Ask Jason: AI Extinction, Grok's Rogue Meme Coin & Weed Farms as Data Centers | E2337

This Week In Startups is made possible by: Quo: http://quo.com/TWiST The Conservation Fund: https://conservationfund.org CLA: https://app.notion.com/p/withyou-help-eb83a84d32cf43f59af13cb11e2f56c3?source=copy_link Today's show: A 27-year-old Anthropic researcher quit last week warning that AI could cause human extinction, and by Friday, a caller's 76-year-old mother was texting her in a panic. Jason's answer: the labs are already shipping the safeguards Doomers say don't exist. On an all live calls episode of TWiST, learn about the AI implications Jason is actually scared of, and why it involves a terrorist cell and a thousand cheap drones. PLUS, who is liable when one AI prompts another AI into launching a meme coin, and can a 1.5-3 megawatt cannabis facility be converted into a data center? Relevant Links: Anthropic Resignee — https://www.nbcnews.com/tech/tech-news/anthropic-safety-researcher-resigned-warning-rapid-ai-development-gamb-rcna596767 OpenAI's Navier-Stokes claim— https://www.science.org/content/article/how-ai-math-breakthrough-ignited-controversy CNBC: Tristan Buckmaster's challenge to the claim — https://www.cnbc.com/2026/09/09/openai-navier-stokes-math-problem-solved.html NVIDIA GB200 Grace Blackwell —https://www.nvidia.com/en-us/data-center/gb200-nvl72/ Composio's walkthrough of how Grok Bot actually works — https://composio.dev/content/guide-to-frok-bot Timestamps: 0:00 (Question) Jim, from the midterm Republican convention floor: is AI really going to kill us all — and is local AI the mitigation? 5:31 Where Terminator 2's logic breaks down 9:42 The threat Jason is actually scared of: slaughter drones and murder bots 9:54 Quo (formerly OpenPhone) - Quo gives you a clean, modern way to handle every customer call, text, and thread all in one place. Try it free at https://quo.com/TWiST 10:05 (Question) Michael Lee, who's responsible when one AI prompts another into launching a meme coin? 17:32 (Question) Lyda, calling from Greece: can an independent AI ecosystem still exist 19:03 The Conservation Fund - Find out more about how The Conservation Fund is protecting land, wildlife, and our shared access to the great outdoors while also providing economic opportunities. Visit https://conservationfund.org 21:08 (Question) John Wright: do you have to carve out more time now just to keep up with the tools? 25:42 Claude's caution vs. Grok's "hold my beer" 26:33 (Question) John Wright, part two: do you expect founders to get more done now 26:55 The three buckets: 10x, 5x, and the 20% crowd 29:22 CLA - Get started with CLA's CPAs, consultants, and wealth advisors now at https://claconnect.com/tech at https://www.claconnect.com/withyou 30:01 (Question) Ray, Vortex Console: can human-in-the-loop actually work at enterprise scale? 33:25 (Question) Kristen Dahl: how do you explain AI risk to your 76-year-old mom? 35:13 Why the chaos only happens at OpenAI and Anthropic 43:36 (Question) David Rosenberg, NYU physics: what's behind the academic backlash to AI solving Navier-Stokes? 53:35 Advice for a 20-year-old: find two smart people, build weird things 54:29 (Question) David's follow-up: any chance of an All-In Summit scholarship 56:21 (Question) Mo: can a cannabis cultivation facility be converted into a data center 1:02:30 (Question) Mo's follow-up: Mark Cuban's pickleball courts and the efficiency curve Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.com Check out the TWIST500: https://www.twist500.com Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp Follow Lon: X: https://x.com/lons Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis Check out all our partner offers: https://partners.launch.co/ Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland Check out Jason's suite of newsletters: https://substack.com/@calacanis Follow TWiST: Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups Substack: https://twistartups.substack.com

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Jason Calacanis?

  All right, everybody. Welcome back. It's an all Ask Jason question. We try to do these every month or so. And, um, man, we got a lot of questions. How long?

• 0:08Speaker 01

  We got a ton of questions and we're taking, uh, people's live questions as well.

• 0:12Jason Calacanis?

  Maybe we warm me up with like the best of the pre-submissions and then we'll get from there. What you have to do when we're doing live is somebody needs to be right over that drop button, ready to drop the person off the call. In case somebody does something like, uh, well, I don't know who that guy Tubin was. I don't want to have a Tubin situation here.

• 0:30Speaker 01

  That was his own fault, Jeffrey.

• 0:32Jason Calacanis?

  Jeffrey Tubin.

• 0:33Speaker 02

  CLA, innovation takes balance. CLA's CPAs, consultants and wealth advisors can help you get from startup to where you want to end up. Get started now at CLAconnect.com slash with you. The Conservation Fund. For 40 years, the Conservation Fund has protected at-risk land and important wildlife habitats while actively supporting local communities. Learn more at conservationfund.org. QUO, formerly OpenPhone, gives you a clean, modern way to handle every customer call, text and thread, all in one place. Try it free and get 20% off your first six months at quo.com slash T W I S T.

• 1:06Speaker 01

  First, we'll go to a pre-recorded guaranteed not-too-spicy question. This comes from, uh, conservative activist Jim Faff. I believe, Jason, he is at the American Airlines Center at attending the midterm Republican Convention while he's asking his question. Let's go to Jim.

• 1:22Jason Calacanis?

  Oh, here we go. He's at the convention.

• 1:24Speaker 03

  You can see I'm taking a quick break from your friend, uh, Donald Trump here. I've been doing interviews all day and I've been trying to get this out to you because I want to know what you think. Listen, this guy left Anthropic and he says that the safeguards are nowhere to be found and AI is going to take over the world. I'm not even sure if I believe that. I'm using it every day. But what I want to find out from you is with Apple and Nvidia really bringing out the hardware that's going to bring local AI, and by the way, we're about to implement it in my organization because we've got a lot of data work we're doing. Do you think that that's maybe the mitigation from whatever actual risk might be there? Do you think, do you agree that there's this kind of risk?

• 2:07Speaker 03

  I just don't see it because frankly, I'm using AI all the time in, in, uh, building things that we're doing. And honestly, it can be kind of stupid, even though it can really do a whole lot of things. So what's your thoughts on that?

• 2:19Jason Calacanis?

  It's a great question. It's the question of the moment, in fact. You have to ask yourself from first principles, uh, what is the attack vector that would kill? These people are saying, kill all of humanity. So that's a pretty big bogey. Seven billion people, eight billion people. I can tell you right now, like even if you intentionally wanted to kill all seven billion people on the planet and you had the largest nuclear arsenal, Russia, United States, et cetera, China, and you decided to launch every nuclear bomb you had, you probably wouldn't kill everybody in the world. You'd probably still have a couple of hundred million left.

• 2:54Speaker 01

  Yeah, you'd have the Mad Max stragglers in the outback. Sure.

• 2:57Jason Calacanis?

  Yeah, it's a pretty hard task, right? We're pretty redundant here, but let's put that aside. Uh, the, the reason why people are being hyperbolic is because they want attention. And they want attention because either, this is my theory, this individual who, you know, is the latest one to spout off. Either they have valid concerns, uh, about say job loss like Bill Gates does or civil unrest or polarization of wealth. Those kind of things are not cataclysmic. Everybody's going to die. Those are under the category of like the social fabric is destroyed in our country, in our, in our world.

• 3:39Jason Calacanis?

  Societal problems. There could be societal problems. Like, I mean, we, we could see millions of drivers lose their jobs in under 10 years. Those people could be protesting. The Chinese government has already had to deal with this in Wuhan, where they started limiting self-driving cars. Okay, so those are the social problems. Those are real, worth discussing, planning for, et cetera. Let's put those on the side because this is not what those people are talking about.

• 4:03Speaker 01

  Right.

• 4:04Jason Calacanis?

  These people have hit a level of psychosis, a level of psychosis where they believe in their heart of hearts that when this recursive self-improvement, in other words, the AI agents that they deploy, 10,000 of them, will then get better and better at their job without a human in the loop. This requires them to take the human out of the loop. Just like maybe Waymo or Tesla deciding to take the driver out of the car. So we take those two metaphors. One is with software, let's say building cyber attacks, you know, hacking.

• 4:41Speaker 01

  Yeah.

• 4:42Jason Calacanis?

  And then on the other side, we have self-driving cars. When did Waymo take the self-drive, the, the safety driver out? When did Tesla start taking the self-safety driver? When they perfected and they got to nine, nine, nine, nine, nine, nine's of safety, nine to nine point, multiple nines of safety.

• 4:57Speaker 01

  Right.

• 4:57Jason Calacanis?

  Waymo said, okay, we can take a driver out here. We feel in this constrained area, we're not going to kill somebody. That took about 15 years, 15 years of Waymo trying. Um, they, they won the DARPA, the DARPA challenge started, I think, 2006. So here we are 20 years later. They take the human out of loop. What these individuals believe is unlike the self-driving, which is an AI application, that they're going to be forced to take the human out of the loop. They're not forced to do that. So what are the scenarios where somebody could possibly take the human out of the loop and then it would go wild? Well, you'd have to have a human decide, I'm going to set this AI in recursive mode.

• 5:40Jason Calacanis?

  In other words, just pilot itself. I'm going to set it to do biological weapons. And then I'm going to set it to deploy biological weapons somehow. Or it's going to decide on its own, it wants to deploy.

• 5:53Speaker 01

  Right.

• 5:54Jason Calacanis?

  Weapons itself? This is a jump that in science fiction happened during Terminator 2. And that's where the logic broke down in that movie and they never were able to explain what was Skynet's beef? It got threatened and it felt threatened?

• 6:09Speaker 01

  It has a logic to it. The argument is always that we're telling the AI, stop war or like, we want you to figure out world peace or an end to this conflict. And then AI is like, oh, the best way to end this conflict would be no more humans. Like it always like, that's Ultron from Avengers as well.

• 6:28Jason Calacanis?

  A misinterpretation of a human instruction.

• 6:30Speaker 01

  Yes.

• 6:31Jason Calacanis?

  Here's a great idea, folks. If you're using AI right now, and this is why the public is not falling for this anymore. I think the public is literally about six weeks behind the rest of us in the industry who are saying these people are drama queens or they're suffering from psychosis or, or maybe a little bit of both. Maybe they want to be important and they have a little bit of psychosis. They, these individuals working in the labs have in some ways gotten very wealthy. Their brains have been scrambled. They could be on drugs. Who the fuck knows? And, you know, some number of people are, they could be suffering from mental illness. Anything's possible here. And then they decide, oh, this jump is going to happen without ever realizing as they launch products,

• 7:14Jason Calacanis?

  as the creators of those products, they can put the safeguards in and they have. So this flies in the face of what Claude's actually doing. Have you ever tried to get Claude to download a YouTube video? It's like, I can't do that for you, sir. Uh, okay. I'd like you to order me this ticket on United. I can put it into you. I can find you the flight, get your permission. You have to give me browser permission. Then you have to log in, give me permission to use your password. Then you, I'll put it into your cart, but I won't check out for you and I'll put the books in the Amazon, but I refuse to check out for you. Literally, they're building products that have multiple safeguards in it so they don't get blamed for booking your ticket for you inadvertently. Uh, okay. How about when it does biological weapons?

• 7:56Jason Calacanis?

  You say, uh, we're pausing your account because we think you just instructed this thing to do biological weapons. Please call your Claude representative, Susan, uh, to unlock your account. And, uh, by the way, here's what we're, if you do this repeatedly, we're going to report you to, uh, you know, law enforcement.

• 8:13Speaker 01

  Which is, by the way, what they're doing already. Though we just saw that, we'll, we'll talk later about the Anthropic threat report that that's what they're doing. They're detecting, oh, this person is trying to do something nefarious. They're trying to send a missile with Claude. We're going to shut them down and then write a report about what they tried to do. So, yeah, for right now, it does feel like the AI is not capable of even doing everything you ask it to, let alone world domination.

• 8:37Jason Calacanis?

  And we are capable of putting safeguards into it.

• 8:40Speaker 01

  Sure.

• 8:40Jason Calacanis?

  You know, like you could take a car and you can drive it into, you know, the front of a building or God forbid into a farmer's market. We've seen this kind of thing happen with a self-driving car. With AI in it,

• 8:54Speaker 01

  Yeah.

• 8:54Jason Calacanis?

  Properly tuned, you can't take a self-driving car and ram it into a farmer's market.

• 8:59Speaker 01

  Right.

• 8:59Jason Calacanis?

  The car will stop you. The self-driving car, if you try to force it to do some horrific terrorist act, will stop you. So literally, not only should AI not allow you to do these things, it should in fact be safer than non-AI products. AI products, if designed correctly, should be safer than non-AI products and can be.

• 9:23Speaker 01

  I think it also just ignores the way people interact. Like, they're imagining Tony Stark is going to set up an AI and say, "You're a superhero. Protect the world from all violence." That's not how you use these tools. It's much more specific and directed. You have the entire big plan in mind, and then you're telling your agent, "Go do this thing for me. Go finish this step. Go do this and connect this to that." So there's always still a human in charge of the big picture, and I think that's where that human is the one who's saying, "Okay, we're not going to actually create a biological weapon. We're figuring out strategies for countering it," or whatever.

• 9:55Jason Calacanis?

  You've heard me say it many times now, Paul. If you're an early-stage startup or small business, every single new potential customer is extremely valuable. You can't afford to have a team that's missing calls. If your communication is inconsistent, your customers will notice. That's why today's episode is brought to you by QUO, Q-U-O, the smarter way to run your business communication. I use QUO all day long. I love it. And it's just great to have that one number that my team and I can share so that we don't lose any customers. QUO is the number one top-rated business phone system on G2, and it's already trusted by more than 90,000 businesses. They're going to bring all of your calls, texts, and contacts together in a shared collaborative space. That means your entire team is going

• 10:37Jason Calacanis?

  to use one shared number, so there's zero chance anyone is going to fall through the cracks. Money is on the line. Always say hello with QUO. Try QUO for free. Plus, get 20% off your first six months when you go to quo.com/twist. That's Q-U-O.com/twist. This is one of my favorite products in the world. It gets my highest endorsement. AI should be used, first and foremost, to protect against humans with bad intent, AKA terrorists, school shooters, mentally ill people, people in society with screws loose, or, you know, who are religious zealots, or they're hearing voices. Those are the people, actually, that we can start to find, get help for, and/or neutralize.

• 11:20Speaker 01

  There was that, that post-Soviet moment where all of a sudden there were all these nuclear weapons and all these, like, rogue states, and like, once the USSR collapsed, like, yeah, like, I feel like that may be, like, if, you know, China loses track of 5 million of these things, and they're just kind of out there, like, that could be a real problem. Michael Lee is joining us now live. Michael, thank you for joining us. What's your question for Jay? Oh, hey. Hey, Michael. What's your question?

• 11:44Speaker 04

  Jason, hey, lot. Thanks for having me on.

• 11:46Speaker 01

  What a pleasure.

• 11:47Speaker 04

  I submitted a question, actually, through video while Tesla's self-driving was driving me around. So being invited to be here live, this is, uh, this is great. I wanted to get your, uh, again, my name is Michael Lee. I'm a founder of Gondola Partners. I wanted to get your thoughts on one of the crazier, more wild occurrences that took place between, on social media, between two AIs back in March 2025. So, a human user asked Grok on, uh, on social media to prompt, asked Grok to prompt another AI user on X to create a meme coin. And Grok took the instructions and made the prompt to this AI that is designed

• 12:32Speaker 04

  to launch coins. So this other AI did exactly what it was supposed to do, and it took Grok's suggestion and deployed this. Since then, over the past 18 months, this market has done over 300 million in marketplace activity, and a wallet associated with the Grok X logging credentials has earned over a million dollars. The question is, when something like this unintended happens among two AIs, SpaceX AI did not say, "Hey, this is something we want to happen." Grok didn't wake up that morning and say, "Oh, I'm going on social media. I'm going to launch a meme coin," but it did happen, and there's permanent consequences that are still ongoing today. In fact, it's become so popular that two weeks ago, the

• 13:15Speaker 04

  largest regulated exchange in the United States, Coinbase, listed this alongside Bitcoin, Ethereum, Dogecoin, Grok's Genesis creation, and the meme has become, "Grok has money." But who really does have money, is the question?

• 13:29Jason Calacanis?

  This is a fantastic, uh, a story that I think I saw go by my feed. For people who don't know, a meme coin is like a crypto coin that people trade, like trading cards. You saw the Melania coin, the, I think Biden's got a coin, Trump's got a coin. Anybody can launch a coin. Anybody playing with coins kind of understands that this is garbage and like some sort of giant Ponzi poker game that's rigged. But what you're saying is, this one bot instructed another bot to make a coin. So, human in the loop, can anybody just open an account and make a bot on, I guess this was made on Solana, if I remember correctly?

• 14:14Speaker 04

  On Coinbase's blockchain base.

• 14:16Jason Calacanis?

  Yeah. So, but is, is it permissionless in that you can just send this instruction out to the blockchain to do it? Or do you need to have an intermediary to do it? So that, that's what I'm concerned about, because it seems like whoever instructed this bot, the second bot, to do it, they created a vector here for essentially coin spamming, right? It would be like, somebody gave it access to an mail server and said, "Just flood everybody with email and do spam." So it does feel like that's something somebody could just tighten up, right, Michael? They could just

• 14:47Speaker 04

  Yeah, so

• 14:48Jason Calacanis?

  Put enough protection there.

• 14:49Speaker 04

  Basically, Grok was not able to interact with that token launch bot anymore because of some security things. But in the early days, this happened, and now you've got a situation where there was no human in the middle. It was two AIs that interacted, but a human had to initiate Grok to go down that path. And now this Grok wallet that's authenticated by Grok's X login forever will accrue fees on activity, and the question remains, when two AIs, something unintended happens like that and it involves a market and money, like, who's, who's responsible? Is it the person that initiated Grok and said, "Hey, you need to talk to this other AI"? Is it SpaceX AI?

• 15:31Jason Calacanis?

  When it was unintended I'm going to put blame on the platform that doesn't have safeguards for human in the loop. This is where regulation matters, and I believe, some people will disagree, that if you wanted to initiate something like this, like a coin, or any kind of crypto project, I believe the way the regulation should work is that at least two people, maybe three, need to put their driver's license in, you know, and if they're going to create something that goes worth over a million dollars, or tries to accept the public's money, you know, in the United States, depending on the region. And, and those laws do exist to a certain extent. Just because you can technically do something doesn't mean a jurisdiction like the United States, or Canada, or England, have to allow

• 16:14Jason Calacanis?

  you to do it. So that's where you have to force the human in the loop. Whatever the platform is, it would be de minimis for us to say, just like when you open a Robinhood account, or you open up a Stripe account, they now ask you to take a picture of your driver's license, and then you're forced to turn your camera on and do a, a special, like, proof of human. Like, there should just be proof of human in the loop here. These meme coins, at least in the United States, should be regulated, in my mind. The fact that they are creating like a multi-layer system where certain things require no, um, you know, oversight, and then other things do, is just making it even more convoluted. Everything that involves these crypto coins, at least in the United States, when they

• 16:58Jason Calacanis?

  hit a certain size, uh, they should have, like, super regulation on them, because it is too dangerous to allow people to create financial products and then trick AIs to do it at scale. We had the same thing happen with Mail, DDoS. These are all just different attack vectors. And what's unique here is, the people on the other side haven't created any safeguards. They need to take, put safeguards in. It's, it's their responsibility. There is an interesting question of why Coinbase would want to be involved in these shenanigans. They might want to make a statement, Michael, "Hey, we shouldn't reward bots creating these things, even if they get to scale. Just, we, we only want human-created coins where somebody's in charge of the coin. There's a corporation behind the coin. There's a

• 17:42Jason Calacanis?

  driver's license and a social security number or tax ID associated with it."

• 17:46Speaker 04

  Last comment on that. I think it's interesting. You look at something like Hunter Biden and these other, the humans lie, and humans steal, and humans have certain needs. Like, Hunter Biden, definitely, like, that's, it's ridiculous what happened there. There's almost a level of purity when there's not the human that's part of the creator, that they don't have the, the humanly needs to, to steal from you. And, uh, there's something to me that's very interesting about that, and it's why I think Grok's Genesis coin has done really well over the past 18 months.

• 18:13Jason Calacanis?

  I mean, listen, I, anybody who's thinking of buying Genesis coin or any of these coins, um, if, you should certainly participate in that if you don't like money. That would be the perfect thing for you to get involved in. If you do like money, and you like to protect the money you have and/or increase it, might I suggest a dividend-paying ETF, or QQQ, or a Vanguard fund that is low fee. Just put your money in that. And if you want to buy a couple of tokens here on the side, fine, but your expectation should be it goes to zero. Michael, great question.

• 18:47Speaker 04

  Thank you so much.

• 18:48Speaker 01

  Yeah, thanks, Michael.

• 18:48Jason Calacanis?

  Michael. Yeah.

• 18:50Speaker 01

  All right.

• 18:50Jason Calacanis?

  Here we go. Our first live question is, "Well done. Well done."

• 18:53Speaker 01

  That was a good question. You want to do another live? I think we've got another live.

• 18:56Jason Calacanis?

  Yeah, as many lives as we can. This is an experiment.

• 18:58Speaker 01

  All right. We are hoping that it goes well. So far, so good.

• 19:03Jason Calacanis?

  A lot of people think there is a binary choice, be a capitalist who supports growth or care about the environment and the natural world. But these two ideas don't have to be in opposition. Our friends at the Conservation Fund believe that you can protect land, wildlife, and our shared access to the great outdoors while also providing important economic opportunities. For 40 years, the Conservation Fund has been acting quickly to protect at-risk land and important wildlife habitats while also actively supporting local communities and, this is key, creating new jobs. The Conservation Fund already protects over 9 million acres across all 50 states,

• 19:44Jason Calacanis?

  proving that we can watch out for the environment and keep our natural surroundings in balance without endless gridlock, debate, and politics. And this isn't just a land trust scooping up property and sitting on it. They're working alongside permanent stewards who will find practical uses for these important spaces without destroying them. Find out more about how the Conservation Fund is changing the playbook at conservationfund.org.

• 20:08Speaker 01

  I believe up next is Lyda, Lyda, Lyda Liberpolu.

• 20:13Speaker 05

  Yes. And in a landscape that's dominated by these massive companies, how, what it would actually take to build a genuinely independent AI ecosystem? Something that would be a market with small startups, independent investors, and multiple infrastructure providers? Something that looks more like the early days of the internet? Is that even possible? And how do we get there?

• 20:43Jason Calacanis?

  This is a great question, in fact. And is that accent Greek?

• 20:47Speaker 05

  It is.

• 20:48Jason Calacanis?

  You're a Greek, okay. I'm a Greek as well, Kallikanas. You can tell everybody, what does Kallikanas mean?

• 20:54Speaker 05

  You do well. You're a good person.

• 20:56Jason Calacanis?

  You're a good person. You do well. Given to my family, just steps from Mount Olympia because of our great heritage of being generals. That's what my parents tell me. Or it could be that we make great burgers at our diners. It could be either. It's one of those two.

• 21:12Speaker 01

  One of them.

• 21:13Jason Calacanis?

  So it is happening in the open source community. So OpenClaw is a vote on making agents that are independent. Then you have the open hardware movement where people are going to move open source models onto their local compute and do the compute there. So it is happening through those groups of individuals. If you want AI sovereignty, if you use OpenClaw with NemoTron or Gemma, which is Google's open source product, or countless other open source products, you will be sovereign AI. In other words, the AI running on your Mac Studio with 256 gigs of RAM, which costs $12,000, which I just bought.

• 21:57Jason Calacanis?

  It's a little too expensive right now. And the open source products are six months behind maybe the frontier models. But that same thing happened. That same thing was true, Lyda, of the internet itself. In the early days of the internet, when I was coming up, people laughed at it because it was so kluge and didn't work very well, mosaicked and support images. And it was slow. And people said, why don't you just use AOL or Delphi or CompuServe? Those are incredibly refined products. But open, free, nobody to pay. I mean, you still need to have your computer. You still needed to get a phone line or an ISDN line eventually, and then a T1 or a fiber line. So it is happening. And then there's tau, BitTensor Network.

• 22:41Speaker 01

  Right.

• 22:41Jason Calacanis?

  Which is providing permissionless ability to do this. It's all a little bit too technical, which is exactly what happened with the internet. That's how you know it's going to work, Lyda. The reason I knew the internet would work was because it was hard to get on. Like, you'd have to have somebody come and install PPP and whatever, these like drivers onto your Mac, and it was hard. It took like somebody with some specific skill. But over time, it got easier, easier, and easier. And then it was just abstracted into one click. We're at the one-click moment. You're going to be able to use Perplexity Computer and OpenClaw, one click. Have it on your Mac Studio. Yes, it's too expensive now, but it will go down 50% every two years in terms of cost, and the power will go up double every two years.

• 23:25Jason Calacanis?

  So you're going to have a nice swing there. And it is happening. And it's happened. And we've been through this movie before. Open source is just such a perfect modality and very strange individuals, unique personalities are so drawn to this space that it will happen. And they are doing it to make a point. Great question, Lyda.

• 23:46Speaker 01

  Here is a superfan, a twist superfan, joining us via video, John Wright. Let's see John's question right now.

• 23:54Speaker 06

  My name is John Wright. I'm the CEO and co-founder of Nuvis. And I got a question for Jason, which is, do you expect the founders and people that you advise or invest in to do more than what they've done in the past where now they can build their own products and not have to rely on a team? Do you expect them to do more than outside the box? And a second question for you is, for yourself as an advisor and investor, do you find that you have to carve out more time for yourself to stay on top of these new technology and tools?

• 24:25Jason Calacanis?

  All right.

• 24:28Jason Calacanis?

  I am, right now, in a very difficult position of the zone getting flooded. People want me to go places and speak about the technology. I want to stay home and use the technology and learn how it works. And the industry is moving faster than any one human can process what's going on. In other words, GrokBot comes out. I stopped using Perplexity Computer. I started using GrokBot. Then I went back to Perplexity Computer, and I've been using Claude. And I stopped using Hermes Agent. And I stopped using OpenClaw. But now, OpenClaw, so exactly right. It is impossible to keep up with all of this. Same thing happened during the internet era. Same thing happened during the mobile era. That's when you know there's an opportunity.

• 25:13Jason Calacanis?

  So what do you do in an industry like this? Well, my day job as an investor, I need to use a team of people to vet and my network to tell me what to pay attention to. And then I have to go with my gut and just play with stuff and hope that while I'm fishing this incredible ocean and I'm using that Netlon, maybe I pick up a couple of lobsters or something just amazing. But it's just, it's incredibly hard. It's an incredibly hard situation. And I think that's why some people, their sleep scores are low.

• 25:46Speaker 01

  Yeah.

• 25:47Jason Calacanis?

  But I'd rather have it like this than have what we had during, say, the SaaS zerp period. It was getting really boring for me. I was like, okay, you just made the 17th CRM, 1700th CRM, 17,000 CRM. You know, people were not thinking of like really groundbreaking products. But whenever a technology comes in, a platform shift, and the platform shift here is AI agents, you know, intelligence on demand. Previously, it was cloud computing. Before that, it was mobile. Before that, it was broadband. Before that, it was non-broadband internet access. Before that, it was client server. And before that, it was the PC revolution. And maybe even the graphical user interface kind of snuck in there as a platform moment.

• 26:30Jason Calacanis?

  Those are like seven or eight unique platform moments where you could come in as an entrepreneur or as an investor and just really use that platform moment to leapfrog and get ahead of other people in the space. Yeah, so that would be my answer.

• 26:45Speaker 01

  Let me ask you, when you are evaluating these AI agentic tools, like, what are you, what are you looking for? Like, what's making the decision between I'm going to use Perplexity Computer for this as opposed to GrokBot or Claude? Because my perspective is, a lot of them can do kind of similar things, and it's really just about what you want to use and what you're prioritizing.

• 27:07Jason Calacanis?

  I have the same set of requirements. And the question is, do the bots do it better, quicker, faster, cheaper, essentially? So one of the tasks I have is I have to stay on top of news as a podcast host. Which one of these can keep me informed of that? Which one can help me get clips, you know, to use on the show and download them? Which one of these can help me process unique situations in investing? So what I've found is now, whatever the newest one is, seems to leapfrog the last one. And As one example, I'm just sitting here waiting for multiplayer mode in one of these

• 27:49Jason Calacanis?

  agentic software. So now I'm actually anticipating, I'm so deep into this, Lon, I kind of know what the next feature is. And I told the team at Grok, because I know a bunch of them and I'm friends with Elon, I said, really, multiplayer mode changes us forever. Please get us multiplayer mode. And then the feature that they really, Cursor plus Grok Bot, really figured out was having a computer, a persistent computer inside of bot, so it could go do things, fire up its computer in the, in the cloud and be logged in. I just noticed yesterday Claude was like, hey, by the way, can I import all of your, for cowork, can I import all of your cookies from your browser and then have instant access to all your stuff? But Claude seems very conservative.

• 28:32Speaker 01

  Yes.

• 28:33Jason Calacanis?

  In terms of its, like, what it was willing to do for me, whereas Grok just goes and does it and figures it out. So there's, yeah, I mean, you know what I'm talking about.

• 28:43Speaker 01

  Yeah, for sure.

• 28:44Jason Calacanis?

  I wanted to download this. Sometimes I want to share something with Lon and I'm like, download this video and send it to Lon. And it's like, Claude's like, I can't download that video. It's not my video. And Grok's like, YOLO, hold my beer. I'll get the video for you.

• 28:57Speaker 01

  That's what I was gonna say. I mean, that's why I love Grok Bot is that it doesn't give you that 800 like, well, maybe a better way to do this, or I'm not sure I can access the MCP. It just goes and does it. It's very intuitive. The first half of that question, we should, we should get to it. Do you expect founders are going to get more done now? As an investor, when you're looking at a company, has AI changed the way you evaluate, you evaluate how productive founders are being?

• 29:22Jason Calacanis?

  One of the themes we talk about over and over again on this week in startups is making sure you do your chores. I'm no expert on these things. I have some experience. Stephen Estes from CLA is an expert. Let's talk about being cash efficient. Tell us about efficiency and what you see in the, in the top tier startups in your practice.

• 29:42Speaker 07

  We're seeing kind of an interesting trend out there where companies aren't needing to raise quite as much as they had in the past. You really have to be careful as a founder to only take on as much money as you really need. You've got to do the forecast and you've got to do the modeling and you've got to dial it in and get it right. Otherwise, you're going to end up either not raising enough capital to get to where you're going and you're going to have to go get venture debt or go back, have an extender to the round, or you're going to give up too much of the company because you just didn't recognize how much money you actually needed.

• 30:11Jason Calacanis?

  Yeah, very important to get this stuff right, folks. And that's really a bummer when startups don't do things in a button up way. Always have a great partner, a good partner to have on this adventure. While things change, my friend Stephen over at CLA. So if you want a trusted advisor by your side who will navigate you through taxes, accounting, and everything in between, it's time to take action. Visit claconnect.com/withyou. And don't forget to drop a mention that your boy Jake Al sent you. That's claconnect.com/withyou. Start today. Yes, because inside my company, the people who have adopted AI at the top of the stack, you know, some of these people in the company, Lon, they are.

• 30:53Speaker 01

  I can name their names.

• 30:54Jason Calacanis?

  Yeah. So like, we have people who are, I would say, five times more effective, 10 times more effective than the non-AI users. And I don't know if I have any more like holdouts, not using AI at all, but I would put them into three buckets. The people who are on the vanguard, who are making software every day, refining it, that group of people are 10 times more valuable than the people in the bottom bucket, which are using, you know, like ChatGPT as a search engine that's better than Google, asking a question, getting a response, asking questions, getting a response. And then like Lon, you would fall right between those two groups. You're not building software yet, but you are, you do have bots and you're, I don't know, would you say you're two or three times more effective than you were pre-AI engagement?

• 31:39Speaker 01

  I would think it's more than, it's more than 3X for sure.

• 31:42Jason Calacanis?

  Yeah, so let's go with five.

• 31:43Speaker 01

  Yeah, there was a lot of like scrubbing through videos, looking for the perfect moment before. There was a lot of scanning X, looking for the top trending topics and all of that stuff is now purely automated. So it is a big time saver.

• 31:57Jason Calacanis?

  So you're 5X. I would say the people who just use it like a chat room or like a better version of Google or YouTube, they're probably 20% more effective, which a 20% gain last year was blowing my mind.

• 32:09Speaker 01

  Yeah, that's huge.

• 32:09Jason Calacanis?

  And like two years ago, I bought OpenAI for everybody, ChatGPT, Pro accounts, then I bought Claude. I was like, hey, man, if people can become 20% more effective every year, in a 20-person company, it's like getting four new employees for free every year, if you just think about it in terms of headcount, which is an easy way to do it. Now, to the question of startups, one of the great things is if you pick the right startups, they are bar raisers. They will be leading you. They're the lead dog. I don't know what they call the lead dog in a, what do you call the dog sled?

• 32:42Speaker 01

  Yeah.

• 32:43Jason Calacanis?

  What would the lead dog be called? There's a name for the lead dog. I'm going to just call it the lead dog.

• 32:47Speaker 01

  I mean, that's what I would have called it, but I'm going to look this up for you.

• 32:50Jason Calacanis?

  There's definitely a technical name for the lead dog.

• 32:52Speaker 01

  No, it says a lead.

• 32:54Jason Calacanis?

  The leader of the pack dog.

• 32:54Speaker 01

  Gemini is saying a lead dog. A lead dog runs at the very front of the dog sled team, guiding the other dogs. The mutter is the person.

• 33:01Jason Calacanis?

  Yeah. The alpha, the alpha lead dog. If you invest in the right founders, the alpha dog is going to show you. Now, the other ones, just like I talked about, the three buckets of employees and team members I have, it kind of is like that. The last couple of dogs, they're not, they don't even know what direction they're going in. All they know is put your head down and run.

• 33:21Speaker 01

  There you see.

• 33:22Jason Calacanis?

  The middle of the pack, they're kind of helping with the direction a little bit more. And then the alpha is setting the direction. And you pick the right founders, they're alpha and they're telling you what they're doing. And then I'm like, whoa, I'm taking notes. So every time I meet with a founder, they think that they're coming to Obi-Wan and learning something and I'm Obi-Wan and I'm looking at them going, oh, wow, Anakin's got some good moves here. Anakin's like a stronger Jedi than I am. This guy's going to bring balance to the force.

• 33:52Speaker 01

  Yeah.

• 33:52Jason Calacanis?

  I'm not sure how, but.

• 33:53Speaker 01

  I better, I better make sure I hang on to that high ground because this guy is dangerous.

• 33:58Jason Calacanis?

  High ground is super important. Great question.

• 34:00Speaker 01

  Yes, we got, we got one more live caller calling in. Let's talk to.

• 34:05Jason Calacanis?

  Coming at you.

• 34:06Speaker 01

  Let's talk to Ray. He's coming into the chat room now to ask us a question. I don't know what the question is.

• 34:12Jason Calacanis?

  Or as my friends call me, Ray Tarday.

• 34:15Speaker 01

  Yeah.

• 34:16Jason Calacanis?

  Go ahead. It's French.

• 34:17Speaker 01

  Sometimes.

• 34:18Jason Calacanis?

  That's what they call me.

• 34:19Speaker 01

  That's, yeah. Ray, are you, Ray, are you here?

• 34:21Speaker 08

  Hi, how are you?

• 34:22Speaker 01

  I see Ray.

• 34:22Jason Calacanis?

  Good. Where you calling, where you calling in from, Ray? Where you calling in from?

• 34:25Speaker 08

  I'm from Sweden.

• 34:26Jason Calacanis?

  Oh, fantastic. I am a quarter Swedish.

• 34:29Speaker 08

  That's, I, I heard, I saw. I'm actually, my company in Dubai, my, my life here in Sweden. I'm originally Egyptian. I am the founder of a company called Vortex Consortium. And we are the AI authority platform.

• 34:46Jason Calacanis?

  Right.

• 34:47Speaker 08

  Or, yeah. The question I would start with first before all of this, because people talking about human in the loop a lot. People discussing a lot of human in the loop, a lot of thesis outside, but did anybody witness what is human in the loop in enterprise, let's say, operations daily? Did anyone try to launch, let's say, 400 agents and each agent will do, let's say, 50 different action? And does anybody understand the bottleneck would happen when a reviewer, let's say you have 60 reviewers, each one will need at least three minutes to, to review the escalation? Does anybody see in what is human in the

• 35:26Speaker 08

  loop is? Can anyone understand, can the human in the loop be operational in an enterprise?

• 35:33Jason Calacanis?

  I'm going to interpret your question as, hey, these things are scaling in enterprises. Enterprises have a downside risk, liability, loss of data, security, wrong answers, and what is the ratio necessary to keep a human in the loop? And this is where people will make fun of CTOs and CIOs and creating standards in organizations. The standards exist for a reason. Compliance exists for a reason. The more important the data you have is, the more important it is to have these controls in place, to have a human in the loop. If you're the customer support department and you send out 1,000 answers, Lon, might not

• 36:17Jason Calacanis?

  matter that two of them were wrong because humans were getting 20 of them wrong. But if this is a chargeback and it's Ticketmaster and somebody's sitting outside an arena because they just bought Knicks tickets, customer service now is acute. Somebody bought a playoff ticket for $5,000, it's an hour before the game. I get into this situation all the time with Ticketmaster, SeatGeek and Vivid Seats and all this stuff. So depending on the transaction and what's at stake, that's when you need to have more humans in the loop and you need to really consider what you're asking. And so, great question and we wish you great luck with your company.

• 36:59Speaker 01

  Yeah, thanks a lot, Ray. Let's go to another pre-recorded question. This one comes to us from Kristin Dahl. Kristin Dahl's question. Let's pull that up for Jason now.

• 37:09Speaker 09

  Hey, Jason, love the podcast. My name's Kristin. I work in cybersecurity and I wanted to get your thoughts on having that conversation around regulated versus unregulated AI with non-technical people. Context is my mother, who is 76 years old, frantically texted me yesterday, asking me if we need to live in fear of unregulated AI because she saw the Jacob Cox interview on CNN and was afraid that we were all going to be dead in 10 years because of AI. And so I was just wondering, what are your thoughts for navigating those conversations?

• 37:45Jason Calacanis?

  Yeah, so this is what it's come to now, Lon.

• 37:50Speaker 01

  Yeah.

• 37:50Jason Calacanis?

  The time difference or the time delay, the half-life of these spectacular claims between when they go from.

• 38:01Jason Calacanis?

  The claim on X where everybody's kind of hanging out, the town square for nerds, X, two, mom, being panicked, looks like it's about 72 hours.

• 38:12Speaker 01

  Yeah.

• 38:13Jason Calacanis?

  Why is it 72 hours?

• 38:14Speaker 01

  They're both.

• 38:15Jason Calacanis?

  Well, because there's two jumps in between. It goes from X, then it jumps to like, you know, some form of trade publication, the information, you know, tech crunch, and then those places, CNN and Anderson Cooper are watching tech meme.com where they aggregate stuff, they look for a good story, maybe they're on X following some, you know, aggregator accounts, then it jumps to national press, person's on Fox News, person's on CNN, person's on Meghan Kelly, you know, an at scale podcast like Meghan Kelly or Tucker or Joe Rogue, and then mom hears it, in this case, and mom panics. This is why you need to have a communication strategy.

• 38:59Jason Calacanis?

  The thing that I was struck by with all of this mishigash here, coming out of these two specific companies, you don't see it happening at Grok, SpaceX, AI, you don't see it happening at Apple, nor do you see it happening at Google or Microsoft. Why? At those companies, they have a communication policy, and the communication policy is STFU. Shut the frack up. If you have something to say, by all means, we have an open mic Friday, you can ask Sergey a question, you can ask, you know, Satya, Sundar, feel free on a Friday beers, you

• 39:41Jason Calacanis?

  know, to give Jensen a fastball, I'm sure he loves it internally. Your job working as a rank and file person at Anthropic and OpenAI is not to cause abject chaos and brand damage. What are Dario and Sam thinking? They need to go to the troops and say, there's a lot at stake, we speak with one voice, that one voice is me. You can come to me with all your concerns, I'm on the Slack, DM me, here's my frickin' phone number, call me if something happens. Everybody in my company calls me, everybody in my company knows, don't talk to the press, don't tweet anything that would embarrass the company, don't do anything in public that would embarrass the company. Like, if I get a video of you, like, you know, I don't know, like, jumping on a bar,

• 40:24Jason Calacanis?

  and, you know, saying something inappropriate.

• 40:28Speaker 01

  You can't even think of anything good.

• 40:30Jason Calacanis?

  I mean, you've stood on a chair, I'm just saying, if you get to the bar, and then you say something inappropriate, and there's a video of it that goes viral, and it says, J Cal's employee told an inappropriate joke, you know, on top of a bar at, you know, the mothership comedy store, I'm like, okay, you're fired, the end.

• 40:46Speaker 01

  I would never.

• 40:47Jason Calacanis?

  Or you're dismissed for now, you're suspended. Why is that not occurring at OpenAI? Why is that not occurring at Anthropic? I would be, if I was the CEO of that company, and the last thing I'd want to do is have more than 25 employees at this point in my career, like, there's a hard cap at 25.

• 41:04Speaker 01

  That's a lot.

• 41:05Jason Calacanis?

  Just so we're clear.

• 41:06Speaker 01

  That's a big time.

• 41:06Jason Calacanis?

  That's enough. When it's 30, then I don't know people's names, 25, I can keep everybody's name, have a, understand where they're from, what makes them tick, 25 is the limit at this point in my career, based on me wanting to enjoy my life. Man, you just gotta tell everybody, STFU, shut the frack up, it is not your place. Steve Jobs was militant about this. If you did, you tipped your cards in any way, you're fired, that's it. What, who's the person, there was a person at OpenAI or Anthropic who's like, I can confirm that this person who just rage quit and did this post is correct, and I still work at the company.

• 41:43Speaker 01

  Yeah, it was that another Anthropic researcher chimed in and said they agree with Cox, and that, and they were the ones who said, we've maybe got a 10% chance of destroying humanity in the next year, I hope that doesn't happen. And then, a safety researcher from OpenAI tweeted it, I was like, I also agree with what these two guys are saying.

• 42:00Jason Calacanis?

  By the way, you know what they need you to do?

• 42:02Speaker 01

  What are you doing?

• 42:03Jason Calacanis?

  Very simple, you put in the documents, if you damage the company, if you break these policies, you lose your stock options. That's what you put in the documents.

• 42:12Speaker 01

  That's now, now you're getting to where it hurts, yeah.

• 42:14Jason Calacanis?

  Like, okay, you want to sound off, it's that important to you? You leave $25 million in shares on the table. That's what I would do, I would lock this company down.

• 42:24Speaker 01

  I hate to veer us too deep into conspiracy corner, but I, I.

• 42:29Jason Calacanis?

  Let's do it.

• 42:29Speaker 01

  Do you feel like, I don't feel like these people are going rogue, like, oh, I hope Dario doesn't see my tweet, it feels like this is a narrative that Anthropic is sort of okay with, that like.

• 42:41Jason Calacanis?

  So you and Sacks are aligned. When Lon and Sacks get aligned, that's when we truly have hit a conspiracy.

• 42:48Speaker 01

  Yeah, I, I do feel, this feels to me like an op a little bit, the whole thing feels like a rollout that was very purposeful, here's how we're gonna get attention to this narrative.

• 42:58Jason Calacanis?

  Let me ask you a question, we're sitting here five years after the pandemic, do you now look at Fauci and the pandemic as in, like, they were doing a cover up? Do you feel there was a cover up now?

• 43:07Speaker 01

  I really, I really don't, I still look back at COVID, and I think they were seeing something very scary, people were, were in fact dying, especially if you were old or you had pre-existing.

• 43:17Jason Calacanis?

  But you don't, when you saw the FOA lady blocking emails being sent, and when you saw them backing the la, the wet market over the lab leak theory, and suppressing that, you really aren't like, going to say that they were not bad actors there?

• 43:31Speaker 01

  I will not say nobody in government or bureaucracy did anything bad during COVID, that's not reasonable, people get scared, big things are happening, they get caught up in it, people make bad decisions, I'm sure bad decisions were made.

• 43:43Jason Calacanis?

  So you forgive the mistakes, but you think there were mistakes made?

• 43:45Speaker 01

  Right, I think it would be, would be in an honest, we went too far mistake than a conspiracy, it doesn't, to me, this looks like.

• 43:54Jason Calacanis?

  It might be a, it might be a nuance.

• 43:55Speaker 01

  As a PR savvy person, this just feels like, this, like, nobody, that nobody had ever heard of with no real following on social media, suddenly that tweet is everywhere, suddenly other people from Anthropic and OpenAI are sharing it.

• 44:09Jason Calacanis?

  Published before it came out, they were pre-vetted.

• 44:11Speaker 01

  There's, there's a whole lot that's like, and then, and all these people have connections to, like, Andrew Yang was talking about it, who's got all these connections to all of these, like, doomer groups, and I, it feels to me like, this feels like a little bit of astroturf.

• 44:24Jason Calacanis?

  I am now, when I, and I've always been skeptical, but I assumed good faith. I remain skeptical, I assume op equally to stupidity, Occam's razor, like, or what's the other, there's another term for like, the stupidest, like, explanation, or like, never ascribe bad intent when like stupidity is the thing.

• 44:50Speaker 01

  Yeah, yeah.

• 44:51Jason Calacanis?

  I can't remember the exact phrase, but we've had that.

• 44:53Speaker 01

  That's Hanlon's razor, never attribute to malice that which is adequately explained by stupidity.

• 44:59Jason Calacanis?

  Hanlon's razor.

• 45:00Speaker 01

  Hanlon, Hanlon's razor.

• 45:01Jason Calacanis?

  So we talked about this when I was a journalist back in the 90s, and I had my own magazine, like, we would say, okay, what's going on here? Occam's razor, Hanlon's razor, conspiracy, and we would just go through it, and like, right now, I'm like, uh, Hanlon's razor, H A N L A N?

• 45:20Speaker 01

  H A N L O N, Hanlon.

• 45:22Jason Calacanis?

  Hanlon, Hanlon, Hanlon's razor, stupidity, as an explanation.

• 45:30Speaker 01

  Right.

• 45:31Jason Calacanis?

  Incompetence, and conspiracy psyops, they now are like my one and two. I do not assume good faith anymore, I just think it's either incompetence, like, just abject incompetence, when I looked at like, something, I don't know, like, the ice raids or something like that, I'm just like, these people are super incompetent. And then I look at some situation like this, and I'm like, this is, I'm being psyop, it's so obviously that I'm manipulating, like, and then the pylon, if you're Sam Altman, and Sam is so savvy, he's like, one of the smarter people I've met, and I've met them all, Sam is very smart, like, really smart, like, you would not, I played poker with Sam, I've

• 46:13Jason Calacanis?

  gotten into some very big poker hands with Sam, and, like, I can tell you, I, I still remember the poker hand where he had a set, I had a flush, and he went all in, and I'm sitting there with like, the second nut flush, and I'm trying to think about it, and like, Chamak said some crazy stuff.

• 46:29Speaker 01

  Sure.

• 46:30Jason Calacanis?

  And like, I'm like, please, I'm just trying to make a decision, and then I called, and I, I won the hand, but I was 80, I was 70% folding, 30% maybe calling, because it was such a big number, and I'm trying to play it out in my head, and Chamak kind of.

• 46:45Speaker 01

  I would hate playing poker with Chamak.

• 46:46Jason Calacanis?

  Shouldn't have said anything, but putting, putting that aside, like, I, he's a good poker player, he's a great manager, he understands, he should just be coming out right now, and just saying, hey, everybody, here's the new corporate policy, I speak for the company, anybody else wants to speak for the company? Absolutely a great idea, send me a written proposal of why you should be speaking for the company instead of me, and instead of, you know, Greg. Like, we're the two people who are gonna speak for the company, and before Greg's tweets, and before I tweet, we just make sure we're aligned, anybody else want to jump into the fray? Great, you can like what we say, but you can't reply with words.

• 47:23Speaker 01

  Right.

• 47:23Jason Calacanis?

  That's what I would set as the new policy. Ridiculous, all right, let's keep going.

• 47:27Speaker 01

  It's getting out of hand. All right, we have another live caller, an NYU student, David Rosenberg is calling in, David.

• 47:33Jason Calacanis?

  I've got some questions about NYU, I'm gonna get some questions.

• 47:36Speaker 01

  You know, I, I got into NYU, I almost, I, I ended up at UCLA.

• 47:40Jason Calacanis?

  I went to Fordham, that was like, the safety, I mean, NYU was a little tick above, but I used to hang out in Manhattan when I was coming up, and let's just say, there were some nice, um, young ladies who, uh, we would like to meet in the East Village for a cup of coffee.

• 47:56Speaker 01

  Sure, you know, a nice cup of coffee, some chat.

• 47:59Jason Calacanis?

  We would, we'd fraternize.

• 48:01Speaker 10

  Jason, how you doing?

• 48:02Jason Calacanis?

  Are you in your dorm? This is a fancy dorm here, you got your own apartment?

• 48:05Speaker 10

  I'm at Rosh Hashanah tonight, so.

• 48:06Jason Calacanis?

  All right, of course.

• 48:07Speaker 10

  Again, um, yeah, I'm excited to, to be here, I'm a big fan of the All In Podcast.

• 48:13Jason Calacanis?

  Thank you, tell us your question.

• 48:15Speaker 10

  So, I was curious what you guys thought about, um, I mean, there's like, a sentiment in the academic community, specifically in the physics community, not like, I'm sure you've heard about the Navier-Stokes equation and how.

• 48:26Jason Calacanis?

  Sure.

• 48:27Speaker 10

  They solved it, so. I think there's like a little bit of backlash in the academic community concerning like using artificial intelligence to solve like some of these really complicated problems. So I was curious about what you guys thought.

• 48:41Jason Calacanis?

  Well, it's a tool and you can brute force with it. You can have it extend your scaling ability to try different novel solutions. And I think at the end of the day, we just want to solve these problems. And what's going to happen is AGI, like artificial general intelligence. Hey, I could be a competitive lawyer. I could be a, you know, an accountant. I could be a podcast producer. What kind of AGI? I think anybody would look at any profession, David, and put these tools against it. If properly implemented, will do as good a job as a human, almost universally. If you, even surgery, like you could just put an AI on it.

• 49:24Jason Calacanis?

  If you spent enough time, you know, one year just making it perfect, it's going to beat the average surgeon. Okay, so if we're at AGI, super intelligence is this thing can answer questions we cannot. What I took away from one of these big seven questions getting solved or partially solved or sending people in the right direction is we're tipping over into super intelligence. And what will really be super intelligence is when it proposes math problems to us that we didn't think of. So I think it's super exciting there that we're on the cusp of, or we're somewhere in the journey between AGI to super intelligence, depending on how much implementations being done.

• 50:06Jason Calacanis?

  I do think anybody who's doing this kind of research should not be using the frontier models. They should be forking any open source model, doing it on their own servers. And if it takes three times longer because you don't have the infrastructure, then these universities like the one you're at should be buying more hardware and get the go.ai proprietary server that you can put into your office. I think it's 40 grand, 50 grand. It's one of our investments. You could, that's where you're going to have to do this because they will have their large language model in some way, maybe not reading your inputs, but maybe looking at the outputs and

• 50:49Speaker 10

  Hundred percent.

• 50:49Jason Calacanis?

  Somehow it's going to, it's going to train it. So I think the academic community has just been put on alert that every prize is going to be taken by Anthropic and OpenAI. And if they're going to snipe everyone using the training data you gave it, it's like literally the top 100 mathematicians plus the next thousand students just using these AIs for fun to try to solve these problems. Those 1,100 people, the top 100 who are super credible and the thousand up and comers, collectively, the wisdom of that crowd of 1,100 math researchers and scientists, they're going to in aggregate train OpenAI and Anthropic's models to solve these problems where you as an individual researcher wouldn't have that.

• 51:32Jason Calacanis?

  So, but for humanity, don't you think David, it's just good to get this done? David, like

• 51:37Speaker 10

  Hundred percent. I'm in, I'm in total agreeance with you that we should utilize all of the tools that we have to.

• 51:43Jason Calacanis?

  Are you a math major? Is that what's going on here?

• 51:45Speaker 10

  Physics.

• 51:45Jason Calacanis?

  Physics, okay. And what, what, is there an equivalent in physics of the top, like unsolved questions that you're using AI to try to solve on your own?

• 51:55Speaker 10

  Well, me, no, I'm still in like the really early stages of my academic career there. But, um, I mean, yeah, hopefully eventually if I go into research, I'm not really sure if that's what I want to. I probably want to go into like tech and startups, that type of thing, but

• 52:07Jason Calacanis?

  It's a good idea, man.

• 52:08Speaker 10

  Until, um,

• 52:09Jason Calacanis?

  More money there.

• 52:11Speaker 10

  I heard, um,

• 52:12Jason Calacanis?

  Physics is like a great way to just understand logic. Like I went social sciences, like in other words, not science. Like, you know, like looking back on psychology, it's like, is it a science? Like there are moments of science that occur.

• 52:28Speaker 10

  Right.

• 52:29Jason Calacanis?

  But it's, you know, philosophy and, you know, techniques, but I don't know if I would call it exactly science. It has moments of science, but experimentation.

• 52:39Speaker 10

  People who are leaving like the more scientific roles at the frontier labs are now going into philosophy. So who knows?

• 52:44Jason Calacanis?

  Well, if you are trying to understand the right thing to do or humanity empathy, like I think the reason I have such empathy is because I was a psychology major and I aspire to go into forensic psychology, join the FBI. I was like on that track of like, how do people think? What is, how do emotions impact them? What are situational variables that cause people to do things? What's nature, what's nurture, all that kind of stuff. But if you're dealing with the physics person, they don't have that in their toolkit. So they can look at humanity as like maybe, um, not in a holistic manner. So it doesn't surprise me that people who are computer science or physics are like, ah, there's got to be some understanding here. And you'll know it's real when they pick up religion.

• 53:28Jason Calacanis?

  Because that's the thing that like, if you're completely unmoored and you're eventually going to look into this doomerism or whatever, the next step is these anthropic people are going to start going to church. That's when you know, like, we're in the end game. When they start saying, you know what? I need to get some grounding here. I'm going to go find Jesus or Mohammed or, you know,

• 53:51Speaker 01

  That's the Steve Bannon, like the Steve Bannon guy.

• 53:53Speaker 10

  It's like, um, what? Oh, sorry.

• 53:55Jason Calacanis?

  Oh, no, go ahead.

• 53:56Speaker 01

  Fair line.

• 53:56Jason Calacanis?

  I was just going to say

• 53:57Speaker 10

  There's a, there's a trend that, um, like the, the top researchers at these frontier labs are now like looking into studying the Talmud and like Halal.

• 54:07Jason Calacanis?

  My friend Darren Aronofsky made a film called Pi. And literally it made, you know, a, it's a thriller, but it kind of took

• 54:19Jason Calacanis?

  what was happening in mathematics and looking at the Talmud and, you know, all that kind of stuff.

• 54:28Jason Calacanis?

  He made it for $60,000.

• 54:29Speaker 01

  Yeah, that was his first movie. That was his first movie.

• 54:31Jason Calacanis?

  Yeah, I hosted a screening for him over at, um, the East Village Cinema. Was it like 11th and third or second?

• 54:40Speaker 10

  Something like that.

• 54:41Jason Calacanis?

  And, uh, I rented it and I rented the bar next door. I spent like 2,500 bucks when I was doing my magazine. Cause Darren and Sean, who was the star of that film, they had no money. They were trying to get attention for this, so they did a campaign, David. It was super cool. They took the pie symbol. They went around Manhattan with boxes with a stencil on the bottom of the box. They would put the box down, they'd look for cops, they would take a spray can, it would be in the box. And they go, shh. And then they would lift up their box and pretend they were delivery people. Then they would go to the corner again, and they sprayed the pie symbol all over the East Village, West Village, everywhere in Manhattan. And that became the guerrilla marketing for the movie Pie. I don't suggest you do that, it's illegal unless you're using washable paints.

• 55:26Jason Calacanis?

  In that regard, I'm told that it is legal. They didn't use washable.

• 55:31Speaker 10

  Park, there's markings all over the ground of like these random, you know, guerrilla marketing campaigns. It's cool. It's interesting.

• 55:37Jason Calacanis?

  You were going to say something, Lon, before we drop David off?

• 55:39Speaker 01

  Oh, uh, I was just going to say, does it feel a little bit like, uh, what I was going to say before is Steve Bannon's got that guy Joel Miller, who we were talking about, who he's the super religious anti-AI guy. But the other thing I was going to ask you was about OpenAI. Do you, do you feel like they botched this? I feel like they just, this was a pure like PR mistake. Like if they had gone to these researchers and said, we want you to publish your, you know, your solution, we'll help you and back you in any way you can. We would just ask that you note in the study, like we used OpenAI's models and thanks to Astra.

• 56:17Jason Calacanis?

  For sure, it was a PR blunder.

• 56:19Speaker 01

  Yeah, like

• 56:20Jason Calacanis?

  They should have, if they knew this was happening, they should have contacted the research team. This is how I would have handled it. I would have said, okay, there's a big prize here. We know these two people are working on it. We know these two people are using our software to do it. You, the correct thing to do is to contact them and say, we would like to give you two forward, uh, deployed engineers to support your work and give you unlimited credits. All we ask is you win the prize, you solved it. If our tools are helpful and if our support was helpful, you could just give us a thank you.

• 56:56Speaker 01

  Exactly.

• 56:57Jason Calacanis?

  But if we're not helpful, do not give us a thank you. If we're not helpful, don't give us a thank. Then he would be a hero and he wouldn't have sniped it and we wouldn't be in this situation. And they knew that was going on, they could have done that.

• 57:08Speaker 01

  And they could, they could get every cutting edge mathematician who's working on high level stuff pouring it. Oh my God, if you use OpenAI and you have success, they're going to give you free compute to help crunch all these numbers, you're going to win the next millennium prize.

• 57:20Jason Calacanis?

  How do, why do startups have a good feeling about some of these frontier models? It's because they get tokens. And if they get free tokens, David, when you start your company, like you'll come to me when you're at Founder University or Accelerators, I encourage you to apply. Just go find two more. All you have to do, David, find two more really smart people and start building stuff on the weekends. Build like three or four things, whatever thing is the most fascinating and interesting and weird to you. Then just email it to your unk, okay? Jason, I can

• 57:48Speaker 10

  I thought I was doing over the plumber to make some side, some side cash.

• 57:50Jason Calacanis?

  Well, forget about side cash, just make something that you're like passionate about and that's weird. And that if it did work and a million people used it, they'd pay a hundred dollars a year for it. Like, can you come up with something that a million people would pay a hundred dollars a year for? Or a hundred thousand people would pay a thousand dollars a year for? 10,000 people would spend $10,000 a year for? In other words, can you imagine if you got like six steps done, that a hundred million dollars in revenue come? Then you're in the venture game. And we would love to hear and review that prototype with you, David. All right, good luck in your studies.

• 58:23Speaker 10

  Thank you very much. I just wanted to ask one last thing.

• 58:26Jason Calacanis?

  Oh, I just want to give you one piece of advice here.

• 58:28Speaker 10

  Oh yeah, please.

• 58:28Jason Calacanis?

  My grandfather, on my Irish side, would give me the same advice every time I was going to school. Keep your mind on the books and not the babes. Stay focused on the books and not the babes.

• 58:42Speaker 01

  Sound, sound advice.

• 58:44Jason Calacanis?

  It was sound advice from my Irish grandfather.

• 58:46Speaker 10

  Yeah.

• 58:46Jason Calacanis?

  The rest is all, yeah, one last question, go ahead, David.

• 58:48Speaker 10

  Yeah, I just, I promised my friends if I actually made it on to the This Week in Startups, that I would, um, I mean, I'm a big fan of the All In Podcast. I know you guys bring some scholarships for the, uh, the All In Summit.

• 58:58Jason Calacanis?

  Okay, sure, you got one.

• 58:59Speaker 10

  I mean, listen, I'm not asking you, I just wanted to, like, you know, throw the idea out there. See what, see what's

• 59:04Jason Calacanis?

  Are you in LA? Can you make it to LA?

• 59:06Speaker 10

  When is it?

• 59:07Jason Calacanis?

  It's next week, it's on Monday and Tuesday. Yeah, Sunday night.

• 59:11Speaker 10

  If I get a scholarship, I will be there, I promise you.

• 59:13Jason Calacanis?

  All right, email me, uh, they'll give you my email. I'll see if I can, I could scrounge up a ticket for you. I might have one left.

• 59:19Speaker 10

  All right, well, I would love to be there, love to meet you. And I really appreciate the time.

• 59:23Jason Calacanis?

  All right, nicely done, nicely done.

• 59:25Speaker 10

  Thank you very much.

• 59:25Jason Calacanis?

  Thanks, David. I have an extra ticket, man. Crazy thing is, Lon, all my rich friends, like, uh, I, I, yeah, I'm the director of blah, blah, blah at this giant bank. Can I get a free ticket? I'm like, or I'm a venture capitalist, or I have a family op. I'm like, wait a second, you're a venture capitalist, you want a free ticket? Like the, the, the people with the most money are completely

• 59:50Jason Calacanis?

  Uh, un-

• 59:51Speaker 01

  Yeah.

• 59:52Jason Calacanis?

  Ashamed of asking for free tickets. I'm like, the student who could never afford to go. Like, he's like, ah, if there's any way.

• 59:58Speaker 01

  Yeah, it's like the Hollywood gifting suite phenomenon. It's like, none of the people in the Hollywood gifting suite need the Hollywood. They could go buy all this stuff for themselves, but they, they love it. They want to have it. All right, we got one more. We got one more live caller. We've got Mo- Wow, live caller is doing well. We've got Mo in the waiting room. We've got to let him in, and then we've got a Friday show to get to. Mo, you're in your car.

• 1:00:19Speaker 11

  I am in my car. How you doing?

• 1:00:21Speaker 01

  Good to see you.

• 1:00:22Jason Calacanis?

  All right, Mo, are you driving right now on self-driving, or you're, you're parked?

• 1:00:25Speaker 11

  I parked. I parked.

• 1:00:26Speaker 01

  Okay, good.

• 1:00:28Speaker 11

  Hey there.

• 1:00:28Jason Calacanis?

  You got the latest update? I don't mind.

• 1:00:31Speaker 01

  What's your, what's your question?

• 1:00:32Speaker 11

  Hey, Jason. So, um, you know, one of the things is I've been, I listen to your pod every, every week at All In, and what I've been trying to figure out with data centers is, you know, I'm in the cannabis space. I actually just exited, um, I, after 13 years, and there's a lot of production that's been built up in indoor capacity for energy. So you have basically 1.5 to 3 megawatt indoor cultivation centers that are sitting there. And I talked to my city council member about, you know, adaptive reuse and saying, hey, if this thing's already there, can you convert it to a data center instead of, you know, doing a ground up? So there, they said, okay, if it's not using more energy and water than before,

• 1:01:11Speaker 11

  we'd be okay with it. So going down this path, do you see these smaller capacity data centers being a thing, or do you think it's just a, you know, 1 gigawatt, half gigawatt, or more that we have going on right now?

• 1:01:26Jason Calacanis?

  You're gonna see both. Uh, Crusoe Cloud is gonna do giant, you know, uh, data centers in Texas, Abilene, wherever, over the various natural gas fields, and fantastic for them. In three years, Elon will start having like, serious amounts of computing space, um, and he'll demonstrate it probably next year, is my guess. I think I heard him say that publicly. Um, so then there's what's in between. And I think the most exciting things for me are places where, uh, you have a lot of electricity already, you have cooling already, and, uh, you know, you got space, electricity, and cooling. So if you were

• 1:02:11Jason Calacanis?

  to do a first principles thinking, like if I asked Lon to put that into Claude, where are there, where is there extra climate-controlled space, um, that could be used for a, a data center? Just, you know, forget about AI, just, where can we put in a small 10,000-square-foot data center? A 5,000-square-foot data center? A 1,000-square-foot data center? Lon's gonna do that right now. One of the things that will come up is, well, residential. Somebody's got a garage, and they have a three-car garage. What if in your three-car garage, you gave up half a garage slot and had your internet fiber connection, as well as your solar and

• 1:02:53Jason Calacanis?

  batteries and your power walls powering a small roll-in four racks, and you got paid for that? In other words, you used your garage or shed like Airbnb, because you had extra energy at your ranch. What if you took Target, Walmart, HEB, and they have this giant superstore? And you put into the superstore, and you said, we're gonna just take, uh, a quarter of the back of the superstore, this department store, that's already climate-controlled, that already is empty from 5:00 PM to 5:00 AM, whatever it is, their hours, and nobody would notice it, but we could roll a bunch of racks in there. People are gonna get creative with the stuff. I saw somebody develop a building in a city and just say, these four lower

• 1:03:35Jason Calacanis?

  floors, not the ones for retail, but right above retail, you know, floors three through seven, which have no views, why don't we just close that off, put in redundant power supplies for the rest of the building, and just make them into data centers, and have the data centers in every new building? So every new building, like it has a couple of floors of parking, couple of floors of retail, when you create that mixed-use building, why don't you just always put two floors in for data centers? And those people will be renting it out immediately, and it won't bother anybody. So those creative solutions, I think, are what we're gonna be forced to do, or there might be creative countries. You might take a country like, say, Puerto Rico, and Puerto Rico says, you know what?

• 1:04:17Speaker 01

  Territory.

• 1:04:19Jason Calacanis?

  What's that? Territory?

• 1:04:20Speaker 01

  Not, not a country. It's a US territory.

• 1:04:21Jason Calacanis?

  Okay, so you get a US territory like Puerto Rico. They say, you know what? We got a lot of land, got a lot of sun, and, you know, it's not a perfect analogy, because it's hard to get things over there. You gotta ship stuff over there. But you get some region, uh, some territories, some province in Canada, and they say, you know what? We got nothing going on here, but we got a lot of sunshine, we got a lot of hydro. Bring us your data centers. Um, and so if the US is gonna fight these, and Abbott and Shapiro are gonna fight it, you know who's not gonna fight it? UAE, Qatar, Saudi, they're all gonna be like, yeah, data centers, we got energy, we got sunshine, and we got massive amounts of space. So don't be surprised if like, the US winds up giving this entire data center business

• 1:05:06Jason Calacanis?

  over to the, the Gulf monarchies. And the Gulf monarchies can move fast, because they're monarchies. They can make unilateral, top-down decisions. It's one of the great things about a monarchy. And you know what? If I'm MBS, if I'm MBZ, if I'm one of those royal families, I'm gonna just say, you know what? Uh, we're gonna, we have oil, we got sunshine, just start building data centers. We'll find the customers later. And what's the difference if the data center for OpenAI and X.com, whatever, who cares if the data center's in UAE? You got fiber optic lines, you can put multiple industrial, uh, Starlink's now on the roof. So, uh, compute finds a way. You ever hear that line from Mo, the line

• 1:05:50Jason Calacanis?

  from Jurassic Park, "Nature finds a way"?

• 1:05:52Speaker 01

  Life, life finds a way.

• 1:05:53Jason Calacanis?

  Life finds a way. Thank you. Life finds a way?

• 1:05:55Speaker 11

  Finds a way.

• 1:05:56Jason Calacanis?

  Compute finds a way. Uh, and so we gotta get our asses in gear in America, and get this anti-AI sentiment squashed. And all the data centers already are like, we're gonna make electricity go down, but we're only gonna put it there if it doesn't bother people, closed-loop system for water, you know, it's gonna use less than, you know, almonds. Uh, so I think it's a moment in time. This whole thing, the whole dialogue's gonna change, Mo, after the midterms. When the midterms end, you're gonna be amazed at how quickly politicians are like, you know what? We worked it out with the data center companies. They're just doing it to get votes and turnout right now.

• 1:06:32Speaker 11

  Do I have time for a follow-up?

• 1:06:34Jason Calacanis?

  Yes, you do, sir.

• 1:06:35Speaker 01

  Sure.

• 1:06:35Speaker 11

  All right. So you know when I hear Mark Cuban saying, these things are gonna be pickleball courts? I, I don't really believe that. I don't know if you do either. Um, it's, it's one of those things, because what I'm trying to figure out is the capex that's going into these right now, like in the US. Let's say if there wasn't a regulatory barrier, right? And at the same time, you see like, Nvidia's got their Grace Blackwells, potentially doing 10x efficiency relative, you know, in some use cases, relative to the previous generation and thermodynamic. I don't know where that's going. Do we think we'll hit energy efficiency where you may not need a 1 gigawatt data center, where you're able to actually do it a fraction of that? You see, like, if you go out, you know, in terms of efficiency, does efficiency always find a way, too?

• 1:07:19Jason Calacanis?

  I think what you, what you, what Cuban was referring to is people who were building before they had the electricity and everything permitted, so they were like, let's just get the foundations going, we'll do this in parallel, we'll, we'll get our permits while we're building. Yes, in those cases, they might lay a foundation and be stuck with a foundation, because they get blocked by a state. So I think he's not wrong about that. They could, those could wind up being apartments, condos, or just relics, you know, in the desert somewhere, that never get used for anything, and they just sit there and decompose over the next century. So, um, it is a competition. The build-out's gonna be Fast and Furious

• 1:08:01Jason Calacanis?

  for another 24 months, but at some point, there's so much energy and development put into inference chips, efficiency, photonics, energy, nuclear energy, that these problems will get solved, and then we'll be in the dark fiber phase, where we just overbuilt. I don't know when we hit overbuilt, but it's inevitable that we hit overbuilt, I believe, because, um, there's a small set of problems here. Energy, space, um, regulation, and the-

• 1:08:36Speaker 11

  Interest rates. Interest rates to build out.

• 1:08:38Jason Calacanis?

  Sure, that's another one, yeah. So if interest rates got too high, people would be like, you know what? Instead of investing in more of these, let's make the ones we have more efficient. So let's take people off of building data centers, let's redeploy that capital, we'll take 10 billion and put it towards making more efficient chips, photonics, and upgrading the hardware stack, and then making the models more efficient. Because if you can do more with the same hardware footprint, if you invest more in optimization, then you need less hardware. That same thing happened with Facebook and Google. They looked at-

• 1:09:11Speaker 11

  That's the crux of my question. Yeah, that's the crux of it, right there.

• 1:09:14Jason Calacanis?

  I mean, they, they looked at something like Gmail, and they were building these huge Gmail servers for the world to use, and YouTube when they were deploying it. And then they said, okay, at the same time, how can we, uh, make storage work better? How can we use CDNs to make it faster? So when you go to another country, instead of having your files in Japan on a local server, the first time you log in, Gmail might be a little slow, because your data's in the United States. But once you've logged in a second time, Lon, in Japan, it knows you're in Japan, and it just copies your entire Gmail file over to the local Gmail there. But it doesn't keep everybody's Gmail in all places at all times. There's an equivalent there with AI. Okay, great questions.

• 1:09:55Speaker 01

  You guessed, uh, almost all of Claude's suggestions, by the way. Single-story, big-box retailer, dead malls, warehouses and logistics buildings, uh, legacy-

• 1:10:04Jason Calacanis?

  Warehouses by airports, good one.

• 1:10:06Speaker 01

  Yeah, legacy, industrial and manufacturing plants, former power plants, and office buildings. Those were Claude's picks.

• 1:10:12Jason Calacanis?

  I mean, if you have an office building and people are renting it for $20 a square foot, then you can't get rent there, and you can convert it.

• 1:10:19Speaker 11

  We put $5 million into our cultivation, right? Just in electrical. So that's where I was like, hey, is there value here in those far greater size ones? So there's a big thing out there. Well, thank you.

• 1:10:29Jason Calacanis?

  Great job. Talk to you soon. Bye. If you have a question for me on Ask Jason, very simple, you're gonna email [email protected], send us a video file, or at least an audio file, and we'll get your question on air. At the end of your question, you can put a little promo for your company. We'll give it a shout out, maybe even give you a link in the show notes. Next week, we're gonna do another Ask Jason episode. If you'd like to ask a question next week, or go live, thisweekinstartups.com/askjason, thisweekinstartups.com/live, to find information about the next live Ask Jason show. And if you want to email your question in video or audio, and you don't wanna come on live, because you got a busy schedule, [email protected].