This Week in Startups
00:44:43 · Oliver Cameron, Jason Calacanis
[00:44:43] S2: So, really, if you think of a world model like this, you've got this really attentive learner. You place it on a couch in front of a TV screen, and you show that really attentive learner many, many thousands of years of video. And it's just sat there. It never gets distracted. It's got this notebook, and it's writing notes on every single thing it sees. It's keeping notes on dynamics, on motion, on humans, and everything you could possibly think about. And then at the very end of that, this notebook is all filled out with all of the knowledge that it's learned through that process, and that that notebook is now massive. Right? It it describes so much about the world. And that really is how these models learn, that just simply taking copious notes of how physics evolves over time, how motion evolves, how humans work together, how we don't work together, and it just learns lots of concepts from from simply doing that. And so when it comes to world models, one of the hills that we believe is most important to climb is physics accuracy. Yeah. If we climb physics accuracy, that means the model is learning fluid dynamics. It's learning gravity. It's learning all these concepts. And so what we see now is they they are just from a very simple learning objective, which is predict during training time how the world's gonna evolve next. It learns very general, concepts. [00:46:03] S0: Amazing. Alright. Listen. Continued success on it. Where are you based, and, are you hiring, at the moment? I know you raised some money.
00:46:03 · Jason Calacanis, Oliver Cameron, Lon Harris
[00:46:03] S0: Amazing. Alright. Listen. Continued success on it. Where are you based, and, are you hiring, at the moment? I know you raised some money. [00:46:09] S2: Yeah? Yes. We've raised $370,000,000 from folks like Amazon, AMD, NVIDIA, GV. We're based in the Bay Area, London, and Zurich, and definitely Harvard. [00:46:17] S0: Love it. And, the domain name so people can go apply to work on this vision and and get in on, maybe some of these RSUs, get some equity in this company over the next four years because you're crush it. [00:46:27] S2: Thank you very much. Odyssey dot systems. [00:46:29] S0: Alright. Thank you, Oliver. If you find yourself in Texas, or Tokyo, you know, you just say my name three times, and we'll go get some barbecue, or I'll take you for some, soup less ramen. Yeah. Great job, Oliver. [00:46:42] S2: Thank you, sir. Thanks. [00:46:42] S0: Appreciate your time. [00:46:43] S1: Yeah. His his his system is gonna have to generate Austin for him to come here. So I've gotta fill [00:46:48] S0: it up. Absolutely. Yes. We're gonna have to do some fidelity of Terry Black's beef rib. Beef rib is right now dormant on a CPU. It's just in a cluster somewhere waiting to be popped up. Man, we haven't been to Terry Black's in a while, man. I gotta go get [00:47:01] S1: that beef. We gotta go. [00:47:03] S0: We need to do, like, a meeting. We gotta get back to the standing meeting at a barbecue joint. [00:47:07] S1: What are doing here? Yeah. The last time I was there was the birth of Still Core Capital. You and Mark Jeffrey and
00:47:07 · Lon Harris, Jason Calacanis
[00:47:07] S1: What are doing here? Yeah. The last time I was there was the birth of Still Core Capital. You and Mark Jeffrey and [00:47:12] S0: I Oh, was very Oh man, we hit that beef rib. I mean, it's like the beef rib is so expensive, but it's so good. It's so fatty and delicious. Can only eat like a third of one now at my GLP. Yeah. Youroe.co/twist. Get yourself a GLP. Yeah. Lose 40 pounds. I was at a Halloween party, man. People were like, JCal, you look great. Yeah. You look great. I was like, thank you. [00:47:35] S1: Think about roe.co. It's a two letter domain name, Jason. How did how did they... That that must have been... That must have cost a pretty penny. [00:47:42] S0: Pretty, pretty penny. Pretty penny. [00:47:46] S1: Worth it. Right? I mean, if you're if you're [00:47:48] S0: I need to sell these two domains, mahalo.com and inside.com. The auction on my own auction software didn't really work too well. It didn't hit the reserves. I'm gonna put it... I think I'm just gonna put it on GoDaddy. You ever heard of this GoDaddy? I've heard of GoDaddy. Yeah. GoDaddy is amazing. I do all my register and domain names there. They also have all kinds of services around it. Like, you wanna set up a website, wanna set up mail, you wanna redirect it like, just go in there, .com. That's right. And I just redirect it to somebody disgraceful as a bet on all in. Yeah. And, you know, it's just it's just sitting there in my GoDaddy, and I just redirect the beep beep boop, and it's done. Two secondes. Alright. Listen. Thank you to our friends at GoDaddy. Godaddy..com.
00:47:48 · Jason Calacanis, Lon Harris
[00:47:48] S0: I need to sell these two domains, mahalo.com and inside.com. The auction on my own auction software didn't really work too well. It didn't hit the reserves. I'm gonna put it... I think I'm just gonna put it on GoDaddy. You ever heard of this GoDaddy? I've heard of GoDaddy. Yeah. GoDaddy is amazing. I do all my register and domain names there. They also have all kinds of services around it. Like, you wanna set up a website, wanna set up mail, you wanna redirect it like, just go in there, .com. That's right. And I just redirect it to somebody disgraceful as a bet on all in. Yeah. And, you know, it's just it's just sitting there in my GoDaddy, and I just redirect the beep beep boop, and it's done. Two secondes. Alright. Listen. Thank you to our friends at GoDaddy. Godaddy..com. [00:48:33] S1: That's where to go. [00:48:34] S0: That's the best register. And you you know, you want something that's safe. They'll lock that they'll lock that ish down. Like, you would not believe. I got that locked down, like, Knox. I never lose anything. If my credit card goes bad, they call me. They email me. Hey. Your credit card is... Because I... You know, if you're there for twenty years, every four years, car goes dead. [00:48:52] S1: Yep. I get [00:48:54] S0: the certificates. So then you could put in a backup card. You know? That's a good feature too. You put a backup card in. Like, if it doesn't hit this one, you hit this one. You never wanna lose a [00:49:01] S1: Yeah. I don't wanna [00:49:02] S0: lose I just got an alert from the [00:49:03] S1: Very important.
00:49:03 · Lon Harris, Jason Calacanis
[00:49:03] S1: Very important. [00:49:04] S0: Yeah. I just got l o a a, Loa. Loa. When I was in my Hawaiian domain names. What does that mean [00:49:11] S1: in Hawaiian? Loa? [00:49:13] S0: Loa, l o a a, is to get, obtain, receive, or find. So to find something. L O A apostrophe Like L O I received the letter, I found the letter is I received. Found. Without the X-ray A has a different meaning. It means far. Like a long mountain, a far distance, secondly. So I was like, when I was doing mahalo, which was like, thank you and a heartfelt thank you, appreciation, I got Loa first to find, right? Like a Google competitor. [00:49:50] S1: Right. Was a search company, so that makes total sense. [00:49:53] S0: Yeah. Man, talk about being ahead of your time. Can you imagine everything we did is like the perplexity surge?
00:49:53 · Jason Calacanis, Lon Harris
[00:49:53] S0: Yeah. Man, talk about being ahead of your time. Can you imagine everything we did is like the perplexity surge? [00:50:01] S1: Data centers, Jason, they remain consistently unpopular, and opposition is strongest when people are asked about building them in their own communities. I pulled a few, little bits of data. The, ABC Washington Post Ipsos poll from October 2026 says 66% of Americans supposed data centers being built anywhere in The US, 30% support them, and support peaks among 53% among MAGA Republicans specifically, 80% opposition to data centers among people 30 years old and under. Now Axios reports that OpenAI is planning a PR push for after Tuesday's midterm elections. The idea on the table would be AI speed limits pushing The US to help write global AI standards. Selling AI is something that empowers people day to day, and here's the best one, renaming data centers as quote AI compute campuses. OpenAI, chief global affairs officer Chris Lehane told Axios that the midterms will be a bit of a movie trailer for 2028 when he expects AI to be the signature issue of the presidential election. Jason, my question's for you. Do you agree? Is AI the thing we'll be debating for the next two years into 2028? And two, is it too late for an AI data center rebrand? Can we call them compute campuses? Do you think that's gonna that's gonna fly?