> 00:02:11,360 twenty twenty six is the year the rubber meets the road. 46 00:02:12,000 --> 00:02:14,599 And honestly, you just have to look at the headlines 47 00:02:14,599 --> 00:02:17,120 this week. The smoke is already starting to rise. 48 00:02:17,280 --> 00:02:19,919 Speaker 2: It really is. And our mission today is to really 49 00:02:20,000 --> 00:02:25,479 dissect this argument that the disruption of AI it isn't 50 00:02:25,560 --> 00:02:29,159 really about intelligence and the way we usually think about it. Okay, 51 00:02:29,240 --> 00:02:31,840 the source calls it a timing problem. 52 00:02:31,520 --> 00:02:33,680 Speaker 1: A timing problem. I love that phrasing. I want to 53 00:02:33,759 --> 00:02:36,599 drill into that because usually when I talk to people 54 00:02:36,599 --> 00:02:39,960 about AGI, you know, at dinner parties or just with 55 00:02:40,039 --> 00:02:43,039 friends of the industry, they want to get into philosophical stuff. 56 00:02:43,039 --> 00:02:46,960 Speaker 2: Sure, is it conscious? Is it smart? Does it have feelings? 57 00:02:47,319 --> 00:02:51,240 Can't write a better novel than Hemingway. But you're saying, 58 00:02:51,400 --> 00:02:54,840 or this analysis is saying, that those are completely the 59 00:02:54,879 --> 00:02:56,759 wrong questions to be asking right now. 60 00:02:57,080 --> 00:02:59,599 Speaker 1: They're interesting questions, for sure, they're great for a late 61 00:02:59,680 --> 00:03:02,000 night to eight with a glass of wine, but they 62 00:03:02,120 --> 00:03:05,960 aren't the dangerous questions. The danger the actual shock it 63 00:03:05,960 --> 00:03:07,000 comes from misalignment. 64 00:03:07,080 --> 00:03:07,879 Speaker 2: Misalignment, I can. 65 00:03:07,800 --> 00:03:10,560 Speaker 1: Go out the internet that was a massive technological shift right, 66 00:03:10,919 --> 00:03:12,800 probably the biggest of our lifetimes until. 67 00:03:12,680 --> 00:03:15,639 Speaker 2: Now, but it rolled out over decades. You have the 68 00:03:15,719 --> 00:03:18,439 nineties with dial Remember that awful screeching noise. 69 00:03:19,000 --> 00:03:21,759 Speaker 1: Vaguely, I just remember waiting like five minutes for a 70 00:03:21,800 --> 00:03:23,680 single picture to load exactly. 71 00:03:24,039 --> 00:03:26,560 Speaker 2: Then you get broadband in the two thousands, then the 72 00:03:26,599 --> 00:03:30,039 whole mobile revolution with the iPhone. The point is we 73 00:03:30,120 --> 00:03:34,120 had thirty years to adjust our laws, our manners, our businesses. 74 00:03:34,159 --> 00:03:35,039 Speaker 1: We had a buffer zone. 75 00:03:35,080 --> 00:03:37,960 Speaker 2: We had a buffer zone precisely, we had time to 76 00:03:38,039 --> 00:03:40,879 figure out what email etiquette was before we had to 77 00:03:40,919 --> 00:03:44,000 deal with social media algorithms rewiring our brains. It was 78 00:03:44,000 --> 00:03:45,080 a gradual absorption. 79 00:03:45,479 --> 00:03:46,800 Speaker 1: Okay, I see where this is going. 80 00:03:46,879 --> 00:03:50,439 Speaker 2: But technologies that arrive compressed into a few years, that's 81 00:03:50,479 --> 00:03:54,360 a different beast entirely. That stresses everything until it breaks. 82 00:03:54,840 --> 00:03:58,120 If AGI lands in twenty twenty six, and the metrics 83 00:03:58,159 --> 00:04:02,199 suggests it's landing right now, it lands fundamentally misaligned with 84 00:04:02,280 --> 00:04:03,520 everything we've built. 85 00:04:03,360 --> 00:04:05,479 Speaker 1: Misaligned with our incentives, our legal. 86 00:04:05,240 --> 00:04:09,080 Speaker 2: Structures, our workflows, our expectations, everything. And here's the kicker, 87 00:04:09,120 --> 00:04:10,840 And this is the part that I think keeps the 88 00:04:10,879 --> 00:04:14,280 researchers themselves up at night. It doesn't need to be 89 00:04:14,360 --> 00:04:15,840 perfect to break things. 90 00:04:15,919 --> 00:04:17,879 Speaker 1: That's the good enough principle, right, I've heard that. 91 00:04:18,160 --> 00:04:21,560 Speaker 2: Yes, it just needs to be good enough, cheap enough, 92 00:04:21,600 --> 00:04:24,639 and fast enough to be trusted at scale. Once you 93 00:04:24,720 --> 00:04:28,360 hit that trifecta, the disruption happens whether the machine is 94 00:04:28,439 --> 00:04:29,240 conscious or not. 95 00:04:29,519 --> 00:04:31,560 Speaker 1: So it's not about the soul of the machine. 96 00:04:31,639 --> 00:04:34,319 Speaker 2: The machine doesn't need a soul to take over the economy. 97 00:04:34,360 --> 00:04:36,519 It just needs to be faster than the person trying 98 00:04:36,560 --> 00:04:37,240 to regulate it. 99 00:04:37,319 --> 00:04:39,519 Speaker 1: Okay, let's unpack this because we have a lot to 100 00:04:39,519 --> 00:04:43,240 get through. And frankly, some of this is well thrilling 101 00:04:43,319 --> 00:04:44,959 is in the name of the show, but it's also 102 00:04:45,000 --> 00:04:45,800 a little terrifying. 103 00:04:45,920 --> 00:04:46,439 Speaker 2: It's both. 104 00:04:46,720 --> 00:04:49,480 Speaker 1: Here's our roadmap for this discussion. First, we're going to 105 00:04:49,519 --> 00:04:53,160 look at the evidence. Why twenty twenty six. Why isn't 106 00:04:53,160 --> 00:04:55,639 this just another random guess. Then we're going to talk 107 00:04:55,680 --> 00:04:59,319 about the quiet shift, this move from AI that assists 108 00:04:59,399 --> 00:05:01,319 us to AI that acts for us. 109 00:05:01,600 --> 00:05:04,800 Speaker 2: That's a critical distinction, the move from assistant to agent. 110 00:05:04,879 --> 00:05:06,279 It changes everything huge. 111 00:05:06,399 --> 00:05:08,000 Speaker 1: Then we're gonna look at what breaks first, and here's 112 00:05:08,000 --> 00:05:11,759 a spoiler alert for you. It's not jobs, at least 113 00:05:11,879 --> 00:05:14,199 not in the way you think. And finally, we're going 114 00:05:14,240 --> 00:05:17,560 to talk about the human tool specifically. This this fascinating 115 00:05:17,680 --> 00:05:21,160 and slightly depressing concept the source calls learned irrelevance. 116 00:05:21,480 --> 00:05:24,120 Speaker 2: It's a heavy concept, but it's so essential to understand. 117 00:05:24,160 --> 00:05:26,600 I think it changes how you view your own workday. 118 00:05:26,800 --> 00:05:29,319 Speaker 1: I think so too. So let's jump into part one, 119 00:05:29,680 --> 00:05:34,040 the evidence why twenty twenty six. For years, I mean 120 00:05:34,079 --> 00:05:36,720 literally for years, every time I read a report on AGI, 121 00:05:37,160 --> 00:05:39,279 the date was always ten years. 122 00:05:39,120 --> 00:05:41,240 Speaker 2: Away, always just over the horizon. 123 00:05:41,319 --> 00:05:45,279 Speaker 1: It was a fusion energy, always promising, never quite delivering. 124 00:05:46,000 --> 00:05:49,879 But this source says twenty twenty six different. It's different 125 00:05:49,959 --> 00:05:53,560 because progress is now compounding, not just inching forward. 126 00:05:53,680 --> 00:05:56,360 Speaker 2: That's the key word, isn't it compounding. When you look 127 00:05:56,360 --> 00:05:58,399 at the metrics that are laid out in the source material, 128 00:05:58,439 --> 00:06:00,360 it's not just that the charts are going up. Everyone 129 00:06:00,399 --> 00:06:02,519 knows the charts are going up. It's that the slope 130 00:06:02,519 --> 00:06:03,480 of the line is changing. 131 00:06:03,600 --> 00:06:04,319 Speaker 1: It's getting steeper. 132 00:06:04,399 --> 00:06:07,319 Speaker 2: It's getting steeper. We aren't seeing linear growth anymore. We 133 00:06:07,360 --> 00:06:10,759 are seeing exponential growth in reasoning, in memory scaling, and 134 00:06:10,839 --> 00:06:11,920 in reliable planning. 135 00:06:12,120 --> 00:06:14,959 Speaker 1: Okay, can we pause on reasoning? That word gets thrown 136 00:06:14,959 --> 00:06:18,680 around a lot in marketing materials. What does actual reasoning 137 00:06:18,720 --> 00:06:21,000 look like in the models we are seeing now in 138 00:06:21,079 --> 00:06:24,560 early twenty twenty six, compared to the chatbots of say, 139 00:06:24,720 --> 00:06:26,279 twenty twenty three or twenty twenty. 140 00:06:26,040 --> 00:06:29,319 Speaker 2: Four, it's night and day. It really is. Back in 141 00:06:29,360 --> 00:06:34,720 twenty twenty four, reasoning largely meant very sophisticated pattern matching. Okay, 142 00:06:34,839 --> 00:06:37,120 you'd ask a question and the model would predict the 143 00:06:37,160 --> 00:06:39,959 next likely word based on the vast ocean of text 144 00:06:39,959 --> 00:06:42,319 it had read. It was a kind of mimicry that 145 00:06:42,360 --> 00:06:45,040 looked like thought. I used to call it a parrot 146 00:06:45,040 --> 00:06:46,120 with a library card. 147 00:06:46,519 --> 00:06:49,040 Speaker 1: Right. It could write a nice poem, but it couldn't 148 00:06:49,079 --> 00:06:53,720 plan a complex logistics route without just making stuff up hallucinating. 149 00:06:53,279 --> 00:06:58,360 Speaker 2: Exactly Today, the models we're seeing can perform multi step deduction. 150 00:06:58,800 --> 00:07:01,040 They can look at a problem, break it down into 151 00:07:01,040 --> 00:07:05,279 twelve subcomponents, solve each one of those individually, and then 152 00:07:05,519 --> 00:07:08,360 this is the important part, check their own work for errors, 153 00:07:08,399 --> 00:07:11,959 and then synthesize the answer. That's not mimicry, That isn't 154 00:07:12,000 --> 00:07:15,639 just better writing. That is a fundamental shift in cognitive architecture. 155 00:07:15,800 --> 00:07:18,920 It's the difference between memorizing a math textbook and actually 156 00:07:19,000 --> 00:07:21,079 understanding calculus. 157 00:07:20,399 --> 00:07:22,480 Speaker 1: And tool use. That was one that stood out to 158 00:07:22,519 --> 00:07:24,560 me from the source. It used to be that getting 159 00:07:24,560 --> 00:07:27,720 an AI to use a calculator or browse the web 160 00:07:28,120 --> 00:07:29,920 was this hacky, experimental thing. 161 00:07:30,000 --> 00:07:31,639 Speaker 2: Yeah, you had to kind of coax it, you had 162 00:07:31,680 --> 00:07:32,720 to prompt it perfectly. 163 00:07:33,319 --> 00:07:35,720 Speaker 1: Please mister robot, use the search tool now. 164 00:07:36,560 --> 00:07:40,720 Speaker 2: But now it's default, it's baked in, And that shift 165 00:07:40,879 --> 00:07:45,839 from experimental to reliable is what completely changes the adoption curve. 166 00:07:46,480 --> 00:07:49,839 The source highlights are really staggering statistic on this about 167 00:07:49,959 --> 00:07:52,680 enterprise adoption. I saw that in just a couple of years, 168 00:07:52,680 --> 00:07:55,920 we went from about fifty percent of companies kind of 169 00:07:55,920 --> 00:07:57,439 playing with these tools. 170 00:07:57,120 --> 00:07:59,399 Speaker 1: Right, setting up a sandbox, letting the interns mess around 171 00:07:59,439 --> 00:08:00,800 with it exactly. 172 00:08:00,439 --> 00:08:03,439 Speaker 2: That we went from that to nearly four out of 173 00:08:03,560 --> 00:08:07,480 five companies deply integrating them into core business processes. 174 00:08:07,519 --> 00:08:10,040 Speaker 1: Four out of five. That is eighty percent of the 175 00:08:10,079 --> 00:08:12,279 market in two years. 176 00:08:12,480 --> 00:08:15,480 Speaker 2: And we are not talking about employees using chet GPT 177 00:08:15,639 --> 00:08:17,560 to write a more polite email to their boss. No, 178 00:08:17,759 --> 00:08:20,399 this is real work. This is writing production code that 179 00:08:20,439 --> 00:08:24,000 ships to customers. This is analyzing thousands of pages of 180 00:08:24,079 --> 00:08:27,279 legal contracts in minutes. This is supporting R and D teams. 181 00:08:27,439 --> 00:08:30,279 And the source makes this point really clearly, that kind 182 00:08:30,319 --> 00:08:33,120 of acceleration doesn't happen because a tool is fun. 183 00:08:33,440 --> 00:08:35,000 Speaker 1: No capital is ruthless. 184 00:08:35,399 --> 00:08:39,519 Speaker 2: It is Companies don't overhaul their entire workflow for a toy. 185 00:08:40,000 --> 00:08:42,639 They do it for a tool that directly impacts the 186 00:08:42,679 --> 00:08:43,480 bottom line. 187 00:08:43,600 --> 00:08:45,519 Speaker 1: And speaking of bottom lines, that brings it to the 188 00:08:45,519 --> 00:08:47,000 money right the compute signal. 189 00:08:47,240 --> 00:08:50,799 Speaker 2: This is the ultimate financial reality check. If you really 190 00:08:50,840 --> 00:08:53,399 want to know what the future holds, don't look at 191 00:08:53,399 --> 00:08:56,000 the press releases, look at the bank accounts, look at 192 00:08:56,000 --> 00:08:57,039 the CAPPAC spending. 193 00:08:57,159 --> 00:08:58,600 Speaker 1: Where are they putting their billions? 194 00:08:58,759 --> 00:09:04,440 Speaker 2: Exactly? And the cost of training these single frontier models 195 00:09:04,600 --> 00:09:08,960 has just ballooned. We're talking about single training runs costing 196 00:09:09,080 --> 00:09:10,519 hundreds of millions of dollars. 197 00:09:10,679 --> 00:09:13,080 Speaker 1: Just to put that in perspective for everyone. That used 198 00:09:13,120 --> 00:09:16,080 to be the annual budget for entire national AI. 199 00:09:16,039 --> 00:09:18,480 Speaker 2: Programs, yes, for a whole country. 200 00:09:18,600 --> 00:09:21,320 Speaker 1: Now it's the receipt for one experiment exactly. 201 00:09:21,480 --> 00:09:24,639 Speaker 2: And the logic here is brutal, but it's crystal clear. 202 00:09:25,440 --> 00:09:27,879 Nobody spends hundreds of millions of dollars on a single 203 00:09:27,919 --> 00:09:32,039 training run unless they strongly believe that capability is about 204 00:09:32,039 --> 00:09:32,639 to compound. 205 00:09:32,679 --> 00:09:34,720 Speaker 1: Again, You're not doing that for a small bump. 206 00:09:34,960 --> 00:09:37,480 Speaker 2: You don't make that bet for a five percent improvement. 207 00:09:37,600 --> 00:09:40,120 You don't build a billion dollar data center just to 208 00:09:40,159 --> 00:09:43,399 get a chatbot that makes slightly better haikus. You make 209 00:09:43,480 --> 00:09:45,879 that bet because you're expecting a paradigm shift. 210 00:09:46,159 --> 00:09:48,799 Speaker 1: It really feels like the ground is shifting under our feet. 211 00:09:49,399 --> 00:09:53,360 The source also mentions that even the researchers, the people 212 00:09:53,399 --> 00:09:55,559 inside the labs building this stuff, are the ones who 213 00:09:55,600 --> 00:10:00,879 know the code best, they're admitting that the timelines feel unstable. 214 00:10:01,039 --> 00:10:04,919 Speaker 2: That is a very unsettling word for a scientist to use. Unstable. 215 00:10:05,600 --> 00:10:07,679 Scientists like predictability they do. 216 00:10:08,080 --> 00:10:10,000 Speaker 1: It gives me this image of walking on a floor 217 00:10:10,200 --> 00:10:12,960 that's suddenly tilting. You thought you knew the angle, you 218 00:10:12,960 --> 00:10:15,279 thought you had your balance, and then suddenly the slope 219 00:10:15,320 --> 00:10:17,600 gets way steeper without anyone warning you. 220 00:10:17,759 --> 00:10:19,879 Speaker 2: And that instability is coming from the fact that the 221 00:10:19,919 --> 00:10:22,840 benchmarks keep getting rewritten. We set a test for the AI, 222 00:10:23,240 --> 00:10:25,840 you know, say, passing the bar exam or solving the 223 00:10:25,879 --> 00:10:28,480 math data set, thinking it'll take five years. 224 00:10:28,200 --> 00:10:30,679 Speaker 1: To be it takes five months or five. 225 00:10:30,519 --> 00:10:34,039 Speaker 2: Weeks, so we write a harder test that takes five days. 226 00:10:34,639 --> 00:10:37,480 The predictability of the science is breaking down because the 227 00:10:37,519 --> 00:10:40,759 subject of the study is evolving faster than the study itself. 228 00:10:41,039 --> 00:10:44,360 Speaker 1: Wow. Okay, So if twenty twenty six is the year, 229 00:10:45,039 --> 00:10:47,960 and the money, the tech, the adoption rates all suggest 230 00:10:48,000 --> 00:10:50,759 it is, how does it actually hit us? Because the 231 00:10:50,799 --> 00:10:53,039 source argues it won't be some big press conference. We 232 00:10:53,080 --> 00:10:56,039 aren't going to get a Hello Imagi tweet. 233 00:10:56,200 --> 00:10:59,559 Speaker 2: No, And that brings us to the quiet shift we 234 00:10:59,559 --> 00:11:02,200 need to discus. History tells us that the biggest shifts 235 00:11:02,240 --> 00:11:05,399 always arrive quietly and then all at once. Right, and 236 00:11:05,440 --> 00:11:08,720 the source identifies a very specific technical shift that changes 237 00:11:08,759 --> 00:11:13,840 the entire risk profile. It's this move from assisting to acting. 238 00:11:14,039 --> 00:11:16,919 Speaker 1: This is part two the mechanism of shock, and I 239 00:11:16,960 --> 00:11:19,200 think this is the single most important technical takeaway for 240 00:11:19,240 --> 00:11:21,399 our listeners. Old AI, and by old I mean like 241 00:11:21,679 --> 00:11:24,039 two years ago, which is just crazy to say. 242 00:11:23,919 --> 00:11:25,279 Speaker 2: It's insane and eternity. 243 00:11:25,320 --> 00:11:27,720 Speaker 1: In AI time, old AI was passive. 244 00:11:27,480 --> 00:11:29,440 Speaker 2: Right, It waited you typed in a prompt to give 245 00:11:29,480 --> 00:11:31,320 you an answer, and then it just sat there frozen 246 00:11:31,399 --> 00:11:33,559 until you poked it again. It was a completely reactive loop. 247 00:11:33,840 --> 00:11:37,120 Speaker 1: But the new systems, the ones rolling out right now 248 00:11:37,159 --> 00:11:38,480 in twenty twenty six. 249 00:11:38,600 --> 00:11:41,960 Speaker 2: They're active agents. They decide. They don't just answer a question. 250 00:11:42,080 --> 00:11:45,639 They break goals into steps. They choose which tools to use. 251 00:11:46,279 --> 00:11:48,639 If they try a method and it fails, they retry 252 00:11:48,759 --> 00:11:51,720 with a different one. They keep context over long periods. 253 00:11:51,759 --> 00:11:53,440 They don't just assist, they act. 254 00:11:53,799 --> 00:11:56,519 Speaker 1: Let's give a really concrete example of this, because I 255 00:11:56,600 --> 00:11:58,440 want people to feel the difference in their bones. 256 00:11:58,480 --> 00:12:02,600 Speaker 2: Okay, perfect. Let's take booking a corporate trip. The old AI, 257 00:12:03,600 --> 00:12:07,080 you ask find me flights to Tokyo. It gives you 258 00:12:07,120 --> 00:12:07,879 a list of flights. 259 00:12:08,120 --> 00:12:09,840 Speaker 1: That's it, and then the work starts for you. 260 00:12:10,039 --> 00:12:12,679 Speaker 2: Yes, you have to read the list, pick one, go 261 00:12:12,759 --> 00:12:15,120 to the airline's website, type in your credit card, then 262 00:12:15,159 --> 00:12:17,440 go to the hotel booking site. Do that, then open 263 00:12:17,440 --> 00:12:19,879 your calendar and put it all in. The AI was 264 00:12:19,960 --> 00:12:23,080 basically a fancy search engine, right, it was a librarian. 265 00:12:23,360 --> 00:12:25,080 Speaker 1: It handed you the book, but you still had to 266 00:12:25,080 --> 00:12:26,159 read it and do all the work. 267 00:12:26,240 --> 00:12:30,159 Speaker 2: Exactly Now, the agentic AI of twenty twenty six. You 268 00:12:30,200 --> 00:12:32,720 say one thing, get me to Tokyo for the conference 269 00:12:32,759 --> 00:12:35,879 next Tuesday, And what happens The agent checks your calendar 270 00:12:35,919 --> 00:12:39,360 for the dates, It knows your airline loyalty numbers, It 271 00:12:39,480 --> 00:12:42,279 navigates to the airline site on its own, books the 272 00:12:42,279 --> 00:12:45,639 ticket using your corporate card on file, finds a hotel 273 00:12:45,679 --> 00:12:48,840 that matches your stored preference for having a gym, books that. 274 00:12:49,320 --> 00:12:51,639 Then it emails the team in Tokyo to let them 275 00:12:51,639 --> 00:12:54,440 know when you land, and puts the final itinerary on 276 00:12:54,519 --> 00:12:55,360 everyone's calendar. 277 00:12:55,399 --> 00:12:57,480 Speaker 1: And you did nothing but give the initial command. 278 00:12:57,840 --> 00:13:00,000 Speaker 2: Nothing. That sounds incredibly useful. 279 00:13:00,360 --> 00:13:03,559 Speaker 1: It sounds amazing, but also slightly ominous. 280 00:13:03,600 --> 00:13:07,200 Speaker 2: It is incredibly useful, but it creates what the researchers 281 00:13:07,240 --> 00:13:11,120 in this space call speed asymmetry. And this right here 282 00:13:11,360 --> 00:13:13,039 is the core mechanism of the flinch. 283 00:13:13,600 --> 00:13:16,200 Speaker 1: Okay, so define speed asymmetry for us. 284 00:13:16,360 --> 00:13:20,759 Speaker 2: It's simple math. Really, machines iterate in seconds, institutions and 285 00:13:20,840 --> 00:13:23,679 humans iterate in months or years. When you have an 286 00:13:23,720 --> 00:13:26,720 agent that can act, iterate, retry, and deploy a new 287 00:13:26,720 --> 00:13:29,279 strategy in the span of a few seconds, and you 288 00:13:29,360 --> 00:13:32,039 pair that with a human oversight system that requires a 289 00:13:32,080 --> 00:13:36,039 weekly meeting to approve a single decision, you have a fundamental, 290 00:13:36,240 --> 00:13:37,399 dangerous disconnect. 291 00:13:37,679 --> 00:13:40,120 Speaker 1: The source used this great analogy for that. It's like 292 00:13:40,159 --> 00:13:42,200 trying to fact check a fire hose with a clipboard. 293 00:13:42,480 --> 00:13:44,919 Speaker 2: It's a perfect visual, isn't it. You're just standing there 294 00:13:45,000 --> 00:13:47,399 with your little pen and your piece of paper try 295 00:13:47,399 --> 00:13:51,000 and tick boxes, and meanwhile you are being blasted in 296 00:13:51,039 --> 00:13:54,480 the face with gigabytes of executed decisions per second. 297 00:13:54,639 --> 00:13:56,600 Speaker 1: And we are already seeing this play out, aren't we. 298 00:13:56,960 --> 00:13:58,559 The source mantions code shipping. 299 00:13:58,840 --> 00:14:02,519 Speaker 2: Absolutely, we have AI generated code being pushed to production 300 00:14:02,600 --> 00:14:06,679 repositories faster than human teams can possibly review it. Imagine 301 00:14:06,679 --> 00:14:10,120 a software team. The human engineer writes what one hundred 302 00:14:10,200 --> 00:14:12,559 lines of good code a day if they're lucky. Yeah, 303 00:14:12,799 --> 00:14:16,399 the AI agent writes ten thousand lines. Who is checking 304 00:14:16,480 --> 00:14:17,639 the ten thousand lines? 305 00:14:17,799 --> 00:14:18,320 Speaker 1: Nobody? 306 00:14:18,360 --> 00:14:21,679 Speaker 2: Nobody can. They're running automated tests. Sure, but the deep 307 00:14:21,799 --> 00:14:25,320 conceptual understanding, the intuition of a senior developer looking at 308 00:14:25,360 --> 00:14:28,240 it and saying this feels wrong, that's gone. 309 00:14:28,320 --> 00:14:31,799 Speaker 1: And we're seeing automated recommendations in business for pricing, for 310 00:14:31,919 --> 00:14:36,120 marketing being accepted almost blindly, simply because slowing down to 311 00:14:36,200 --> 00:14:37,600 check them feels too expensive. 312 00:14:37,759 --> 00:14:40,000 Speaker 2: That is the good enough danger we talked about, but 313 00:14:40,080 --> 00:14:44,120 now supercharged with agency AGI doesn't need to be flawless 314 00:14:44,159 --> 00:14:47,759 to be destabilizing. It just needs tempo. If the machine 315 00:14:47,799 --> 00:14:50,399 is right ninety percent of the time and it moves 316 00:14:50,440 --> 00:14:53,279 one hundred times faster than you, you eventually stop checking 317 00:14:53,279 --> 00:14:55,759 the ten percent. You just can't keep up. You let 318 00:14:55,799 --> 00:14:58,840 it run so small errors, small biases, they start to 319 00:14:58,840 --> 00:15:00,679 scale up, but you don't know it them until it's 320 00:15:00,679 --> 00:15:03,039 far too late because you've lost the ability. 321 00:15:02,679 --> 00:15:05,720 Speaker 1: To audit the stream. Okay, so we have this high 322 00:15:05,759 --> 00:15:10,120 speed agenic AI and we have these slow legacy human systems. 323 00:15:10,799 --> 00:15:13,240 This leads us right into part three. What breaks first? 324 00:15:14,000 --> 00:15:16,360 Because the common narrative, the one I hear at every 325 00:15:16,360 --> 00:15:19,000 single dinner party, is oh, the robots are coming for 326 00:15:19,039 --> 00:15:21,759 our jobs math unemployment. That's going to be the first sign. 327 00:15:22,039 --> 00:15:24,159 Speaker 2: And that's a valid long term concern, of course. 328 00:15:24,240 --> 00:15:24,519 Speaker 1: Huh. 329 00:15:24,519 --> 00:15:26,919 Speaker 2: But the source argues very convincingly that it is not 330 00:15:27,000 --> 00:15:29,759 the first thing that breaks. The first casualty is coordination. 331 00:15:29,960 --> 00:15:31,519 Speaker 1: Coordination. Explain that. 332 00:15:31,639 --> 00:15:34,519 Speaker 2: Think about how a company is built, how any organization 333 00:15:34,600 --> 00:15:37,480 is structured. It's a hierarchy built on the assumption of 334 00:15:37,559 --> 00:15:41,559 human speed. We have meetings to synchronize, we have approval chains, 335 00:15:41,840 --> 00:15:45,919 we have escalation paths. That entire structure relies on the 336 00:15:45,919 --> 00:15:48,879 assumption that every part of the system moves at roughly 337 00:15:48,919 --> 00:15:49,639 the same pace. 338 00:15:50,080 --> 00:15:53,759 Speaker 1: Right, I email you, You email me back a day later. 339 00:15:53,840 --> 00:15:55,320 We have a meeting next Tuesday to. 340 00:15:55,279 --> 00:15:58,799 Speaker 2: Decide exactly The whole thing chugs along at a human pace. 341 00:15:59,320 --> 00:16:03,000 But now you insert a component, an AI agent that 342 00:16:03,120 --> 00:16:06,240 moves one hundred, maybe a thousand times faster than the 343 00:16:06,320 --> 00:16:08,879 rest of the org chart, the whole structure just starts 344 00:16:08,919 --> 00:16:09,480 to wobble. 345 00:16:09,600 --> 00:16:10,559 Speaker 1: Can't handle the strain. 346 00:16:10,759 --> 00:16:13,639 Speaker 2: It's like putting a Ferrari engine in a bicycle. The 347 00:16:13,639 --> 00:16:18,519 frame can't handle the torque. The management layer completely loses visibility. 348 00:16:18,679 --> 00:16:22,320 How so well, Decisions that used to take days of 349 00:16:22,399 --> 00:16:27,200 meetings and reports now happen in minutes autonymously. Outputs that 350 00:16:27,320 --> 00:16:29,440 used to trickle out over a week arrive all at 351 00:16:29,440 --> 00:16:32,399 once in a giant data dump. The manager isn't firing 352 00:16:32,399 --> 00:16:35,360 people yet. The manager is just sitting there, completely overwhelmed, 353 00:16:35,360 --> 00:16:38,000 realizing they no longer understand how the work is actually 354 00:16:38,039 --> 00:16:38,679 getting done. 355 00:16:39,039 --> 00:16:41,080 Speaker 1: The source brought up a specific example here that I 356 00:16:41,080 --> 00:16:44,039 found really grounding, a tool called overseer os. 357 00:16:44,320 --> 00:16:47,159 Speaker 2: Yes, this is a prime example of that assisting to 358 00:16:47,240 --> 00:16:48,840 acting shift in the real world. 359 00:16:49,000 --> 00:16:50,840 Speaker 1: So for our listeners who haven't seen this, can you 360 00:16:50,879 --> 00:16:52,799 explain what overseer os does. 361 00:16:53,159 --> 00:16:57,000 Speaker 2: It's a system designed to reverse engineer YouTube channels, but 362 00:16:57,120 --> 00:16:59,039 it doesn't just give you a chart of views. It 363 00:16:59,120 --> 00:17:02,519 goes deep breaks down exactly what's working in their videos, 364 00:17:02,799 --> 00:17:06,200 the hooks, the pacing, the topic selection and its surfaces 365 00:17:06,200 --> 00:17:10,359 patterns instantly and practically hands you a full content strategy and. 366 00:17:10,359 --> 00:17:13,559 Speaker 1: The comparison the before versus after that the source laid 367 00:17:13,559 --> 00:17:15,599 out is just nuts. 368 00:17:15,759 --> 00:17:20,599 Speaker 2: It is previously that kind of analysis deeply studying a competitor, 369 00:17:20,640 --> 00:17:24,359 analyzing their audience retention, their thumbnail strategy. That would take 370 00:17:24,400 --> 00:17:26,960 a team of human analysts weeks. 371 00:17:26,559 --> 00:17:28,119 Speaker 1: To compile weeks of work. 372 00:17:28,200 --> 00:17:31,039 Speaker 2: They'd have to watch hours of video, take notes, build spreadsheets, 373 00:17:31,039 --> 00:17:34,839 have multiple meetings to discuss findings. Overseers does a more 374 00:17:34,880 --> 00:17:38,160 thorough job in minutes minutes, So the role of the 375 00:17:38,160 --> 00:17:42,119 analyst didn't immediately disappear, but the oversight of that analysis 376 00:17:42,200 --> 00:17:47,079 lagged catastrophically behind the execution. The human manager receiving that 377 00:17:47,160 --> 00:17:49,920 report can't possibly verify all the data points in the 378 00:17:49,960 --> 00:17:51,480 two minutes it took to generate them. 379 00:17:51,519 --> 00:17:52,480 Speaker 1: They just have to trust it. 380 00:17:52,559 --> 00:17:55,079 Speaker 2: They have to, and that leads directly to this concept 381 00:17:55,079 --> 00:17:57,039 the Source calls the phantom organization. 382 00:17:57,200 --> 00:17:58,079 Speaker 1: It's a haunting term. 383 00:17:58,279 --> 00:18:00,599 Speaker 2: It is. The ORG chart on the wad wall looks 384 00:18:00,640 --> 00:18:03,240 exactly the same. The headcount in the HR system is 385 00:18:03,240 --> 00:18:05,799 the same. Everyone is still showing up to the office 386 00:18:05,880 --> 00:18:09,079 or logging on from home, but the humans are no 387 00:18:09,160 --> 00:18:10,640 longer steering the ship. 388 00:18:10,720 --> 00:18:11,640 Speaker 1: They're just passengers. 389 00:18:11,720 --> 00:18:15,359 Speaker 2: They're just rubber stamping the machine's output because the machine 390 00:18:15,400 --> 00:18:18,839 is moving too fast for them to actually provide meaningful 391 00:18:18,960 --> 00:18:20,279 direction or oversight. 392 00:18:20,599 --> 00:18:22,119 Speaker 1: I feel like we've all been in that one meeting 393 00:18:22,119 --> 00:18:25,200 where you just say yes to something because you didn't 394 00:18:25,200 --> 00:18:27,039 have time to read the fifty page report they sent 395 00:18:27,079 --> 00:18:29,640 an hour before. Right now, I imagine that happening fifty 396 00:18:29,680 --> 00:18:32,079 times a day for every major decision. 397 00:18:32,359 --> 00:18:35,880 Speaker 2: That is the coordination collapse. You lose control, not because 398 00:18:35,920 --> 00:18:40,039 you're lazy or incompetent, but because you are fundamentally outpaced. 399 00:18:40,160 --> 00:18:44,519 Speaker 1: And this transitions perfectly into Part four, the economic and 400 00:18:44,599 --> 00:18:48,640 human toll, because if companies are running on this phantom autopilot, 401 00:18:49,079 --> 00:18:51,599 what does that do to the broader economy? The Source 402 00:18:51,680 --> 00:18:54,519 predicts something it calls a silent crash. 403 00:18:54,559 --> 00:18:57,480 Speaker 2: This is a fascinating idea because it challenges our traditional 404 00:18:57,559 --> 00:19:00,519 view of a recession, usually a crash that means GP 405 00:19:00,680 --> 00:19:03,079 goes down, stock markets tank, everyone. 406 00:19:02,799 --> 00:19:05,160 Speaker 1: Panics, yeah, red arrows everywhere. 407 00:19:05,319 --> 00:19:08,039 Speaker 2: But the source argues, we might see an economic shock 408 00:19:08,160 --> 00:19:10,960 without a recession, so GDP. 409 00:19:10,640 --> 00:19:12,519 Speaker 1: Could actually go up. We could look prospers on. 410 00:19:12,480 --> 00:19:15,400 Speaker 2: Paper, it could skyrocket. Productivity is going through the roof 411 00:19:15,480 --> 00:19:19,480 because of these agents. But the value the profits from 412 00:19:19,480 --> 00:19:23,720 that productivity, the source argues, it concentrates heavily and very 413 00:19:23,799 --> 00:19:26,599 quickly on those who control the speed. 414 00:19:27,160 --> 00:19:29,119 Speaker 1: Speed based monopoli exactly. 415 00:19:29,440 --> 00:19:32,000 Speaker 2: We used to worry about monopolies based on who owns 416 00:19:32,000 --> 00:19:35,720 all the resources like oil. Then we worried about monopolies 417 00:19:35,759 --> 00:19:39,039 based on network effects, like who has all the users 418 00:19:39,079 --> 00:19:41,720 on their social platform. This is a new kind of 419 00:19:41,720 --> 00:19:45,319 monopoly based on pure tempo right. The companies that can 420 00:19:45,359 --> 00:19:48,400 price their products, ship their goods, and negotiate their contracts 421 00:19:48,440 --> 00:19:51,839 faster than their rivals win every single time. It stops 422 00:19:51,839 --> 00:19:54,759 being about acting better. It's about acting before the competition 423 00:19:54,799 --> 00:19:57,039 can even finish processing the data. 424 00:19:57,079 --> 00:19:59,559 Speaker 1: It's like high frequency trading, but for everything. 425 00:19:59,680 --> 00:20:02,880 Speaker 2: For every applied to logistics, applied to marketing applied to 426 00:20:02,960 --> 00:20:06,480 hiring applied to R and D, and the disconnect there. 427 00:20:06,640 --> 00:20:10,240 The social problem is that while productivity is rising, we're 428 00:20:10,279 --> 00:20:14,279 making more stuff faster. Wages for the average person don't 429 00:20:14,319 --> 00:20:18,079 necessarily follow the games flow to the margins, They flow 430 00:20:18,119 --> 00:20:20,319 to the system owners. They flow to the people who 431 00:20:20,400 --> 00:20:22,599 own the AI that's moving at light speed. 432 00:20:22,920 --> 00:20:26,480 Speaker 1: So the economic pie gets bigger, but the slices get 433 00:20:26,759 --> 00:20:29,279 impossibly thinner for the average worker. 434 00:20:29,200 --> 00:20:32,880 Speaker 2: Or disappear entirely, and that leads to the psychological toll. 435 00:20:33,240 --> 00:20:35,119 This is the part of the analysis that I found 436 00:20:35,160 --> 00:20:38,559 most profound. The concept of learned irrelevance. 437 00:20:39,119 --> 00:20:41,240 Speaker 1: This one hit me hard when I read it. I 438 00:20:41,240 --> 00:20:43,440 think a lot of people are feeling the edges of 439 00:20:43,440 --> 00:20:45,279 this in their jobs already, even if they don't have 440 00:20:45,319 --> 00:20:46,160 a name for it. 441 00:20:46,160 --> 00:20:49,279 Speaker 2: It's distinct from a more well known term, learned helplessness. 442 00:20:49,880 --> 00:20:51,960 Learn helplessness is when you feel like you can't do 443 00:20:52,000 --> 00:20:54,880 anything to change your situations. You give up, you stop trying. 444 00:20:54,880 --> 00:20:56,680 Speaker 1: Okay, so learned irrelevance is different. 445 00:20:57,000 --> 00:21:00,799 Speaker 2: It's more subtle and maybe more insidious. You still show 446 00:21:00,880 --> 00:21:03,079 up to work, you still participate in the meetings, you 447 00:21:03,079 --> 00:21:07,160 are still employed, but you realize, deep down in your gut, 448 00:21:07,279 --> 00:21:11,079 that your input, your opinion, your experience carries less and 449 00:21:11,160 --> 00:21:13,680 less weight simply because you are too slow. 450 00:21:13,799 --> 00:21:15,319 Speaker 1: It's the erosion of human intuition. 451 00:21:15,720 --> 00:21:20,319 Speaker 2: Yes, that's our superpower, right. Humans rely on gut feelings, 452 00:21:20,640 --> 00:21:24,319 We make decisions based on incomplete information and years of 453 00:21:24,400 --> 00:21:28,559 pattern matching in our heads. But AI relies on comprehensive data. 454 00:21:29,079 --> 00:21:32,559 And when the machine is consistently right, or at least 455 00:21:32,559 --> 00:21:35,759 consistently faster and good enough, you start to second guess 456 00:21:35,799 --> 00:21:36,279 your own gut. 457 00:21:36,440 --> 00:21:39,000 Speaker 1: You hesitate. You think, do I override the system's recommendation? 458 00:21:39,079 --> 00:21:41,599 Do I trust myself? I mean, the machine analyzed ten 459 00:21:41,599 --> 00:21:43,480 million data points in three seconds, and I. 460 00:21:43,440 --> 00:21:47,160 Speaker 2: Just I have a hunch precisely, And that hesitation creates 461 00:21:47,200 --> 00:21:50,960 decision paralytus. Do I overcare the system? Do I trust myself? 462 00:21:51,000 --> 00:21:53,119 And eventually you just stop fighting it. You defer to 463 00:21:53,160 --> 00:21:55,559 the machine, not because you were fired, not because you 464 00:21:55,559 --> 00:21:58,039 were forced to, but because arguing with a system that 465 00:21:58,079 --> 00:22:00,000 operates at the speed of light just feels pointless. 466 00:22:00,319 --> 00:22:04,279 Speaker 1: It's a quiet form of disengagement. It's not a loud panic. 467 00:22:04,640 --> 00:22:08,079 It's a slow fading away. You become a spectator. In 468 00:22:08,119 --> 00:22:08,680 your own job. 469 00:22:08,759 --> 00:22:11,799 Speaker 2: That is the human lag. We stop knowing what is real, 470 00:22:11,920 --> 00:22:15,160 what is earned, what was a human idea versus a 471 00:22:15,200 --> 00:22:18,880 machine suggestion, And that creates a crisis of meaning that 472 00:22:18,960 --> 00:22:22,440 is much much harder to solve than a crisis of employment. 473 00:22:22,880 --> 00:22:25,279 Speaker 1: And naturally, when we feel this profound loss of control, 474 00:22:25,519 --> 00:22:28,000 we look for a savior. We look to our institutions, 475 00:22:28,000 --> 00:22:29,759 We look to the government. We say, hey, there should 476 00:22:29,759 --> 00:22:32,039 be a law slow this down. Of course, but part 477 00:22:32,079 --> 00:22:34,640 five of our discussion today explains why the source is 478 00:22:34,680 --> 00:22:37,640 so pessimistic that institutions can save us. 479 00:22:37,720 --> 00:22:40,119 Speaker 2: Yeah, it all comes back to the clocks, the fundamental 480 00:22:40,119 --> 00:22:41,039 mismatch of clocks. 481 00:22:41,119 --> 00:22:42,799 Speaker 1: Election speed versus compute speed. 482 00:22:42,839 --> 00:22:45,279 Speaker 2: Think about the speed of a democracy. It is designed 483 00:22:45,319 --> 00:22:48,079 to be slow. That is a feature, not a bug. Right. 484 00:22:48,680 --> 00:22:51,759 We have election cycles every few years. We have public consultations, 485 00:22:51,799 --> 00:22:54,799 we have legal reviews, we have budget cycles. These things 486 00:22:54,839 --> 00:22:57,039 take years or best months. 487 00:22:57,079 --> 00:22:58,319 Speaker 1: In AI, speed. 488 00:22:58,160 --> 00:23:01,519 Speaker 2: Training cycles for new models are measured in weeks. Deployment 489 00:23:01,559 --> 00:23:04,920 cycles for new applications are measured in days or even hours. 490 00:23:05,240 --> 00:23:07,519 By the time a regulator sits down to draft a 491 00:23:07,599 --> 00:23:11,400 rule to regulate a specific AI behavior, that behavior has 492 00:23:11,440 --> 00:23:15,079 already shifted. The technology has changed three times since the 493 00:23:15,079 --> 00:23:16,480 committee was even formed. 494 00:23:17,240 --> 00:23:19,920 Speaker 1: The source used the social media precedent here, and it's 495 00:23:19,960 --> 00:23:21,640 a really chilling comparison. 496 00:23:21,720 --> 00:23:25,200 Speaker 2: It's the perfect historical analogy. Look at social media regulation. 497 00:23:25,799 --> 00:23:29,000 It arrived after our attention spans were rewired. It arrived 498 00:23:29,039 --> 00:23:32,599 after our politics were fundamentally reshaped. It arrived a decade 499 00:23:32,680 --> 00:23:33,960 after the damage was done. 500 00:23:34,039 --> 00:23:36,359 Speaker 1: We are always always fighting. 501 00:23:36,039 --> 00:23:38,680 Speaker 2: The last war, and with AI the timeline is even 502 00:23:38,759 --> 00:23:41,519 more compressed. We don't have a decade of a wild 503 00:23:41,559 --> 00:23:44,240 West social media era before the rules come in. We 504 00:23:44,319 --> 00:23:46,920 might have eighteen months. And there's another trap here. The 505 00:23:46,960 --> 00:23:49,559 source points out the global coordination trap. 506 00:23:49,759 --> 00:23:51,960 Speaker 1: This is the classic prisoner's dilemma of AI. Right. 507 00:23:52,079 --> 00:23:55,839 Speaker 2: It is if one country, say the US or the EU, says, okay, 508 00:23:56,319 --> 00:23:58,400 we are going to tap the brakes. We are going 509 00:23:58,440 --> 00:24:01,079 to slow down development to ensure safety and alignment. 510 00:24:01,160 --> 00:24:05,079 Speaker 1: What happens another country season opportunity and speeds up exactly. 511 00:24:05,160 --> 00:24:07,160 Speaker 2: Nobody wants to fall behind. Nobody wants to be the 512 00:24:07,240 --> 00:24:09,640 nation that missed the Industrial Revolution two point zero because 513 00:24:09,640 --> 00:24:13,799 they were too cautious. So even if policymakers understand the risks, 514 00:24:13,920 --> 00:24:17,319 and many of them do. The global incentive structure forces 515 00:24:17,440 --> 00:24:19,240 everyone to keep their foot on the gas. 516 00:24:19,559 --> 00:24:22,000 Speaker 1: So to sum it all up, we have a technology 517 00:24:22,000 --> 00:24:24,720 that moves at the speed of light, institutions that move 518 00:24:24,759 --> 00:24:28,079 at the speed of bureaucracy, and a global competitive environment 519 00:24:28,119 --> 00:24:30,440 that punishes anyone who tries to slow down. 520 00:24:30,599 --> 00:24:32,799 Speaker 2: It really does feel like a timing collapse. 521 00:24:33,000 --> 00:24:35,720 Speaker 1: That's the summary of the whole situation, isn't it. The 522 00:24:35,799 --> 00:24:38,640 disruption of twenty twenty six isn't about a robot waking 523 00:24:38,680 --> 00:24:41,599 up and saying hello. It's about the collapse of timing. 524 00:24:42,160 --> 00:24:44,960 We're playing an old game, a game of careers and 525 00:24:45,079 --> 00:24:48,039 laws and gradual change, while the rules of the game 526 00:24:48,079 --> 00:24:50,519 are being rewritten in real time by a player who 527 00:24:50,559 --> 00:24:51,920 moves a million times faster. 528 00:24:52,079 --> 00:24:54,599 Speaker 2: The damage comes from that misalignment. 529 00:24:54,039 --> 00:24:58,079 Speaker 1: Yes, between human systems, which are gradual and linear, in 530 00:24:58,160 --> 00:25:01,759 AI systems which are sudden and exponential. When those two 531 00:25:01,839 --> 00:25:05,160 lines on the grapt diverge far enough, the system snaps. 532 00:25:05,599 --> 00:25:06,680 Speaker 2: That is the flinch. 533 00:25:06,759 --> 00:25:10,279 Speaker 1: It's a sobering realization. We started this episode talking about 534 00:25:10,279 --> 00:25:12,799 twenty twenty six as if it's some date in the future. 535 00:25:13,119 --> 00:25:16,039 But if this analysis is right, these mechanisms are already 536 00:25:16,079 --> 00:25:18,119 grinding into gear all around us. 537 00:25:18,599 --> 00:25:21,240 Speaker 2: And as we wrap up, there is one final thought 538 00:25:21,319 --> 00:25:24,200 the source leaves us with, and it's a provocative one. 539 00:25:24,319 --> 00:25:25,480 Speaker 1: It's that we. 540 00:25:25,440 --> 00:25:28,200 Speaker 2: Won't all agree on when the moment happened. We're not 541 00:25:28,240 --> 00:25:30,200 going to get a notification on our phones that says 542 00:25:30,240 --> 00:25:31,720 AGI has now arrived. 543 00:25:31,799 --> 00:25:32,839 Speaker 1: There's no starting gun. 544 00:25:33,000 --> 00:25:36,440 Speaker 2: No, we will only agree afterwards that something fundamental shifted. 545 00:25:36,519 --> 00:25:37,519 It'll be in hindsight. 546 00:25:37,640 --> 00:25:40,400 Speaker 1: The shock doesn't announce itself with a press release. 547 00:25:40,519 --> 00:25:44,000 Speaker 2: It arrives quietly in the code, in the workflows, in 548 00:25:44,079 --> 00:25:46,519 the silence of a manager who doesn't check a report 549 00:25:46,559 --> 00:25:49,839 anymore because they just don't have time. It arrives quietly, 550 00:25:50,119 --> 00:25:51,079 and then it's all at once. 551 00:25:51,359 --> 00:25:53,839 Speaker 1: That is well, it's a lot to process, but that 552 00:25:53,920 --> 00:25:56,160 is why we do this show, to try and look 553 00:25:56,200 --> 00:25:58,799 the future in the eye, even when it's moving faster 554 00:25:58,880 --> 00:25:59,240 than we are. 555 00:25:59,519 --> 00:26:03,759 Speaker 2: Absolutely, awareness is the first step to navigating this kind 556 00:26:03,759 --> 00:26:04,240 of speed. 557 00:26:04,559 --> 00:26:07,880 Speaker 1: So here's my question for you, our listener, as you 558 00:26:07,960 --> 00:26:12,920 sit here with us in February twenty twenty six, looking 559 00:26:12,960 --> 00:26:15,960 at this speed asymmetry that we've been talking about, do 560 00:26:16,039 --> 00:26:19,359 you feel learned irrelevance creeping into your own. 561 00:26:19,200 --> 00:26:20,880 Speaker 2: Work That's the real question, isn't it. 562 00:26:20,960 --> 00:26:23,759 Speaker 1: Are you steering the ship or are you just rubber 563 00:26:23,759 --> 00:26:26,480 stamping the speed? Are you the analyst with the insight 564 00:26:26,680 --> 00:26:28,759 or are you becoming the clipboard holder in front of 565 00:26:28,799 --> 00:26:29,440 the fire hose. 566 00:26:29,519 --> 00:26:30,960 Speaker 2: We really want to know what you think. 567 00:26:31,039 --> 00:26:33,799 Speaker 1: Let us know your stand in the comments. Are we 568 00:26:33,839 --> 00:26:37,519 ready for the flinch? This has been thrilling Threads. Thanks 569 00:26:37,559 --> 00:26:39,319 for listening and good luck out there. 570 00:26:39,480 --> 00:26:41,799 Speaker 2: Stay curious and try to stay fast.