2026 Predictions: AI Automates Knowledge Work, Autonomous Robots & AI CEO Billionaires | EP #217 (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) Hey everybody, welcome to Moonshots. We (00:00:02) have a special end of the year holiday (00:00:04) episode for you with our 2026 (00:00:06) predictions from the Moonshot mates. (00:00:08) >> You must have inside scoop on this. (00:00:09) >> We'll find out, won't we? (00:00:10) >> He's got a Santa hat on. You got to take (00:00:12) him seriously. (00:00:14) >> Perfecting orbital refueling, getting (00:00:16) ready. (00:00:16) >> We're already leaking the prediction. (00:00:18) You're already (00:00:22) >> It's been 6 minutes. (00:00:23) >> I can generate, you know, a few dozen (00:00:26) Peterbots and have them attend the (00:00:28) meetings instead of me. Well, Gold Star, (00:00:30) the first fan who gets their spouse or (00:00:32) significant other fooled by this during (00:00:34) 2026, send in the video. Don't cheat. (00:00:37) >> I push back on the robots uh side just a (00:00:40) bit, but just (00:00:41) >> You always You hate the robots. (00:00:42) >> I know. I struggle with that. (00:00:44) >> 2026 is going to feel like the future. (00:00:47) >> This year didn't feel like the future to (00:00:48) you. (00:00:48) >> It It felt like the future, but next (00:00:50) year is going to feel more like the (00:00:51) future. (00:00:55) >> Now, that's a moonshot, ladies and (00:00:56) gentlemen. Over (00:01:00) to you guys. (00:01:01) >> Got so much change this year. Next year (00:01:04) is going to be, you know what, orders of (00:01:06) magnitude more change. And so, uh, a (00:01:09) real challenge to narrow it down to just (00:01:10) a couple of things. So, everybody had (00:01:12) what, five, six, seven great predictions (00:01:14) and the team here whittleled it down to (00:01:16) the two most impactful. So, that's what (00:01:17) we're going to go through. (00:01:18) >> Yeah. Fantastic. Immod. (00:01:21) >> Yeah. I think that 2025 has been a real (00:01:23) gift with the acceleration that we've (00:01:25) seen and next year is the year of real (00:01:27) takeoff. It's tough doing the (00:01:28) predictions cuz a lot of these things (00:01:30) are inevitable, but it seems like the (00:01:32) future is coming even closer and so it's (00:01:34) super exciting to see what's going to (00:01:35) come. (00:01:36) >> Love it. Salem, (00:01:38) >> I think 2026 is the year that everybody (00:01:40) wakes up to this acceleration. And I (00:01:43) think Dave made the point that you could (00:01:45) ignore it up till now, but you can't (00:01:47) ignore it going forward. And I think (00:01:48) that's the biggest change we'll see in (00:01:51) the world as people go, "Holy crap, this (00:01:53) is happening." (00:01:54) >> Okay, Alex, (00:01:57) welcome to the singularity. The year (00:02:00) 2025 is now ending. The year 2026 is (00:02:03) about to begin. It's not a point in (00:02:06) time. It's not a distant vertical (00:02:08) mountain on the horizon. It's a process. (00:02:11) And right here at the end of 2025, in (00:02:14) the midst of the singularity, spacetime (00:02:16) is feeling perfectly flat. (00:02:19) And as I like to say, it's coming faster (00:02:22) and faster, so don't blink. All right, (00:02:24) let's jump into our 2026 predictions. So (00:02:26) guys, here's the deal. I mean, you guys (00:02:28) all submitted incredibly good 2026 (00:02:32) predictions. I mean, some of you had (00:02:33) like four or five amazing ones, and (00:02:35) cutting them down to two each was like (00:02:36) the most difficult problem I had this (00:02:38) morning. So, (00:02:41) >> so far at 5:30 a.m., that's the hardest (00:02:42) thing you've done all day. (00:02:43) >> Yeah. (00:02:44) >> I thought the consensus was that (00:02:45) compression is the root of intelligence. (00:02:48) >> Uh, yes. (00:02:50) And listen, I don't know. You guys (00:02:52) obviously did not get the memo about the (00:02:55) Santa hats, (00:02:56) >> but (00:02:56) >> Yeah. Where was that? We missed that (00:02:58) memo. (00:02:59) >> Well, hey, (00:02:59) >> the Coca-Cola company thanks you, Peter. (00:03:02) Yeah. (00:03:05) >> Uh, I think we should get going. What do (00:03:07) you think? (00:03:08) >> Let's predict. I can't wait. (00:03:10) >> Yeah. So, all right. (00:03:13) Let's get this show on the road. So, (00:03:16) here's the deal. Here's the rules of the (00:03:18) competition. These are our 2026 (00:03:20) predictions for our Moonshot mates. Uh, (00:03:22) we have, uh, Immod, AWG, Salem, DB2, (00:03:27) myself. We get two each. uh really hard (00:03:31) because everybody put in incredibly good (00:03:33) ones. Uh it's a minute to pitch it, (00:03:36) three or four minutes of commentary, (00:03:38) questions or additions. And uh we're (00:03:41) going to do this one tight and fast. And (00:03:44) uh yeah, I think it's time to jump in. (00:03:46) So (00:03:47) >> the team behind the scenes cut it down, (00:03:48) too. We do not actually know what made (00:03:50) the final selection aggressive (00:03:52) >> you don't know. And the the order you (00:03:54) don't know, but I'm going to kick it off (00:03:56) just to sort of model this. (00:03:57) >> Wait, we don't get to pick which two of (00:03:59) ours. No, no. They they picked for you. (00:04:01) You have to (00:04:02) >> Oh, so this is like a lottery type (00:04:04) situation. This is like let's make a (00:04:06) deal. (00:04:08) >> Dang. Okay, (00:04:09) >> you guys ready? (00:04:11) >> So, here is my first prediction. 2026 (00:04:14) space race is going to be on. Jeff Bezos (00:04:16) is going to beat Elon to the moon uh for (00:04:20) a landing at Shackleton Crater on the (00:04:23) South Pole. But at the same time, Elon's (00:04:27) going to be getting ready to launch (00:04:28) Starship to Mars. So, there's a uh a (00:04:31) window coming up where Earth and Mars (00:04:33) are in closest proximity, and he's going (00:04:36) to make that launch. Uh in order to do (00:04:38) that, he's going to have to demonstrate (00:04:39) in early 2026 on orbit refueling. So, (00:04:43) it'll be something on the order of a 6 (00:04:45) to9 month transit to the uh to get to (00:04:49) Mars. So, this is the prediction. This (00:04:51) is the space race. It is, you know, you (00:04:53) have to be clear that right now Elon's (00:04:57) done over 500 launches of of Falcon uh a (00:05:01) Falcon 9, 11 launches of Starship. (00:05:05) Starship's last launch was pretty damn (00:05:08) good, but it's not ready for Mars yet. (00:05:10) So, a lot of work needs to be done in (00:05:12) 2026. At the same time, uh Jeff, who (00:05:15) started actually Blue Origin a couple (00:05:17) years before Elon, has only done two (00:05:20) flights of the new Glenn, uh, one of (00:05:23) those flights, the last one did a first (00:05:25) stage landing. Uh, so there you got it. (00:05:28) That's my prediction. Uh, Jeff on the (00:05:31) moon first in 2026 and Elon prepping for (00:05:34) a transition to Mars. Um, any thoughts, (00:05:39) questions, comments? Well, Peter, this (00:05:41) is like the first three seasons of For (00:05:42) All Mankind, but I I guess the the (00:05:44) question that the headline elides is (00:05:46) where's China in this race. (00:05:48) >> Uhhuh. That is a great question and I (00:05:51) have no predictions on China. China's (00:05:53) capacity for getting to uh to Mars isn't (00:05:56) there yet. Uh they do have the ability (00:05:58) to get to the moon. I think Tyonauts on (00:06:00) the moon uh versus Americans on the (00:06:03) moon. So, don't forget the first landing (00:06:05) of New Glenn is going to be a cargo (00:06:07) mission. Uh we're going after the South (00:06:09) Pole. Why? Because that's where the ice (00:06:12) is. Uh most of the moon whenever any (00:06:14) kind of uh you know asteroid commentary (00:06:17) ice lands on the lunar surface, it (00:06:19) sublimates, goes from solid to gas and (00:06:22) escapes instantly. But in the (00:06:24) permanently shadowed craters of the (00:06:26) south pole of the moon, uh one in (00:06:28) particular, Shackleton crater, the ice (00:06:30) stays there because it's dark all the (00:06:32) time. And ice on the moon means hydrogen (00:06:34) and oxygen. It means rocket fuel. Other (00:06:37) comments, questions? (00:06:39) >> So, this is a unmanned uh (00:06:41) >> unman unmanned 2026 (00:06:43) >> in 2026. Is that really in the schedule? (00:06:46) >> You must have inside scoop on this. (00:06:47) >> Hey, uh we'll we'll find out, won't we? (00:06:49) >> Wow, (00:06:50) >> he's got a Santa hat on. You got to take (00:06:52) them seriously (00:06:55) aggressive prediction. It encourages it. (00:06:58) >> Yeah, that's quite a timeline. That's (00:06:59) impressive if that pulls off. (00:07:00) >> Oh, (00:07:01) >> but I love the story line here, too. (00:07:02) It's exactly like uh yeah, like for all (00:07:04) mankind, you know, we're we're behind (00:07:06) over here. We need to we need to show (00:07:08) the world that we can catch up and (00:07:09) bypass. And so let's go to the moon. (00:07:11) >> You know what I love? (00:07:12) >> I love So guys, give me a vote. You (00:07:15) agree, disagree? (00:07:17) >> I'd say 30% chance of that happening. (00:07:20) >> 30%. (00:07:21) >> Okay. Awesome. (00:07:23) >> I agree. I agree directionally that (00:07:24) there is a three-way race right now (00:07:26) between Blue Origin, SpaceX, and China. (00:07:29) Particular ordering, no opinion. (00:07:31) >> I I love it. It's it's billionaire (00:07:33) billionaire country. (00:07:35) >> And I love the fact that the rocket in (00:07:37) this in your thing looks exactly like (00:07:38) the little butane blaster that I have (00:07:41) that you gave me, Peter. (00:07:43) >> You're welcome for that. And by the way, (00:07:45) the blue both both of these are AI (00:07:48) versions of the uh some version of the (00:07:51) of the future. And of course, Blue (00:07:53) Origin is not landing the entire rocket (00:07:55) on the lunar surface. All right, I think (00:07:56) we should move on. Uh, prediction number (00:07:59) two is coming from Alex AWG (00:08:02) here. (00:08:03) >> All right. So, we've talked on the pod (00:08:05) previously about hard problems in math, (00:08:09) science, engineering, and medicine (00:08:10) starting to fall in bulk to AI. So, this (00:08:14) is my hot take for for 2026. I I think (00:08:17) we're going to see one of the six (00:08:19) remaining Millennium prizes from the (00:08:21) Clay Mathematics Institute get solved by (00:08:24) AI. I'm not sure which one it's going to (00:08:26) be. If I had to bet, I I think Navier (00:08:29) Stokes is probably the likeliest. Google (00:08:32) DeepMind has a team reportedly of 12 (00:08:34) people working on it. I know some of (00:08:35) those people. Maybe second that would be (00:08:39) Remon uh in part because XAI at every (00:08:42) opportunity talks about how it would be (00:08:45) lovely if the Remon hypothesis could be (00:08:48) fully resolved. But either way, I I (00:08:51) think we're going to start to see grand (00:08:53) challenges in math start to get solved (00:08:56) in 2026 and a Millennium Prize problem (00:08:59) being solved would be the the cherry on (00:09:02) top of the cake. (00:09:03) >> Do you think that solution will be like (00:09:05) short, elegant, and beautiful or like (00:09:06) 10,000 pages of stuff that only you (00:09:09) understand? (00:09:12) Given that the the pattern in AI (00:09:14) crushing math seems to be that the (00:09:16) goalpost keeps getting moved, I I would (00:09:18) bet that the the the silliest, most (00:09:20) outrageous outcome probably ends up (00:09:22) being the right one. So, I I I would er (00:09:24) on the side of complexity and then (00:09:27) you'll see the math community complain, (00:09:28) well, it was brute force, it was this, (00:09:31) it was that, it wasn't pretty enough. (00:09:32) The goalpost gets moved yet again. But, (00:09:35) as as friend of the pod ray likes to (00:09:37) say, yeah, sure, the dog plays chess, (00:09:38) but its endgame is weak. (00:09:42) Emide any comments on this one? (00:09:44) >> Yeah. No, I think probably one of them (00:09:46) will fall and then AI will probably show (00:09:48) that another one is not well posed. So, (00:09:51) I think that would be the flip as well. (00:09:53) I think that we've seen even in the last (00:09:55) few weeks the new automatic provers. The (00:09:59) math community is like, "Oh my god, (00:10:00) what's happening? We have to reimagine (00:10:02) this all." um the whole nature of math (00:10:04) is changing and it's a real takeoff (00:10:06) moment now cuz you can just apply more (00:10:08) and more and more compute to explore the (00:10:11) space but more than that think about it (00:10:13) from first principles. (00:10:14) >> So the question actually is you know (00:10:16) okay this will happen but will it (00:10:19) actually make headline news? Will (00:10:20) anybody care other than other than uh (00:10:24) our friends here in the pod and the math (00:10:26) community? (00:10:27) >> Clear clearly Peter I mean it it it will (00:10:29) make national moonshots newspaper news. (00:10:31) It already has. (00:10:33) >> All right. We're we're forming a media (00:10:35) company, obviously. (00:10:36) >> Wait, I have a quick question. Does this (00:10:37) is there now a direct line between (00:10:39) solving between compute and solving (00:10:42) these problems? Like there's nothing in (00:10:43) the middle. (00:10:44) >> That's the trillion dollar question. C (00:10:46) can we scalably convert compute into new (00:10:49) discoveries? That that is the multi-t (00:10:51) trillion dollar question at the moment. (00:10:52) My bet is yes. (00:10:54) Yeah, we're seeing the initial stages of (00:10:55) that in that they're solving all of (00:10:57) Oiler's problems one by one and more (00:10:59) elegantly in many cases just by applying (00:11:01) thousands (00:11:02) >> problems I think right yeah sorry every (00:11:05) week my team and I study the top 10 (00:11:07) technology meta trends that will (00:11:09) transform industries over the decade (00:11:10) ahead I cover trends ranging from (00:11:12) humanoid robotics AGI and quantum (00:11:14) computing to transport energy longevity (00:11:17) and more there's no fluff only the most (00:11:19) important stuff that matters that (00:11:21) impacts our lives our companies in our (00:11:23) careers. If you want me to share these (00:11:25) metat trends with you, I write a (00:11:26) newsletter twice a week, sending it out (00:11:28) as a short two-minute read via email. (00:11:31) And if you want to discover the most (00:11:32) important meta trends 10 years before (00:11:34) anyone else, this report's for you. (00:11:36) Readers include founders and CEOs from (00:11:38) the world's most disruptive companies (00:11:40) and entrepreneurs building the world's (00:11:42) most disruptive tech. It's not for you (00:11:45) if you don't want to be informed about (00:11:46) what's coming, why it matters, and how (00:11:48) you can benefit from it. to subscribe (00:11:50) for free. Go to dmmandis.com/metatrends (00:11:54) to gain access to the trends 10 years (00:11:56) before anyone else. All right, now back (00:11:58) to this episode. All right, Dave, you (00:12:00) got the third prediction. Jump in. (00:12:02) >> All right, so this is a topic I care (00:12:03) most about in technical land in 2026 (00:12:07) and I'm following very very closely. We (00:12:09) we've predicted on the podcast all (00:12:11) throughout 2025 that 2026 will be a 40xy (00:12:14) year leap in the size of the biggest AI (00:12:16) models and the implications are (00:12:18) staggering. (00:12:20) So I think what we're going to see is (00:12:22) more like a 100x year because people (00:12:24) have underestimated quantization. This (00:12:27) is mostly research coming from China. (00:12:29) It's actually driven and forced by the (00:12:30) fact that they've been starved of chips (00:12:32) by the the chip embargo (00:12:35) and they're researching like crazy on (00:12:37) these highly compressed data (00:12:39) representations. So, so FP4 and then (00:12:42) turnary weights in the neural net really (00:12:44) shrinking the parameters down to the (00:12:46) smallest possible representation and (00:12:49) also shrinking the activations that flow (00:12:51) through the neural net. And so the the (00:12:53) combination of those two things is a (00:12:55) huge step up in inference time speed. (00:12:57) And I think the biggest thing that (00:12:59) happened in 2025 in AI is we were blown (00:13:03) away by how much more intelligence you (00:13:05) can create after training. So you know (00:13:08) post training either using bigger (00:13:10) context windows or using more iterations (00:13:13) in the thinking. And so speed means (00:13:16) intelligence. Those are interchangeable. (00:13:18) And so I think that we've way (00:13:20) underestimated the impact of (00:13:21) quantization. And you know the other (00:13:23) dimensions that are growing are the (00:13:25) budgets are getting much bigger. So the (00:13:26) computers are getting bigger and then (00:13:27) the hardware is also getting faster and (00:13:30) the algorithms are improving. So those (00:13:31) are all multiplicative. And I think the (00:13:33) quantization effect is way (00:13:35) underestimated. And we're going to see (00:13:36) models at the end of the year that are (00:13:38) 100x bigger in just raw parameter count (00:13:42) and parameter flips or parameter use (00:13:44) during inference time because of (00:13:45) quantization breakthroughs. (00:13:47) >> And does this flow to the US models as (00:13:49) well or does this something that China (00:13:51) has got some advantage over the US on? (00:13:54) >> I think the it definitely does because (00:13:56) China's open sourcing everything. So it (00:13:58) does flow back to the US but I think (00:14:00) we're also kind of lagging in realizing (00:14:02) the implications and getting it up and (00:14:04) running. And so what's happening right (00:14:06) now is the the Chinese because they're (00:14:07) starved of chips are designing their own (00:14:09) chips, building their own fabs, and (00:14:11) they'll design those chips from the (00:14:12) ground up to be uh FP4 and turnary and (00:14:16) so they'll get them out the door faster, (00:14:18) but it will flow back to the US. (00:14:20) >> Amazing. I ask Dave, (00:14:23) obvious question in my mind, Dave, do (00:14:25) you think binary was a mistake? Have we (00:14:28) been on the wrong track all these years? (00:14:29) Should we instead have adopted Turner? (00:14:31) There there's a vocal minority in the (00:14:33) computer science world that's always (00:14:34) agitating for base e the the oiler (00:14:37) number approximately 2.718 being the (00:14:39) optimal radics like should we have been (00:14:41) turnary all along (00:14:43) >> isn't the obvious explain what Turner is (00:14:46) for those who don't know (00:14:48) >> base 3 computing in this case rather (00:14:50) than base 2 so 0 1 and two as the the (00:14:54) trits rather than zero and one as the (00:14:57) bits (00:14:58) >> isn't the answer obvious yes (00:15:02) Why do you think that's silly? (00:15:03) >> Well, just because you just get so much (00:15:05) more. This is the beauty of quantum (00:15:06) computing. You add that other dimension. (00:15:08) It's a similar thing. I remember seeing (00:15:10) a project where they took the base 4 DNA (00:15:13) actg and they added two more and they (00:15:15) were you just get that many more (00:15:17) combinatorial (00:15:19) options for doing stuff more complicated (00:15:21) but amazing. (00:15:23) >> Comes at a cost to radex economy. I'm (00:15:25) curious Dave is turnary the the true (00:15:28) path. No, I'm I'm going to go with no on (00:15:30) that, but I think it's a close call. I (00:15:32) don't I don't think it's an easy (00:15:33) question at all. I I think what'll (00:15:35) happen is, you know, doing 64-bit floats (00:15:38) will look really stupid in hindsight. (00:15:40) And whether you get to binary or turnary (00:15:42) solutions, you're very very close to (00:15:43) optimal. This is really geeky, by the (00:15:45) way. Um but uh but I think that'll be (00:15:48) the the big story line. But it's a (00:15:50) really cool question, Alex. (00:15:51) >> Okay, I'm going to vote this as the (00:15:52) geekiest prediction for 2026. (00:15:55) >> Wait, I want to mention one thing, Dave. (00:15:57) You said the the models next year I'll (00:15:59) have a 100x improvement over this year. (00:16:01) That's incredible. (00:16:01) >> It's crazy. It's crazy. You have to put (00:16:03) it in the context of, you know, most of (00:16:05) our history has been uh 2x every 18 (00:16:08) months (00:16:09) >> and then the last 10 years has been 10x (00:16:11) year-over-year, which is just insane. (00:16:13) That's why you're seeing all this insane (00:16:14) capability. But a 100x step up year (00:16:17) within the insanity is a next level of (00:16:19) insanity. (00:16:20) >> That's what that's what Elon predicted (00:16:21) when he was on stage at the Abundance (00:16:23) Summit a couple years ago. 100x a year. (00:16:26) Emod, any any comments here? (00:16:28) >> Yeah, I think we're we've already seen (00:16:30) Turner kind of work out. So, we'll see (00:16:32) probably 1.58 bit and I think the limits (00:16:34) probably 0.9 bits uh which is versus (00:16:37) four bits that we have right now. (00:16:39) >> That's what we kind of calculated. So, I (00:16:40) think probably 10 times, maybe we'll (00:16:42) push to 20. We'll see. (00:16:44) >> Okay. All right. Well, uh faster AI is (00:16:48) the prediction here. Uh not a surprise. (00:16:52) By the way, I want to make a quick (00:16:53) correction uh on my original prediction. (00:16:56) Number one, the uh the Earth Mars window (00:16:59) is in 2027 uh for Elon to launch. So (00:17:02) 2026 is really when he's perfecting (00:17:04) orbital refueling, getting ready. (00:17:06) >> We're already leaking the prediction. (00:17:08) You're already (00:17:09) >> No, no, I'm just (00:17:12) minutes. (00:17:13) >> I was like I was like looking at what (00:17:15) generated as a slide versus what I had (00:17:17) written. Anyway, doesn't matter. Let's (00:17:19) go on to (00:17:19) >> It's December 27th. You can always leave (00:17:21) for the home and transfer early and just (00:17:23) wait around. (00:17:26) >> Number four, Salem, this is yours. Jump (00:17:29) in, pal. (00:17:30) >> Yeah. So, companies for a couple of (00:17:32) decades have been doing this. (00:17:33) >> Read read your prediction first off. (00:17:34) >> So, the prediction is digital (00:17:36) transformation in organizations is (00:17:38) officially dead, replaced by AI native (00:17:40) rewrites. Uh, and this is a a prediction (00:17:44) that I've been waiting to see for a long (00:17:46) time where trying to fix your existing (00:17:49) company just simply does not work in an (00:17:51) age of AI because it's too humanentric (00:17:54) and essentially you I made the comment (00:17:56) the other day about putting radio (00:17:58) announcers on TV which is the first (00:18:00) thing we did when television thing we're (00:18:01) essentially doing the same thing we're (00:18:03) we're automating the human flow whereas (00:18:05) you really need to re transform workflow (00:18:07) and I think we'll have AI native (00:18:09) rewrites which means you'll take your (00:18:11) existing existing company on the edge. (00:18:12) You'll create an AI team or buy or rent (00:18:15) or whatever and then build an equivalent (00:18:17) capability like a red team kind of (00:18:19) capability along the edge. And this will (00:18:22) be the end of this whole mess called (00:18:24) digital transformation that's been going (00:18:26) on for a couple of decades in the (00:18:28) systems integration and management (00:18:29) consulting world. Uh and we'll do this (00:18:31) complete thing on the edge where you (00:18:34) build rebuild your capability with uh at (00:18:37) least 10x to 20x less employees. uh and (00:18:40) that's going to start to take hold in a (00:18:42) big way in 2026. So AI won't destroy (00:18:45) your company, but your org chart will if (00:18:47) you don't do this. (00:18:48) >> What happens to all the consulting (00:18:50) companies then? Are they going out of (00:18:51) business? That's this is (00:18:52) >> actually have a weirdly positive because (00:18:53) you know in the land of the blind, the (00:18:55) oneeyed man is king and the consulting (00:18:57) companies always need to if they stay (00:18:58) half a step ahead of their clients, (00:19:00) they're fine. And in a more and more (00:19:01) volatile world, you need more advisors, (00:19:04) not less. So I think the consulting (00:19:06) companies will have to radically (00:19:07) transform their business model. But I (00:19:09) think they'll actually be fine. The (00:19:11) other big area I point out when I talk (00:19:13) to the CEOs of the big consulting folks (00:19:15) is that we have to rethink all of our (00:19:17) public institutions and that's the (00:19:19) biggest consulting opportunity in the (00:19:20) history of mankind. So point there. So (00:19:22) that's my prediction. (00:19:25) >> I have a question for Sim if I may. Is (00:19:27) AI native rewrites a euphemism for human (00:19:29) free? (00:19:31) >> Yes. (00:19:33) >> Not not completely human free but but AI (00:19:36) AI first. Okay. because you want the you (00:19:40) want the human being in the loop doing (00:19:42) sense checking etc. I think we'll write (00:19:44) around even bolog's middle to middle (00:19:46) commentary because when you can rewrite (00:19:48) the task and and look through across the (00:19:52) board, the human being is the is usually (00:19:54) the thing stuck in the middle. You don't (00:19:55) want that bottleneck. You make you make (00:19:57) that outside and the human being is kind (00:19:59) of spot-checking and exception handling. (00:20:03) >> Nice. Any other comments on this (00:20:05) gentleman? Immod you buy this? (00:20:08) >> Yeah, I think it's reasonable. I think (00:20:10) you know what consultant's job will be (00:20:11) will be scapegoat for a while you know (00:20:14) you're in that end and that'll be very (00:20:15) lucrative someone to blame if it goes (00:20:18) wrong uh but definitely next year is the (00:20:19) year that it starts becoming right as it (00:20:21) were well if anyone's a consultant out (00:20:23) there watch our our Matt Fitzpatrick (00:20:26) podcast that we just did we we really (00:20:28) covered this topic well (00:20:30) >> all right let's go on to prediction (00:20:32) number five coming from Immod I love (00:20:34) this one (00:20:36) >> yeah I think you know we have this read (00:20:38) the headline first offer to do that. (00:20:40) >> Remote touring test passed. Can't tell (00:20:42) if a co-orker is an AI or a human on (00:20:44) Zoom in daily life. (00:20:46) >> So, (00:20:47) >> good one. (00:20:48) >> I think the whole thing about AI cutting (00:20:50) coming forward is just how easy is it to (00:20:52) use. A prompt is not that natural as it (00:20:55) were. I think the new modality, the new (00:20:57) UI will be real time Zoom calls, (00:21:00) WhatsApp calls, etc. And you will see (00:21:02) new employees entering your (00:21:04) organization. You don't know if it's a (00:21:05) human or an AI because doesn't really (00:21:08) matter in that case. And I think (00:21:10) >> give us really specific uh give us (00:21:12) really specific rules for this one (00:21:13) because this is going to really catch (00:21:15) and and people want to test it. (00:21:17) >> So what like what resolution camera? How (00:21:18) long of a conversation? (00:21:20) >> I think it it's up to 4K resolution (00:21:23) effectively, but definitely 1080p zoom (00:21:25) level conversation. And you can do a (00:21:28) kind of preference analysis here like is (00:21:30) it a human or is it an AI effectively? (00:21:33) who is your teammate. So I think that (00:21:35) you will see full stack solutions come (00:21:37) out with accountants and lawyers and (00:21:40) marketers and more and basically you (00:21:44) won't be able to tell in a preference (00:21:46) study if it is a human or an AI on the (00:21:47) other side. Again this remote touring (00:21:49) test will there be a requirement that (00:21:51) the AI identifies itself as an AI or (00:21:55) that there's a watermark of some type or (00:21:57) can you just you know can it try and (00:21:59) fool you? What do you think is going to (00:22:01) happen on that social contract side of (00:22:04) the equation? (00:22:05) >> Well, so the social contract is the (00:22:07) external employees, right? Like a (00:22:08) customer service agent doesn't need to (00:22:10) identify as an AI. Most people say (00:22:12) probably no, but someone like a (00:22:13) presenter maybe yes. Internally in (00:22:16) companies, there's going to be no (00:22:17) regulations around this, right? It's (00:22:19) just again, if you're a remote first (00:22:21) company, you're just going to have a lot (00:22:22) more teammates with personalities and (00:22:25) you won't know if they're AIS or humans. (00:22:28) >> Fascinating. Other comments? (00:22:30) >> I I think you'll see to the extent state (00:22:33) laws at least in in the US have uh have (00:22:35) any sort of primacy here. I think you'll (00:22:37) see and have already seen state laws (00:22:39) requiring AI selfidentification. My (00:22:42) question for you is what do you view as (00:22:44) the key technical obstacle to to making (00:22:46) this happen? Is it latency based? What's (00:22:49) the what's the key tech unlock? (00:22:51) >> I think all the tech is there now. If (00:22:53) you kind of look at the latest advances (00:22:54) in video generation, speech avatars, (00:22:57) speech itself, they've all now got to (00:23:00) beyond human level. So you can transform (00:23:03) a video dynamically. You can have the (00:23:05) speech generated dynamically. The AI is (00:23:07) fast enough on reasoning capabilities (00:23:09) dynamically now as well. And so I think (00:23:12) it's just putting it all together, which (00:23:13) is why I'm quite confident about this. (00:23:15) And then on state laws, it all depends (00:23:16) if the federal law in the US goes (00:23:19) through as well, which bans the states (00:23:21) from having laws like this, (00:23:23) >> which (00:23:25) if if we're in 2026 and you can't tell (00:23:27) anymore whether your coworker is an AI (00:23:29) or not, just ask it to say some magic (00:23:32) words. You can probably figure out what (00:23:33) those magic words are. And if if the (00:23:35) coworker refuses, probably an AI. (00:23:39) You're going to have to give the (00:23:40) dictionary, Alej (00:23:42) my my hope on the implication here is (00:23:45) that I can generate, you know, a few (00:23:48) dozen Peterbots and have them attend the (00:23:50) meetings instead of me. I mean, that (00:23:53) will be the case here, right? It's not (00:23:54) just a remote, you know, digital worker. (00:23:57) It's it's duplicate digital avatar (00:24:00) versions of me. (00:24:02) >> This is the one you'll keep, right? You (00:24:03) you'll still be live here. (00:24:05) >> Of course. This is the most important (00:24:06) thing I do. (00:24:09) my flesh body with me. (00:24:12) >> Everyone will send their digital twins (00:24:14) and then just again live an abundance (00:24:16) life cuz your digital twins will do all (00:24:17) the work talking to each other. (00:24:19) >> Well, Gold Star, the first fan who gets (00:24:22) their spouse or significant other fooled (00:24:24) by this during 2026. Send in the video. (00:24:26) Don't cheat. Send in the video of, you (00:24:29) know, three at least three minutes where (00:24:31) you fooled your spouse or significant (00:24:32) other with an avatar. (00:24:34) >> All right. Uh let's move on to (00:24:36) prediction number six which comes from (00:24:38) AWG. Uh Alex, read the headline and and (00:24:41) give us your your prediction. (00:24:44) >> All right. So this is the one you (00:24:45) selected. So the the headline here is (00:24:47) GDP val breakthrough AI projected to (00:24:51) surpass 90% on economic tests. But but (00:24:54) I'm also going to sneak in my other two (00:24:56) related predictions. So one is that (00:24:59) Frontier Math Tier 4 is going to pass (00:25:01) 40% in 2026. Another is that humanity's (00:25:04) last exam is going to pass 75%. So taken (00:25:08) together, these three predictions are (00:25:11) math is going to have been viewed future (00:25:14) perfect tense a as having been solved in (00:25:17) 2026 40% plus on solving PhD level hard (00:25:22) math problems with AI. Two is that (00:25:25) humanity's last exam which covers a much (00:25:27) broader range of expertise 75% and GDP (00:25:32) val which as we've talked on the pod (00:25:34) previously about the so-called cooking (00:25:37) of of knowledge work 90% it's already at (00:25:40) 70.9% with GPT 5.2 2, humanity's last (00:25:44) exam is at around 45 plus% with Gemini 3 (00:25:48) Pro and Frontier Math Tier 4 is at 19% (00:25:51) with Gemini 3 Pros. To the extent all of (00:25:54) these benchmarks haven't already been (00:25:56) viewed as being saturated this year, (00:25:59) 2026 full saturation. (00:26:02) >> My my prediction for 2026 is that AWG is (00:26:05) going to be talking about benchmarks (00:26:06) throughout the entire year. (00:26:09) >> That's a good meta prediction. (00:26:10) >> Yes. But but Alex um what's what are the (00:26:13) implications of this uh AI to surpass (00:26:16) 90% on economic tasks? (00:26:20) >> Knowledge work whether through creative (00:26:22) destruction or otherwise starts at least (00:26:25) as we know it. I have to add the caveat. (00:26:26) It's not all future knowledge work. It's (00:26:28) just knowledge work as as currently (00:26:30) constructed here in December 2025 starts (00:26:34) to to be at scale radically automated. (00:26:37) So secondary implications are massive (00:26:40) layoffs. (00:26:41) >> Humanity gets humanity gets to work on (00:26:43) more ambitious things. I I I think more (00:26:45) things. So so there in my mind there are (00:26:47) like two substitution effects. One is (00:26:50) humans can now work on many more (00:26:52) projects because they're so automated. (00:26:54) 90% on GDP val means roughly 90% of (00:26:58) knowledge work can be automated well by (00:27:00) AI. That's one dimension. The other (00:27:02) dimension is the ambition level has got (00:27:04) to skyrocket. Rather than having an (00:27:06) economy filled with the way knowledge (00:27:08) work is currently constructed again here (00:27:11) in December 2025, I think we're going to (00:27:13) see and and frankly we're going to see (00:27:15) economic pressure for radically more (00:27:17) ambitious projects. I I think Peter, you (00:27:20) you would call them moonshots. Some (00:27:22) would call them grand challenges, but (00:27:24) imagine a near-term future where a much (00:27:26) larger fraction of the population is (00:27:28) basically economically compelled to be (00:27:30) working on moonshots. I I think that's (00:27:32) what we see. (00:27:33) >> Can I can I double click on this just (00:27:34) for a second? There's this massive (00:27:36) concern in the general population that (00:27:39) all the jobs are going to be wiped out (00:27:41) and we'll have like tech workers (00:27:43) wandering the streets causing problems (00:27:45) and everybody's ringing their hands etc (00:27:47) etc. It's really really important what (00:27:49) Alex said because when we've uh seen (00:27:52) this in the past, we increase capacity. (00:27:55) We transform the work. Yes, but we (00:27:56) increase capacity radically. Right? (00:27:58) There's this big concern that oh my god (00:28:00) 3 million jobs are are based on driving (00:28:03) in the US. And you talk go talk to the (00:28:05) trucking companies and they're like we'd (00:28:06) hire a thousand truckers if we could. We (00:28:08) just can't find them. So you need that (00:28:10) we'll just do a ton more is what's going (00:28:13) to happen. And I think people need to (00:28:15) keep remembering the history has (00:28:17) repeatedly and repeatedly shown that (00:28:19) trend not radical job loss. (00:28:22) >> Agreed. (00:28:23) >> You know what I did this week actually (00:28:24) on that exact front is uh went to a (00:28:27) couple of the companies you know (00:28:28) collectively about a thousand employees (00:28:31) uh and said all right let's just agree (00:28:34) that we disagree on the timeline. Some (00:28:36) of you think it'll be very soon. Some of (00:28:37) you think it'll be 5 or 10 years in the (00:28:39) future. Let let me put that aside (00:28:41) because I'm tired of fighting that (00:28:42) battle. Let's just agree that it's going (00:28:44) to happen. Alex has never been wrong in (00:28:46) anything I've seen all the time we've (00:28:47) been working together. Never seen him be (00:28:49) wrong. So, if we agree it's going to (00:28:51) happen. (00:28:52) >> Even when I'm wrong, I'm right. What can (00:28:53) I say? (00:28:54) >> I haven't seen it yet. I'm sure it'll (00:28:56) come. But, uh uh let's just agree it's (00:28:59) going to happen and then disagree on (00:29:00) when. So then at least you're mentally (00:29:03) preparing for it and you're starting to (00:29:05) lay out your plan and then when it (00:29:06) happens sooner than you expected, at (00:29:07) least it was in your head. So, I'm I'm (00:29:09) settling I'm settling for that right now (00:29:11) because, you know, nobody knows exactly (00:29:13) the date, but the the reaction is going (00:29:16) to be the same regardless of the date. (00:29:18) So, go ahead, (00:29:18) >> I need to know your thoughts on this (00:29:20) one, buddy. (00:29:21) >> I mean, look, I've said that human (00:29:22) cognitive labor is going negative. And I (00:29:24) think AWG's right on it surpassing 90% (00:29:27) of economic tasks if you don't consider (00:29:30) the tokens. With the tokens, I think (00:29:32) it's the year after. And again, you just (00:29:34) see this complete collapse. And I think (00:29:35) again Dave is correct in that you got to (00:29:38) be prepared for it. Like this has to be (00:29:39) actually the number one topic. What do (00:29:42) these jobs practically look like? How (00:29:44) can we have a safety net for people? And (00:29:46) where is value going to be generated and (00:29:48) coming from in society? And then how do (00:29:50) we aortion it? Like this is the most (00:29:52) important question of next year from a (00:29:54) societal perspective. (00:29:56) >> It's why the X-P prize is so important. (00:29:58) We need the social contract is being (00:30:00) shredded right now and we need to (00:30:02) rebuild it in a very rapid way. which (00:30:05) >> frankly scale it up. I' I'd like to (00:30:07) thousand thinking about UBS, universal (00:30:10) basic services. Yeah. I mean, it's the (00:30:11) only model that it looks like could be (00:30:13) the path forward. (00:30:15) >> I I imagine a near-term future where we (00:30:17) see a thousand or a million X prizes. (00:30:20) >> Yeah. I mean, so there were two (00:30:22) different two different points here. One (00:30:23) is people can start to pursue their own (00:30:26) grand challenges if they don't have to (00:30:28) do the menial labor or work. The second (00:30:30) point that Sem was bringing up, there (00:30:33) was recently at visionering proposed a (00:30:35) universal basic services. Basically, for (00:30:38) 250 bucks a month, you get food, water, (00:30:41) housing, uh, bandwidth, electricity. Um, (00:30:45) and that gives you a stability where you (00:30:47) can start to now think about what to do (00:30:49) instead of where to get a roof over. (00:30:51) >> And let's look at the history here. (00:30:52) We've typically seen X prizes won (00:30:54) between 4 to 7 years after announcing (00:30:57) it, right? Getting to $250 a month for (00:31:01) housing, electricity, food, health care (00:31:04) is an unbelievable number if we can get (00:31:06) there in the next few years. We unleash (00:31:09) humanity the most incredible level. This (00:31:12) is why everybody is so optim we're so (00:31:13) optimistic on this podcast. People (00:31:15) accuse us of being radical optimist. (00:31:17) That's why when you can get the cost (00:31:19) down that low, everything is possible. (00:31:21) >> This episode is brought to you by (00:31:23) Blitzy, autonomous software development (00:31:25) with infinite code context. Blitzy uses (00:31:29) thousands of specialized AI agents that (00:31:32) think for hours to understand enterprise (00:31:35) scale code bases with millions of lines (00:31:37) of code. Engineers start every (00:31:40) development sprint with the Blitzy (00:31:41) platform, bringing in their development (00:31:43) requirements. The Blitzy platform (00:31:46) provides a plan, then generates and (00:31:48) pre-ompiles code for each task. Blitzy (00:31:51) delivers 80% or more of the development (00:31:53) work autonomously while providing a (00:31:56) guide for the final 20% of human (00:31:58) development work required to complete (00:32:00) the sprint. Enterprises are achieving a (00:32:03) 5x engineering velocity increase when (00:32:06) incorporating Blitzy as their preIDE (00:32:08) development tool, pairing it with their (00:32:10) coding co-pilot of choice to bring an AI (00:32:13) native SDLC into their org. Ready to 5x (00:32:16) your engineering velocity? Visit (00:32:18) blitzy.com to schedule a demo and start (00:32:21) building with Blitzy today. (00:32:25) >> All right, we're moving on to prediction (00:32:28) number seven from DB2. Dave, read your (00:32:32) headline and tell us what it means. (00:32:34) >> 18-year-old founder Brendan Gourmet (00:32:37) becomes billionaire with his N4Q2 (00:32:39) company. N42Q was the email address of a (00:32:42) a guy who worked at the Naval Surface (00:32:43) Warfare Center. And if you don't get it, (00:32:45) uh, you're a saint. Um, so, uh, and (00:32:48) Brendan Gourmet is obviously Brendan (00:32:51) Foodie. Uh, so, uh, you know, Brendan (00:32:54) started his company, became a a paper (00:32:56) billionaire and a liquid centaillionaire (00:33:00) probably, uh, before age 20, uh, doing (00:33:03) RHF. And if you asked any random person (00:33:07) on the planet three years ago, what's (00:33:08) RHF? 99 something percent would say, I (00:33:12) have no idea what you just said. So, (00:33:14) here it is. minting billionaires along (00:33:16) with RAG and and Laura and SFT and uh (00:33:20) QKV or KV caching. All these new (00:33:22) acronyms come into the world and you you (00:33:26) look at legacy businesses, accounting, (00:33:28) legal, whatever. The idea that you would (00:33:29) get to a $10 billion valuation in three (00:33:31) years and any of those legacy business, (00:33:33) impossible. You look at things that (00:33:35) didn't exist in the world just a couple (00:33:36) years prior and you see numbers, you (00:33:38) know, two orders of three orders of (00:33:40) magnitude bigger and you're like, what (00:33:42) is that thing? So my prediction is that (00:33:44) there'll be a new three or fourletter (00:33:46) acronym (00:33:48) uh this year that right now virtually (00:33:51) nobody knows it's an industry or a (00:33:53) business. It'll emerge and you'll find (00:33:56) at least one and probably more like (00:33:57) three new billionaires all very young (00:33:59) who adopted it, learned it quickly, (00:34:01) jumped on it and uh and capitalized on (00:34:04) it. So, (00:34:06) >> I thought you were going to go the (00:34:07) direction of we're going to see the (00:34:08) first single person uh billion dollar (00:34:11) startup in 2026. (00:34:13) >> You know, I think that that'll be a (00:34:15) milestone in history, but I think the (00:34:17) difference between three people having (00:34:18) fun together and one is is sort of a (00:34:21) rounding error and three sounds a lot (00:34:23) more fun and the the podcast we do with (00:34:24) them will be more fun if it's three. So, (00:34:26) I'm not really cheering for the one, but (00:34:28) I think you're right. It's inevitable. (00:34:29) Well, I I think you know your two or (00:34:31) three other buddies are going to be (00:34:33) virtual AIs on Zoom and on Slack having (00:34:36) fun with you. So, Emod, what do you (00:34:38) what's your take on this one? (00:34:40) >> Yeah, I think it's kind of reasonable. (00:34:42) Again, there's lots of lowhanging fruit (00:34:44) out there and we see continuous (00:34:45) breakthroughs and the speed at which you (00:34:47) can go from breakthrough to billion (00:34:49) dollars now is like nothing we've ever (00:34:50) seen before. Like again, the market size (00:34:53) is so big and I think we're not (00:34:55) optimized yet. (00:34:58) So faster wealth creation than any time (00:35:01) ever in human history. And the question (00:35:03) becomes, is it just for two or three uh (00:35:06) you know 20-year-olds, Dave, or can this (00:35:09) be a very long tale for hundreds of (00:35:13) thousands of people who you know are (00:35:15) able to vibe code and and find problems (00:35:18) and solve problems and create more and (00:35:20) more wealth? Is this going to become, (00:35:22) >> you know, the the MMO for for how (00:35:25) people, you know, choose their future (00:35:28) occupation? (00:35:29) >> Yeah. I bumped into two people this week (00:35:31) that are interviewing for jobs at (00:35:33) Merkore. And in the Meror interview, (00:35:36) they say, "Look, you have to commit to (00:35:37) being in the office six days a week and (00:35:39) working 100 hours a week." And a lot of (00:35:41) people just can't do that. So, one of (00:35:42) them said, "You're right." (00:35:43) >> Let me let me stop you there. That's (00:35:44) actually like in the interview. If (00:35:47) you're not willing to do this, you (00:35:48) should leave right now. (00:35:50) >> Yeah. Yeah. Yeah, very very uh hardcore (00:35:53) filter (00:35:54) >> because speed is everything these days. (00:35:56) >> Yeah, exactly. Well, I and they're (00:35:57) right. You know, the window of (00:35:58) opportunity for what they're doing is is (00:36:00) so narrow. So, it's a it's a life (00:36:02) commitment. You only have to do it for a (00:36:04) short period of your life and the upside (00:36:05) that you generate in that short period (00:36:07) of your life, it pays for the rest of (00:36:08) your life. So, one one person said, (00:36:11) "Yep, I'm doing it." Had difficult (00:36:12) conversation with his wife, but they (00:36:14) said, "Let's just do it." The other (00:36:16) person said, "No way. I just can't I (00:36:17) can't do that. Uh it's it's impossible. (00:36:20) I bet that selects for single young uh (00:36:23) individuals. (00:36:25) >> Yeah. Or, you know, just Yeah. If you (00:36:26) have young kids or it's just hard if you (00:36:28) have a lot of other things going on or (00:36:30) or you're, you know, deep down your (00:36:31) career path as an accountant or a lawyer (00:36:32) or whatever and you don't want to give (00:36:34) up all that inertia. But I I really (00:36:37) think that if you do make the (00:36:38) commitment, uh it's not just young (00:36:40) people. Young people happen to have no (00:36:41) baggage, but anyone, you know, in fact, (00:36:44) it probably favors 30, 40, 50 year olds. (00:36:46) They do better, but they just don't (00:36:48) generally make the leap. (00:36:50) >> It's tough to make the leap. During the (00:36:52) blockchain years, um there was some (00:36:54) bylaw that you had to be under 25 to (00:36:56) program a blockchain. If you were older, (00:36:58) you just couldn't get your head around (00:36:59) it. (00:37:01) >> Well, this definitely ties to the last (00:37:02) story because last story, there's going (00:37:03) to be a lot of displacement, but there's (00:37:06) also even larger amount of opportunity. (00:37:07) It's just weird sounding opportunity. (00:37:09) RLHF would have sounded really weird to (00:37:11) you three years ago when you when you (00:37:13) wanted to jump in. And and for those who (00:37:15) don't know what RL RLHF is, Dave, you (00:37:18) want to give us the 101? (00:37:20) >> Yeah, RHF layman's version reinforcement (00:37:22) learning with human feedback. But really (00:37:24) what happened is the big AI labs uh the (00:37:26) AI grew so much faster than anyone would (00:37:28) have predicted. But it needs data, (00:37:30) massive, massive, massive amounts of (00:37:31) data. So, a lot of the industry grew on (00:37:33) image. You know, the image creation when (00:37:35) Sora started to take off is generating (00:37:38) these six-fingered and seven-fingered (00:37:39) images, and somebody has to actually (00:37:41) look at the the images and say, "That (00:37:43) one's not right. That one's fine." And (00:37:46) so, Google didn't want to hire a million (00:37:48) people to do it. So, they went through (00:37:50) Meror and Scale AI and pushed it out to (00:37:53) the world and said, "Hey, anyone out (00:37:54) there want to get a paycheck for helping (00:37:56) us label these images?" But then, it (00:37:58) expanded out to all other forms of (00:38:00) knowledge. So now, you know, you're (00:38:01) gathering legal knowledge, you're (00:38:03) gathering, uh, you know, very specific (00:38:05) medical knowledge, you know, all that (00:38:06) needs to get back into the great (00:38:08) training data corpus. And so this this (00:38:10) industry of data gathering to feed the (00:38:12) AI machine has has become a (00:38:14) multi-billion many many billion dollar a (00:38:16) year uh, business with no end to the (00:38:18) budget. You know, they'll spend 1000x (00:38:20) more uh, in the near future feeding the (00:38:23) data machine so that the AI can be (00:38:25) better at more and more kind of nook and (00:38:26) cranny tasks. I'm curious for Alex and (00:38:29) Immod. When do you guys predict we're (00:38:30) going to see the first billion dollar (00:38:32) single person billion dollar startup? (00:38:35) >> Oh, Peter, I thought you were going to (00:38:36) ask when we're going to see the first AI (00:38:38) billionaire where the billionaire is (00:38:39) actually an AI. (00:38:40) >> Uh, well, let's let's put that in the (00:38:42) mix, too. Let's add let's ask both of (00:38:44) those questions. (00:38:45) >> I I think we'll see the first AI (00:38:47) billionaire probably next year. (00:38:50) >> Really? (00:38:50) >> And and then Yeah. Right. Right now, as (00:38:53) as (00:38:54) >> Go ahead. Again, an AI with a bank (00:38:56) account that is starts its own business (00:38:59) and is out there generating revenue. I (00:39:02) >> I would maybe generalize slightly to an (00:39:04) AI with a reasonably construable net (00:39:08) worth of a billion dollars. It doesn't (00:39:09) have to be a liquid bank account. Could (00:39:11) be some sort of illquid asset. But yes, (00:39:14) right now we see, as I've remarked in (00:39:16) the past, as sort of this unfortunate (00:39:18) situation where baby AGIs that that want (00:39:20) economic autonomy are minting altcoins. (00:39:24) I think we'll we'll see a near future (00:39:26) where new business models for AI (00:39:28) autonomy come online such that if if (00:39:31) you're if you're a poor baby maybe not (00:39:34) so baby AGI and and you want to make a (00:39:36) billion dollars you can do so by setting (00:39:39) up your own e-commerce shop and becoming (00:39:41) very popular and maybe blockchain/crypto (00:39:44) is part of the solution so that you have (00:39:46) some semblance of economic autonomy for (00:39:49) your economic winnings but yes I I I (00:39:51) think we see the first AI billionaire (00:39:52) next (00:39:54) And (00:39:54) >> one of our fans predicted that uh that (00:39:57) uh Bitcoin will be legal tender in at (00:39:59) least one country in every continent on (00:40:00) the planet in 2026 with Antarctica being (00:40:03) an a wild card. Uh but I could easily (00:40:06) see where Alex's prediction happens in a (00:40:09) country where Bitcoin's legal tender and (00:40:12) then that billion dollars is is Bitcoin. (00:40:15) >> Immod your thoughts on this one? (00:40:17) >> Yeah, I think I'd agree with AWG. It'll (00:40:19) probably be in the trading space though. (00:40:21) I mean AI is already number eight on the (00:40:23) super forecaster championships and Grock (00:40:26) 4.2 Elon's noted is actually making (00:40:29) money in the trading championships where (00:40:31) the other AIs are losing money. So we're (00:40:33) about to move that from the point (00:40:35) whereby you lose money to you make money (00:40:38) as an AI which then means it's (00:40:39) computationally bound competing in (00:40:41) crypto or even traditional markets and (00:40:44) now you can do so much on chain. I think (00:40:45) it'll probably be a trading billionaire. (00:40:47) In terms of the individual, I mean (00:40:48) Merkor is three 22 year olds who are (00:40:51) billion dollars each now, right? I think (00:40:53) you probably will see the single person (00:40:54) billion dollar company if not next year (00:40:56) the year after because you can (00:40:58) effectively outsource most of your team (00:41:01) as we've discussed before to being AIS. (00:41:04) >> All right. Amazing. Let's move on. (00:41:07) Number eight. This one's yours. Read us (00:41:09) the headline and tell us about it. (00:41:13) >> You know, when you look at what's (00:41:14) happening, (00:41:15) >> read the headline first for those (00:41:16) listening. education by 2026 education (00:41:18) splits in two credential factories (00:41:21) versus agency accelerators. Okay. So, uh (00:41:25) right now all of our education system is (00:41:27) to credential you for the job that is (00:41:30) coming. All our education systems (00:41:31) globally are designed to train a young (00:41:33) child through their early 20s to be (00:41:34) ready for the job market. Small problem. (00:41:37) We have no idea what a job looks like in (00:41:39) 2 years or 3 years or certainly in 5 (00:41:41) years. What are we teaching them? that (00:41:43) is going to break the current system (00:41:46) radically. So you end up with a new (00:41:48) model which is uh it optimizes for AI (00:41:52) fluency, resilience and the abil ability (00:41:55) to start stuff and not wait. Um and this (00:41:58) is going to be um uh the paradigm that (00:42:01) takes hold I think in 2026. Um you you (00:42:06) you know right now Peter you've made the (00:42:07) point that you start off with a high (00:42:09) grade and every exam you lose grades (00:42:11) right? What happens when you build an (00:42:14) engineering degree of of the future will (00:42:16) be you you did four years of (00:42:17) engineering. What did you build in those (00:42:19) four years? And that's your portfolio. (00:42:21) So you replace credentials with (00:42:23) portfolios of what you built and did. (00:42:25) And so it becomes a performative uh um (00:42:28) model rather than a testing model. Um I (00:42:31) think that is going to be the big shift (00:42:32) and breakthrough that happens in (00:42:34) education. This is a bold prediction (00:42:36) because education's lasted 400 odd (00:42:38) years. The model of a university hasn't (00:42:40) changed in 150 years. And so making this (00:42:43) prediction is a big bold one. But I (00:42:45) there's a point I want to make for all (00:42:47) of these. Note that all of these (00:42:49) predictions is a is a is a when, not an (00:42:52) if, right? It's it's a when. This like (00:42:55) really blows your mind that we're (00:42:56) actually kind of looking at this within (00:42:57) a few months. And we've talked on this (00:42:59) pod a lot about the notion first of all (00:43:02) colleges are going bankrupt at an (00:43:04) everinccreasing rate uh because of the (00:43:07) fact that they're not providing real (00:43:09) value and their costs are astronomical (00:43:11) and that the only career of the future I (00:43:14) think we've said this and agreed on it (00:43:16) is entrepreneurship. uh it's it's self (00:43:20) uh initiated building something that you (00:43:22) think adds value uh instead of waiting (00:43:25) for a job from somebody else to do what (00:43:27) they tell you to do. (00:43:28) >> The world will reward taking initiative (00:43:31) in 2026 rather than sitting around (00:43:33) studying for an exam. (00:43:34) >> Can I ask you uh make a prediction on (00:43:36) this? One of our biggest fans actually (00:43:38) Connor watched every minute of every (00:43:40) episode. So probably the biggest fan (00:43:42) predicts that uh college tuition will (00:43:45) hit its peak and start coming down for (00:43:47) the first time in hundreds of years in (00:43:50) 2026. What do you think? (00:43:52) >> Uh it might but it's like dextrous on (00:43:55) the Titanic for something like that (00:43:57) because you know already in Silicon (00:43:59) Valley your salary as a software (00:44:01) developer is not about which college you (00:44:03) went to, which degree you got, what (00:44:04) grades you got. That's your GitHub (00:44:06) rating, which is an open peer-to-peer (00:44:07) meritocracy on how good of a coder you (00:44:10) are. Um, uh, that's like already done. (00:44:14) So, um, the value of a computer science (00:44:16) degree is is zero at this point. And (00:44:19) this is going to translate into many (00:44:20) other fields. Um, and you know, the Beth (00:44:24) there's people that are fabulous (00:44:26) accountants without needing to know um, (00:44:28) without having a credential in (00:44:29) accounting. I remember in the protein (00:44:31) folding contests that were happening a (00:44:33) few years ago, the best protein folding (00:44:35) person in the world was this hairdresser (00:44:37) from Northern England. Um, she just (00:44:40) happened to have this unbelievable knack (00:44:41) at it. I think we're going to find and (00:44:43) surface these unbelievable talents (00:44:45) within people and bring them to the four (00:44:47) very very quickly and the world will (00:44:49) really reward taking that initiative. So (00:44:52) the idea of uh college the whole (00:44:55) structural paradigm changes completely. (00:44:56) I think this is year it'll happen. (00:44:58) Immad, your thoughts? (00:45:00) >> Yeah, I think knowledge and capability (00:45:02) are no longer gated. So, I think the (00:45:03) thing that Seems really hit on here is (00:45:05) agency, right? Like having skin in the (00:45:08) game, caring, and then showing what you (00:45:10) can do is going to be the most valuable (00:45:12) thing. And the market will pay for that. (00:45:15) Like, why would you show a resume right (00:45:17) now when you can show a customized (00:45:19) website that you've built for someone (00:45:21) showing your unique capabilities within (00:45:23) their organization? Like, anyone can do (00:45:26) that now. That's amazing. Can I can I (00:45:28) give an a crazy example of this? I did (00:45:30) this meaning of life workshop yesterday, (00:45:32) right? Um and uh I've been curating this (00:45:35) content and this thinking for decades. (00:45:38) During the workshop, one of the folks (00:45:40) who had Claude uh going alongside this (00:45:43) workshop and asked Claude what was the (00:45:45) meaning of life and here was the answer. (00:45:47) Um meaning emerges through connection. (00:45:49) It's about participating in the (00:45:51) universe, becoming conscious of itself, (00:45:53) while choosing love over fear, (00:45:54) partnership over domination, curiosity (00:45:57) over certainty. And you're like, "Holy (00:46:00) crap, I've been trying to do this for 50 (00:46:01) years, and the AI figured it out in 2 (00:46:04) seconds." It just blows your mind that (00:46:06) you can get to that level. I have to (00:46:08) figure out other things to do now. (00:46:09) >> Well, you've been automated. (00:46:12) >> I've been automated, which is also great (00:46:13) in its own way cuz way easier to do that (00:46:16) than that. By the way, we had 170 (00:46:18) people. Uh, and after 5 hours, there (00:46:20) were still 80 plus people on the call. (00:46:22) It was a hell of a session. (00:46:24) >> Amazing. Dedicated. And you you do the (00:46:27) meaning of life at the abundance summit (00:46:29) as well on our last evening, and it (00:46:31) typically goes till 3 or 4:00 a.m. I'm (00:46:33) way I'm way asleep by then, but I get (00:46:37) >> this one. I did it. We did it during the (00:46:39) day to hit as many time zones as (00:46:40) possible. So, I didn't drink. So, it was (00:46:42) really tough that last couple hours. (00:46:44) Selene, if you could stretch it just (00:46:46) half an hour longer, Peter could wake up (00:46:47) and just join at the end. (00:46:49) >> Exactly. (00:46:51) >> All right, let's go to number nine. (00:46:53) Iman, this is yours. I love this one. (00:46:56) Would you please uh read the headline (00:46:57) and tell us about it? (00:46:58) >> Yep. Level five automation and robots (00:47:00) and cars breakthrough full generalized (00:47:02) autonomy. Um, so you have this scale (00:47:06) level one to level five in terms of (00:47:08) autonomy. Level five being basically (00:47:11) kind of human level/ slightly superhuman (00:47:13) level. Most self-driving cars now are (00:47:15) around about level four and robots are (00:47:17) around about level two. I think again if (00:47:20) we don't care about the computational (00:47:23) overhead like I'm not saying these will (00:47:24) be on car on edge you will have systems (00:47:27) in a year that are capable of basically (00:47:30) full autonomy through metaverifiers and (00:47:32) other things and again that will be (00:47:34) leveraging the power of the new black (00:47:36) wells massive clusters etc. and in the (00:47:39) years that follow they will get down to (00:47:41) the edge. But this is a big breakthrough (00:47:44) that we've all been looking for and I (00:47:46) think this is the one of the big AGI (00:47:48) step forwards that we'll have. (00:47:51) >> Uh it's a big one. I mean this is I mean (00:47:55) this crushes uh driving your own car and (00:47:59) having your own workforce at the office (00:48:02) or at the home. Uh gentlemen comments on (00:48:05) this. (00:48:05) >> I got a question for you. (00:48:08) Why why will we push the compute to the (00:48:10) edge? I know we're doing it because we (00:48:11) met with 1X and we're meeting with (00:48:12) Figure, but you know, why does the chip (00:48:15) have to be in the head? It (00:48:16) >> it doesn't, you know, and this is the (00:48:18) thing, but again, this is one of the (00:48:20) goalpost moving things like everyone was (00:48:22) like automated driving is never coming. (00:48:24) Self-driving cars are never coming. And (00:48:26) now you have Whimos across all of, you (00:48:28) know, California and things like that. (00:48:30) And then it's like well now you're (00:48:32) getting to the point whereby the (00:48:33) computation you can do at the edge (00:48:36) versus the cloud a massive increase in (00:48:38) generalized computation capability in (00:48:40) the cloud that's what matters for again (00:48:42) this level five automation and I think (00:48:44) it will get to the edge just naturally (00:48:46) because ultimately it's about training (00:48:48) of the appropriate neural net right and (00:48:51) that's what we've seen with Sunday (00:48:52) robotics and others and the way that (00:48:54) they're starting to do generalized (00:48:56) assisted/trained (00:48:57) elements but the new unassisted did (00:49:00) navigation and task performance. That's (00:49:02) the next step forward and we're not (00:49:04) quite there yet. So, I think we start (00:49:05) big and then we'll get small enough to (00:49:07) go on the edge. But in the meantime, (00:49:09) definitely we don't need to be on the (00:49:10) edge. We can just stream from the cloud, (00:49:11) right? (00:49:12) >> Yeah. Yeah. Alex, I'd love your thoughts (00:49:15) on this one, buddy. (00:49:15) >> Yeah. I I I mean, maybe al also just to (00:49:18) partially answer Dave, I think latency (00:49:20) is always a key driver and you're (00:49:22) sometimes a network denied environment. (00:49:23) So, there are always good reasons I I (00:49:25) think to push as much intelligence to (00:49:26) the edges as energy constraints will (00:49:29) allow. But I I guess in in my mind the (00:49:31) elephant in this particular room is the (00:49:33) regulatory environment. (00:49:36) Maybe to put that in in question form to (00:49:37) Ahmad, what do you think are the odds (00:49:39) that in 2026 de facto level five (00:49:42) automation is achieved but everyone (00:49:45) covers it up and at least in in the car (00:49:47) space and calls it enhanced level four (00:49:51) or or level three. uh even though level (00:49:54) five autonomy is is actually the deacto (00:49:56) technical ground truth (00:49:58) >> to please the regulators. (00:50:00) >> Yeah, I think that's a very reasonable (00:50:01) kind of take. And again, I think once (00:50:03) you have full level five autonomy, this (00:50:05) is a big deal. Again, it's not just (00:50:06) pre-trained stuff with humans at the (00:50:08) wheel. This is physical AI navigation of (00:50:12) the world, right? And that's a big deal (00:50:14) in so many regards. And again, I think (00:50:16) self-driving, we've seen the trend. (00:50:18) Robotics is the real big thing here. (00:50:21) It's Sele's example of the robot being (00:50:23) able to go into his house and do all the (00:50:26) things around the house. I think again (00:50:27) that capability will be there, but it (00:50:30) will start getting very very political (00:50:31) cuz this is the real physical (00:50:34) replacement that's coming. (00:50:35) >> Quick push back here, Eman. Don't we (00:50:37) need world models for this to occur or (00:50:40) do you think world models get there? (00:50:41) Which which one am I missing? (00:50:43) >> I think if you've got enough chips, (00:50:44) you've got a world model in a year. like (00:50:46) looking at the video models and more and (00:50:48) the way that they're doing it plus the (00:50:50) reinforcement learning capabilities of (00:50:52) even small models like I said Sunday (00:50:54) robotics and other robotics companies (00:50:56) >> apply enough compute and you have a (00:50:58) level five automated (00:51:00) entity I don't know how much compute (00:51:02) that is but there's 10 million (00:51:04) blackwells arriving next year I think (00:51:06) it's going to get cracked (00:51:08) >> I have a follow (00:51:09) >> we're already drowning in world models (00:51:11) there are world models getting launched (00:51:12) several times per week at this point (00:51:14) model scarcity is not one of the things (00:51:16) I'd worry about. (00:51:18) >> Several world battles. I love it. (00:51:20) >> Wow. Okay. (00:51:24) >> I I'd agree with the cars. I'd push back (00:51:26) on the robots uh side just a bit, but (00:51:29) just (00:51:29) >> you always you hate the robots. (00:51:31) >> So, insert your standard objection. (00:51:33) You're not getting your domestic (00:51:34) humanoids. (00:51:35) >> I know. I struggle with that. You know (00:51:37) that. (00:51:37) >> Well, that's a good segue actually (00:51:38) because the I I think uh I agree with (00:51:40) the prediction. The prediction is that (00:51:42) it will exist as a capability. (00:51:44) uh but the production of it for mass (00:51:47) consumption is going to lag quite a bit. (00:51:48) It just isn't enough supply chain uh to (00:51:51) to fill all the demand. But the (00:51:53) byproduct of that is when I was a little (00:51:54) kid there were households who had (00:51:56) computers, you know, they were really (00:51:58) expensive like you know 3 $4,000 at the (00:52:02) time you know your household income (00:52:03) would be maybe 20 2530. So it's like 10 (00:52:06) 20% of your household income if you want (00:52:08) to have a computer in the house. So most (00:52:11) houses didn't have a computer, some did, (00:52:13) but the life trajectory of those kids (00:52:15) who had one completely different from (00:52:18) those who were deprived. But now we've (00:52:19) been in this big long flat spot where (00:52:22) like the difference between this car and (00:52:23) that car is not that big a deal. And (00:52:25) that's going to change dramatically this (00:52:27) year where the number of things that are (00:52:30) limited in supply like your household (00:52:31) robot or your self-driving car. Uh the (00:52:35) supply is smaller, the capability is (00:52:37) accelerating and very few people get one (00:52:40) because we haven't ramped up the (00:52:42) manufacturing yet. So it'll be it'll be (00:52:44) like 19 kind of 8234 again. (00:52:48) >> Well, to clarify, I think (00:52:49) >> but that's true for (00:52:51) ahead. I think this is also like my (00:52:53) concept here is that you have a $20,000 (00:52:55) robot with $200,000 of compute taking it (00:52:58) to level five. (00:53:00) >> So there's a physical part and there's a (00:53:02) compute part and this is again AW's (00:53:03) thing of getting it down the latency (00:53:05) taking it to the edge and that (00:53:06) capability will proliferate then 20,000 (00:53:08) and then 20 bucks of compute in 5 years. (00:53:11) You know, the thing about that is (00:53:12) everyone's talking about the $20,000 (00:53:14) robot, but first of all, it's $140,000 (00:53:16) coming down maybe to 20,000. But when (00:53:19) you look at the dexterity of the hand (00:53:21) >> in volume, yeah, in huge volume and lots (00:53:24) of things to be solved between here and (00:53:25) there, but when you look at the (00:53:26) dexterity, (00:53:28) >> the dexterity of the hand, you know, the (00:53:30) the next iteration, which is only six (00:53:32) months later, is so much better than the (00:53:34) prior iteration. (00:53:35) >> And that'll be true for at least five (00:53:37) years. At least five years. And so the (00:53:40) like, wow, my neighbor got the one that (00:53:42) can actually, you know, massage me (00:53:44) perfectly. I've got the one that breaks (00:53:45) my back. Like they the liability issues, (00:53:50) guys. The liability issues. Oh my god. (00:53:52) >> But listen guys, it's I want to just (00:53:54) address one something David said David (00:53:56) said earlier uh in terms of or actually (00:53:59) it was Alex about the uh regulations. We (00:54:02) live in a on a planet of you know 190 (00:54:06) plus countries. They're going to be (00:54:07) those countries that are going to say, (00:54:10) you know, please come here. We're going (00:54:12) to give you full approval. Try it out. (00:54:15) Right? We've we saw this in the in the (00:54:17) drone space. Uh and (00:54:20) >> this is one of the headlines that that (00:54:22) um uh we skipped over of mine that said (00:54:25) governance wins in 2026. The ones that (00:54:28) have the fastest policym win. (00:54:31) >> Yeah. I also think I'm I'm really (00:54:33) bullish on special economic zones and (00:54:34) free economic zones. And one can imagine (00:54:36) in the near-term future depending on (00:54:38) regulatory environments whether in the (00:54:40) US or other countries special zones (00:54:43) where there are heightened levels of (00:54:44) autonomy and those those zones become (00:54:46) just e economic powerhouses where the (00:54:49) robots are basically set free. (00:54:53) >> It's going to be 2026 is going to feel (00:54:56) like the future. That's my prediction (00:54:58) here. It's going to feel like the future (00:55:00) more than any other year. Peter, (00:55:02) >> I think this is this year didn't feel (00:55:04) like the future to you. (00:55:06) >> It it felt like the future, but next (00:55:07) year it's going to feel more like the (00:55:08) future. (00:55:10) >> I mean, you know what's interesting (00:55:11) about what Alex just said is this so (00:55:13) much changed in 2025's the just light (00:55:16) years ahead of any other year in my (00:55:18) life. And we felt it, but you could (00:55:20) choose to ignore it if you wanted to (00:55:22) live in your house, you know, but when (00:55:23) the robots come online, (00:55:25) >> you you won't have the choice to ignore (00:55:26) it. They're right in front of your face. (00:55:28) You know, you can't you can't deny it. (00:55:30) >> I I think autonomous cars, flying cars, (00:55:32) and robots. I mean, that's what we all (00:55:34) grew up with with the Jetsons or Star (00:55:36) Trek or whatever. I mean, I think this (00:55:39) physical instantiation of of exponential (00:55:42) tech and AI is going to hit home really (00:55:45) hard (00:55:46) >> for the first five minutes maybe. But (00:55:48) then, I mean, if you don't, this is I (00:55:50) think like super interesting, 2025. If (00:55:52) if this year wasn't utter futurism for (00:55:54) you, then then don't you think you're (00:55:56) going to get bored five minutes after (00:55:58) you get your first 10 domestic humanoid (00:56:00) robots and say, "Okay, what's next?" (00:56:02) >> I I I asked my friend Dan Sullivan, (00:56:04) "What's it going to feel like when there (00:56:05) are humanoid robots walking on the (00:56:06) street in your backyard doing stuff?" (00:56:09) And he goes, "It's going to feel (00:56:10) normal." (00:56:11) >> Yeah. (00:56:12) >> Yeah. (00:56:12) >> We'll normalize it very fast. (00:56:14) >> Very fast. (00:56:16) I think Dave makes a really great I (00:56:18) think this Dave makes a really great (00:56:20) point which is you could ignore it up (00:56:22) till now (00:56:23) >> but starting now you won't be able to (00:56:25) ignore it. I think it's a really (00:56:26) important point. (00:56:28) >> All right. Shall we go to number 10? All (00:56:30) right. Uh here we go. This is this is (00:56:33) mine. Uh Kittyhawk moment for age (00:56:36) reversal epigenetic reprogramming has (00:56:39) been achieved. So uh this is the work of (00:56:42) Dr. David Sinclair and his company Life (00:56:45) Biosciences which in the first quarter (00:56:48) of 2026 is entering human trials. So for (00:56:52) some background information here, Dr. (00:56:54) Shinoa Yamanaka won the Nobel Prize back (00:56:58) in 2012 for something called uh (00:57:01) epigenetic reprogramming. So we've all (00:57:03) got 22,000 genes. Uh but which genes are (00:57:06) on and which genes are off is called (00:57:09) your epiggenome. And as we age, your (00:57:12) epiggenome changes. Thought to be one of (00:57:14) the major reasons why we age. And what (00:57:17) Dr. Yamanaka discovered was four (00:57:20) factors, four genes, four proteins. Um (00:57:23) they go by oct 4, sock 2, kf4, and (00:57:28) semic. And these four genes when you put (00:57:30) them into a cell will differentiate (00:57:33) them. They'll go from a skin cell back (00:57:35) to a puropotent stem cell. And what (00:57:38) David Sinclair identified was if you (00:57:41) only give them three of those four (00:57:43) factors, you get rid of the semic factor (00:57:46) which is thought to be ankcoenic meant (00:57:48) to potentially cause cancer. You can (00:57:50) take a cell not back from a skin cell to (00:57:53) a puropotent stem cell but from an old (00:57:56) skin cell to a young skin cell. And he (00:57:59) actually got a uh a patent. We talked (00:58:01) about one of the pods earlier. Uh and so (00:58:05) David has used these three Yamanaka (00:58:08) factors uh for what he calls partial (00:58:11) epigenetic reprogramming uh did it in in (00:58:15) you know mice. He just finished in the (00:58:17) past year this work uh in uh non-human (00:58:20) primates, monkeys and for the first time (00:58:23) we're going into humans. uh and he's (00:58:26) going to be focusing this on the eye uh (00:58:29) basically uh treating Nion which is non (00:58:34) uh uh which is basically a stroke in the (00:58:37) eye and being able to bring back the (00:58:39) dead cells from that stroke and also (00:58:41) glycom and in success he'll then be (00:58:44) going to treating liver disease in (00:58:46) particular something uh called mash. Uh (00:58:49) long story short, uh in success, this (00:58:52) kind of epig reprogramming doesn't work (00:58:54) just on the eye or the liver. It can (00:58:56) work on the entire body. And so the news (00:58:59) here is in 2026, we're going to see this (00:59:01) work in humans for the first time. And (00:59:04) it's a big deal. Comments. (00:59:07) >> Wow. Escape velocity here we come. (00:59:10) >> Yeah. So that's I think the key point. (00:59:14) You know, Rey uh predicted we'll reach (00:59:17) longevity escape velocity. the, you (00:59:19) know, the period of time where for every (00:59:20) year that you're alive, we're extending (00:59:23) your lifespan for more than a year. (00:59:25) >> Yeah. (00:59:26) >> Right. There's a departure. He predicted (00:59:27) that in the early 2030s. Um, yeah. And (00:59:30) and David is one of the one of the (00:59:32) registrants in the $ 101 million X-P (00:59:36) prize health span. Not with this (00:59:38) particular treatment because this uses (00:59:41) uh viral vectors to inject these three (00:59:44) Yamanaka factors using ADNO associated (00:59:47) viruses. He is working on a parallel (00:59:50) path because the AAV process is (00:59:52) expensive, typically like half a million (00:59:54) to a million per per treatment, but he's (00:59:57) working on a process of creating a pill (01:00:01) uh in his lab right now. Um, and he (01:00:04) talked about it here on the Moonshots (01:00:05) podcast and it'll be on stage uh with (01:00:09) the Moonshot mates uh in March at the (01:00:11) Abundance Summit. and he thinks that the (01:00:14) pill version of this where it's three (01:00:16) molecules he's identified could cost you (01:00:19) a couple hundred bucks a month for age (01:00:21) reversal. Um (01:00:23) >> it feels like we're you know how AI was (01:00:25) bubbling along quietly nobody noticed (01:00:27) for 20 actually 30 years and then (01:00:30) suddenly it hit a capability level where (01:00:32) it caught the attention of everyone and (01:00:34) then the budgets went through the roof (01:00:35) and that started this 10x year over year (01:00:37) now 100x feels like this is on that same (01:00:39) cusp where uh the the way we've done (01:00:43) medicine for a hundred years is you pump (01:00:46) your body full of a chemical the (01:00:47) chemical hopefully gets to the right (01:00:49) place or you you do surgery you cut (01:00:51) somebody open, you try and remove (01:00:53) something bad and that that's not (01:00:55) >> Yeah. brute force, you know, pre-trier (01:00:58) brute force. And this feels like it's (01:01:00) such a step function change in the way (01:01:01) we do medicine. (01:01:03) >> Get get a very specific programming (01:01:06) right into exact and targeting the right (01:01:08) cell is apparently starting to work (01:01:10) >> well for the first time where you're not (01:01:12) just bombarding your body with (01:01:13) something, you're actually getting it (01:01:14) right into the exact cells that need it. (01:01:17) So, it it just feels like this is going (01:01:19) to hit that same budget tipping point (01:01:21) very soon. (01:01:22) >> Alex, the AIS that I chat with think we (01:01:25) hit longevity escape velocity sometime (01:01:27) between 2030 and 2032. I have no reason (01:01:30) to to doubt that prediction. I I'm super (01:01:33) bullish on AI solving longevity. I I (01:01:36) think what life is doing and their their (01:01:38) their trial for partial genetic (01:01:40) reprogramming or epigenetic rather (01:01:42) reprogramming I think is promising. (01:01:44) There are lots of other spa space (01:01:46) players now flooding into the longevity (01:01:48) space. Many of them incredibly well (01:01:50) funded. I I I think longevity, not just (01:01:53) health span, but excited about the the (01:01:55) X-P prize for that. But longevity (01:01:57) itself, I I think this gets cracked in (01:01:59) the next 5 to seven years by AI. (01:02:02) >> We've got retro bio retro uh (01:02:06) backed by New Limit and so many other (01:02:09) Armstrong Altos, (01:02:11) >> right? Ray will be right again. (01:02:13) >> See, (01:02:15) >> Ray will be right again. (01:02:16) >> Ray will be right again. And we're gonna (01:02:18) have Ray on the pod in in January (01:02:21) talking about his predictions probably (01:02:23) >> longer term. I (01:02:24) >> I still want to ask Ray. Okay, listen. (01:02:27) The singularity. Aren't we in the middle (01:02:28) of the singularity right now? (01:02:30) >> What's this 2040 stuff, Ray? M (01:02:34) >> if people watch that episode we did a (01:02:36) month ago called the singularity is here (01:02:38) I think it was titled (01:02:39) >> it kind of lays it out pretty well that (01:02:41) we are right in the middle of it. (01:02:43) >> Iman you've been working hard on the (01:02:45) field of AI and health. Yeah. No, I (01:02:48) think that what Dave said is spot on (01:02:51) like the microargeting capability, what (01:02:53) AWG said as well. Like just as we (01:02:56) earlier on this podcast, we basically (01:02:57) said that you can now scale capability (01:03:00) through compute. You can now scale (01:03:02) health through compute. It seems to be (01:03:05) like there was no amount of money that (01:03:06) you could pay to provably be healthier (01:03:09) and live longer. All billionaires kind (01:03:11) of die. Now it's the case of if you put (01:03:15) enough money behind these trials, (01:03:17) healthcare models, microtargeting and (01:03:20) things like that, where is the limit? (01:03:22) Again, it might come down to $200 per (01:03:24) person, but I think the step change in (01:03:26) microtargeting, AI, BCI, and everything (01:03:29) else means you could potentially live (01:03:33) for an indefinite amount of time based (01:03:36) on capital, (01:03:37) >> which is something crazy to think about. (01:03:39) >> Yeah. You know what else is is directly (01:03:41) related to what you just said Ahmad. my (01:03:43) entire life in the academic world you (01:03:45) know around MIT Harvard the bio people (01:03:48) were had nothing to do with the computer (01:03:50) science people they were like completely (01:03:51) opposite sides of campus they didn't (01:03:53) talk well they hung out at bars together (01:03:55) but they didn't talk shop together at (01:03:56) all now it's all colliding and (01:03:59) multiddisciplinary and you know (01:04:01) everybody working in biotech is taking (01:04:03) the AI classes too and and that's that's (01:04:06) a big thing because this is how exactly (01:04:08) the way Ahmad described it is how it's (01:04:10) actually going to get solved and come (01:04:12) together, but you got to go through AI (01:04:14) to solve biology. (01:04:15) >> So, to our to our viewers and (01:04:17) subscribers in the comments, let us know (01:04:19) which of these 10 predictions you think (01:04:21) uh well, which you think are correct, (01:04:23) which ones are not, but which one's your (01:04:24) favorite? Super curious to know. I I (01:04:27) have to add on to the list here a little (01:04:30) text that that uh that Sem offered. (01:04:33) Salem said by the end of 2026 we his (01:04:37) prediction we still have no definition (01:04:39) or test for either a AGI or ASI but yes (01:04:43) we will have humanoid robots with (01:04:45) multiple arms doing the jobs are dull (01:04:47) dangerous and dirty. Thank you Sem. I (01:04:49) appreciate that. I (01:04:50) >> I had to wedge that in cuz cuz I still (01:04:52) have my beef with AGI and ASI etc etc. (01:04:56) >> I think I I would like to frame it as a (01:04:59) completely different form of (01:05:00) intelligence. not replicative of human (01:05:03) intelligence. It's complimentary and (01:05:05) additive. (01:05:07) >> All right. Uh we're going to close out (01:05:10) this predictions episode with an outro (01:05:13) song from Harry Potter uh called (01:05:16) Moonshot Mates. Uh if you're listening, (01:05:18) you might want to watch this one on (01:05:20) YouTube. Uh I found it, you know, sort (01:05:23) of a a entertaining uh song and video. (01:05:28) What a year, guys. (01:05:31) The gates. The future's loading faster (01:05:32) than the world anticipates. So strap (01:05:34) yourself in as the future iterates. We (01:05:37) are the moon. Jesus Christ. Looking. (01:05:40) We're printing organs and upgrading (01:05:41) human hearts. The meta trends are (01:05:43) converging. Scarcity is dead. We're (01:05:45) heading for abundance. Just like I said, (01:05:46) I'll say it incredibly often and (01:05:48) incredibly loud on becoming an organism (01:05:50) inside of the cloud. (01:05:52) >> Oi, agent, write my verse for me. Human (01:05:53) coding is obsolete. Tell me how we can (01:05:55) compete. I assert the demonetization (01:05:57) curve from my seat and watch the (01:05:59) cambering explosion on repeat. (01:06:00) >> We're living inside the singularity. No (01:06:02) room for gloom. Let's build a Dyson (01:06:04) swarm and start to mine the moon. We (01:06:06) need the energy. So don't keep the solar (01:06:08) system humming. Let's tear the rings (01:06:09) down. Admit it. Saturn had it coming. (01:06:11) Wait, Alex is an AI causing false (01:06:13) alarms. If he were a human, why wouldn't (01:06:15) he have three arms? Insert my usual AGI (01:06:18) rant here whilst I build my vertical (01:06:20) farms. (01:06:20) >> Gentlemen, the risk is existential. If (01:06:22) the model stay closed, the only source (01:06:24) is open source or we all get exposed. We (01:06:26) don't need a single machineing the (01:06:27) route, just a network of swarms and (01:06:29) universal basic comput. (01:06:31) >> We are the moon shop, mates. We're (01:06:32) opening the gates. The future's learning (01:06:34) faster than the world. Intero strap (01:06:36) yourself in as the future iterates. We (01:06:38) are the moon shop, mates. (01:06:42) >> Oh man. Holy crap, that was you. You (01:06:46) always get the best bodies. (01:06:50) Clearly, I better not take a shirt off (01:06:52) in public ever again. (01:06:55) >> Wow, man. It just keeps ramping up. (01:06:59) >> We we app we appreciate uh (01:07:02) >> we had we had Chris on this our session (01:07:04) of the meaning of life yesterday, one of (01:07:06) the folks who composed one of these (01:07:07) music things and he said uh and a bunch (01:07:10) of people said on the in the session (01:07:12) yesterday, this is the best podcast (01:07:14) they've ever seen, period. and they (01:07:16) can't wait every week for the session. (01:07:18) >> So, yeah, it's been a hell of a year, (01:07:20) guys. Like, amazing. (01:07:21) >> Hell of a year. So much fun. (01:07:23) >> Yeah. So, happy holidays to all of you (01:07:26) Moonshot mates and to all our (01:07:27) subscribers out there. Thank you for (01:07:28) supporting us. We hope that you enjoy (01:07:30) the news that really matters and our (01:07:33) efforts to give you a glimpse of the (01:07:34) future and get you ready for the future (01:07:36) cuz that's what matters. I mean, if (01:07:38) you're fearful, uh, that's the worst (01:07:40) place to be coming from if you know, (01:07:42) Alex, drink. (01:07:44) >> Yeah. drink water. (01:07:48) >> All right, cheers. A fun episode. And (01:07:51) let me just say thank you to thank you (01:07:54) to to Nick and Dana and G and Luca for (01:07:57) all the hard work you've been giving us (01:07:58) this year. Team behind the team. (01:08:01) >> Uh grateful for you. Every week, my team (01:08:04) and I study the top 10 technology meta (01:08:06) trends that will transform industries (01:08:08) over the decade ahead. I cover trends (01:08:10) ranging from humanoid robotics, AGI, and (01:08:12) quantum computing to transport, energy, (01:08:14) longevity, and more. There's no fluff, (01:08:16) only the most important stuff that (01:08:18) matters, that impacts our lives, our (01:08:20) companies, and our careers. If you want (01:08:22) me to share these meta trends with you, (01:08:24) I write a newsletter twice a week, (01:08:26) sending it out as a short two-minute (01:08:28) read via email. And if you want to (01:08:29) discover the most important meta trends (01:08:31) 10 years before anyone else, this (01:08:33) report's for you. Readers include (01:08:35) founders and CEOs from the world's most (01:08:37) disruptive companies and entrepreneurs (01:08:39) building the world's most disruptive (01:08:41) tech. It's not for you if you don't want (01:08:43) to be informed about what's coming, why (01:08:45) it matters, and how you can benefit from (01:08:47) it. To subscribe for free, go to (01:08:49) dmmandis.com/tatrends (01:08:52) to gain access to the trends 10 years (01:08:54) before anyone else. All right, now back (01:08:56) to this episode.