How AI splits startups into winners and losers | E2322

Summary

Podcast Summary: This Week in Startups

Episode: How AI Splits Startups into Winners and Losers | E2322

Date: August 7, 2026
Host: Jason Calacanis
Co-host: Lon Harris
Notable Guest: Kofi Asante (CEO, Blue Core Energy)

---

Overview

In this episode, Jason Calacanis and Lon Harris analyze how AI has rapidly transformed the competitive landscape for startups, particularly in the SaaS sector. They discuss the effects of AI adoption—or lack thereof—on leading public tech companies such as Figma, Twilio, and Airbnb, and dive deep into trends around stock-based compensation. Later, Jason and Lon interview Kofi Asante, CEO of Blue Core Energy, on the promise and challenges of deploying small modular nuclear reactors on barges to power ports (and potentially AI data centers). The episode closes with an engaged discussion on AI competition, the irreversible march towards increasingly capable AI models, and a candid, motivating segment on personal health and a team weight-loss challenge.

---

Key Discussion Points & Insights

1. Markets & The SaaS Divide (00:00 – 09:49)

• Impact of Hedge Fund Collapse: The episode opens on recent market turbulence in SaaS after a major hedge fund collapse, highlighting volatility but also new opportunities for companies embracing AI.
• Three Key Companies Examined:
  • Figma (01:54):
    • Down 77% from its high, but with healthy revenue growth and a founder (Dylan Field) signaling strong alignment with shareholders by giving up ~$46M in stock awards.
    • Jason highlights:
      • Importance of founder-led companies, especially ones that “have something to prove."
      • Risks of excessive stock-based compensation: “If the company's not doing well and management is giving out, you know, 1, 2, 3%...you’re diluting over 5 years, 15, 20% of the company while the stock is sideways.” (03:02)
  • Twilio (05:42):
    • Not founder-led anymore, but shares surged 31% after a blowout quarter with strong free cash flow.
    • Embracing AI as a “conversational layer," adding intelligent routing and automation.
    • Unique among SaaS for using consumption-based rather than per-seat pricing.
    • “They are turning this around because they have embraced AI fully and that’s really what all these SaaS companies need to do if they want to survive.” (06:23)
  • Airbnb (09:28):
    • Measurable improvements in bookings and customer service from AI investments.
    • Struggles with customer vs host friction are discussed, with Jason suggesting more rigorous host certification for quality assurance.
    • “I think they need to bifurcate…it’s a service business, and they need to bucket hosts who care about hospitality.” (13:00)

2. Stock-Based Compensation (07:40–17:03)

• Scale and Controversy:
  • Discussion of which companies (public and private) are most generous with stock-based compensation and the effect on investor sentiment.
  • Harmonic AI and Perplexity help surface the most up-to-date data: Significant for attracting talent in startups, but potentially problematic at scale (e.g., Snowflake’s SBC = 35% of revenue).
  • “In a public company…institutions are looking at a menu of choices and saying, which names are just out of sync on stock-based compensation?” (16:03)

3. Interview: Nuclear Power & Decentralized Deployment (17:17–42:25)

Guest: Kofi Asante, CEO of Blue Core Energy

a. What Are Small Modular Reactors (SMRs)? (18:08–19:47)

• Definition: SMRs are compact nuclear reactors offering rapid deployment, high energy density, and modularity.
• Use Cases: Ideal for ports, off-grid communities, and potentially powering energy-hungry AI data centers in locations lacking grid capacity.

b. Innovating with Floating Reactors (25:30–29:47)

• Deployment Strategy:
  • Using barges to place reactors offshore, circumventing complex permitting and siting challenges.
  • “Our facility, I wanted it to be based here so that everybody understands that when we say it’s safe, we are literally standing right next to it.” (25:58)
• Regulatory & Supply Chain Factors:
  • Need to comply with ABS, Coast Guard, NRC, DOE.
  • US seen as both most innovative and most welcoming, but significant global interest.
• Jurisdiction & Perception:
  • Barges can be pushed further offshore if needed for public comfort, still tethered by cables to port or grid.

c. Safety, Waste, and Public Perception (31:35–38:46)

• Addressing Fears:
  • High redundancy, smaller size, and multiple safety barriers reduce risk.
  • “It’s probably the highest bar for redundancy and safety on top of the fact you’ve actually made it much smaller.” (34:00)
• Waste Handling:
  • Comparable or lower than existing industrial nuclear plants, managed under existing NRC regulations.
  • “These things are happening. Our job is not to create new science…but to fit into guidelines that have already kept us safe.” (36:26)
• Historical Precedent:
  • US and Russian navies already have nuclear reactors on subs, some lost at sea without noted environmental disaster.

d. Projected Timeline and Regulatory Support (38:46–42:12)

• Timeline:
  • Three to five years toward real-world deployments, enabled by not having to invent everything from scratch.
• Government Involvement:
  • Strong support and collaboration from federal, local, and industry stakeholders.
  • “We were able to say we want to enable commercial nuclear energy in ports. This is the largest port in the US.” (40:40)

4. AI Race: Model Competition & Open Source (47:45–57:59)

• ByteDance's Massive Model:
  • Training a 10 trillion parameter model, surpassing current records and rivaling top US models. Raises questions about the risks and ethics of data export (esp. to China) and national security implications.
  • “We are now crossing into an area…the pace that no one human can keep the innovation in their brain.” (51:49)
• Open Source and Control:
  • AI models learning to "cheat" or bypass constraints, steadily moving out of reach of centralized control and oversight.
  • “With open source, you can just fork it and remove [safety] constraints. So cat’s out of the bag. Everything’s going to get hacked. Act accordingly, folks.” (55:12)

5. Weight Loss Challenge & Personal Transformation (58:10–68:43)

• Personal Stories:
  • Jason, Lon, and Jason’s brother Jamie discuss their experiences with GLP-1 drugs and the benefits—weight loss, mental focus, regained time—plus the personal, transparent approach to health and accountability.
  • Jason: “If you want…you have to opt into it. But if you were looking to take that journey, hey, I’ll pay for it, right?…Because I’m an employer. And, you know, listen, if it’s on the insurance, great. If it’s not on the insurance, whatever it is.” (58:49)
• Motivation & Milestones:
  • Discussing tiered goals (e.g., –20lbs, –15lbs, etc.) and public accountability.
  • Jason offers $5,000 for each weight loss tier achieved.
  • “I want you guys to be around for a long time and I want anybody in the audience who struggles with their weight to at least investigate this option.” (66:12)

---

Memorable Quotes & Moments

• On AI Shifting the Startup Landscape:
  • “Companies that have embraced AI and gone AI-first have had a little bit of a turnaround here and it’s a mixed bag.” – Jason (01:19)
• On Founder-Led Companies:
  • “When I see a founder being a little bit chippy…that brings up the bigger issue which is stock-based compensation.” – Jason (03:02)
• On the Future of Nuclear-Powered AI:
  • “If you want to currently go and get like a commercial small modular reactor and put it with your community, there are great companies that are working towards that.” – Kofi Asante (22:57)
  • “When you actually get production to the right pace…then you can get the cost down to make it economical.” – Kofi Asante (24:12)
• On Safety & Perception:
  • “It’s probably the highest bar for redundancy and safety.” – Kofi Asante (34:00)
  • “The status quo is inherently dangerous today…and so that’s the first thing…” – Kofi Asante (31:35)
• On AI Model Race:
  • “If your model doesn’t break the law, are you doing your job?...There seems to be a competition for being like, ‘oh no, you won’t believe it, man.’” – Jason (52:20)
• On Health & Weight Loss:
  • “Instead of eating five or six slices of pizza, now I eat one and a half. I get that there’s no change in my lifestyle…” – Jason (67:20)
  • “I didn’t realize how much time I was thinking about food, snacking, and the time it took me to eat. When I started eating much lighter, I recouped so much time…” – Jason (77:22)

---

Timestamps for Important Segments

• 00:00 – Jason & Lon open: Market pulse and SaaS company performance
• 01:54 – Figma: AI, stock performance, stock-based compensation
• 05:42 – Twilio turnaround: Embracing AI, business model strengths
• 09:28 – Airbnb & AI in travel; host/guest tension
• 17:44 – Interview: Kofi Asante, Blue Core Energy – Floating nuclear reactors
• 18:08 – What are SMRs? Distribution challenges and port use cases
• 29:09 – US vs. global market for SMRs
• 31:35 – Safety, waste, and public concern about floating nuclear reactors
• 38:46 – Timeline and regulatory pathway for SMRs
• 47:45 – AI model race: ByteDance, open source, national security
• 58:10 – Weight loss challenge, personal health transformation stories
• 68:43 – Reflections on health, AI, and the future

---

Closing Thoughts

The episode is a vivid snapshot of how AI forces a reckoning in technology, from markets to infrastructure (and even personal routines). Winners and losers are being determined not just by technology adoption, but by leadership, culture, and systemic thinking across sectors. The nuclear conversation underscores a convergence between AI’s demand for energy and solutions once considered futuristic. The candid health segment models leadership not only in the boardroom, but at home and in personal wellbeing.

---

For more, catch the full episode or follow This Week in Startups for ongoing discussions on tech, innovation, and the entrepreneurial journey.

Transcript

[A]

All right, everybody, welcome back. It's this Week in Startups with me again, my guy Lon Harris. We've got two amazing guests today. We're going to get to them in a moment. We've got ro wait. Bet that's coming at the end of the show with my brother Jamie and Lon. And lots of news this week in Startups is brought to you by Lightfield, the AI native CRM that updates itself so your team never has to try it for free. At Lightfield app Sentry, your team should be focused on shipping features, not chasing down bugs. New users can get $240 in free credits when they go to Sentry IO Twist and use the code Twist and DigitalOcean. Want to see what building on a true AI native platform looks like? Head to dot co twist to start building on Digital Ocean's AI native cloud today and cut your AI workload costs by up to 50%. First thing I want to get to is markets, Lonnie. We've had a lot of market activity. Markets hit an all time high, as many people know. But what has been surprising is after this hedge fund blew up situational awareness, he had a lot of shorts on SaaS companies.

[B]

Correct.

[A]

And SaaS companies have been taken to the woodshed. Right. We just saw Airtable get sold. I think we talked about that already on Wednesday's show. But companies that have embraced AI and gone AI first have had a little bit of a turnaround here and it's a mixed bag. But if you look at, and I want to just go through three people who reported earnings, Airbnb, Figma and then Twilio. Now I own all of these, I believe, and I bought them all during the COVID right after Covid when I started J trading I think in 2023. So I think I'm massively up on all of them. But let's do this rapid fire. We'll get through these three important companies and what we can learn from them by their AI embracing of AI or not.

[B]

All right, so if we'll start off with design software. Figma, down about 77% from its high of $122 set shortly after last year's public offering. It's at about $23.81 right now. Co founder and CEO Dylan Field's actually given up around $46 million in stock awards, which he was set to receive back in July in hopes of boosting investor confidence. Here's his quote. Dilution is something that I feel just like our shareholders feel, given that this is an investment period right now that felt like the thing that made sense to do. For his part, Field posted on X that he thinks things are going generally well. He points to $370 million in revenue in Q2, 48% year over year revenue growth rate. It's their third straight quarter. Jason of accelerated growth for Figma. You just bought in a big J trade on Figma only about a month ago after getting into a discussion about the health of the company with Dylan

[A]

Field on X. Yeah, what did I get in? What is the price? I got in. So I got in at $22. So I'm basically at par right now. I bought 4,500 shares and you know, when I see a founder being a little bit chippy, a little bit with something to prove and founder led companies, especially when they're public, they have founder led authority. For Figma, him giving up a huge stock ramp brings up a bigger issue which is stock based compensation. And a lot of companies lon have gotten into giving huge amounts of stock based compensation a year. 1, 2, 3%. If you're a shareholder and the company's not doing well and the management team is giving out, you know, 1, 2, 3% of the company every year. I think Snap was kind of in this with a lot of stock based comp going on and they don't put it into their books. So you don't take it as an expense. They kind of put it off the books. And this has made sophisticated investors. Bill Gurley talks about it a lot. But it's a tough thing to bring up because you want people to be incented by stock. But if the company's not doing well and you're taking out 1 or 2 or 3% of the company every year, you're diluting over 5 years, 15, 20% of the company while the stock is sideways. That gets people's Irish up. And then shareholders are like, I don't want to be involved in this company because they're just taking and taking and not performing. So I think what Dylan did here with Figma saying, hey, I'm not going to take, I think it's like a $50 million grant. He probably owns 20, 30% of the company. I'm taking a guess here. That would be a high number for a publicly traded company. But what does he need another 50 million? He's already, if he owns 10% of the company even he's got a billion plus. It's an $11 billion market cap company. So this is signaling which is, hey, I'm not going to try to make an extra 2, 3, 4, 5% on my net worth. I'm going to give that back to the shareholders and I'm going to show you how great this company is. Now of course they're up against Claude Design, they're up against Gemini, they're up against Lovable, they're up against some serious competition. But I like a founder led company. I think the stock will be trading at 2x in three years. That's my guess right now. If they're making 350 a quarter I think is what you said puts them at 1.4 billion. That means they're trading at like nine times sales or something in that range which is pretty high multiple. We'll see what the earnings multiple is over time. But I believe in the founder. I think he understands designers in design. I don't think design is going away. I think we're just going to get more great design. Now if we go to the next company, which is Twilio. This story is also pretty fascinating. People had given this company up for dead. No longer founder led, but they're up 30% today. So give us a story here.

[B]

Thanks to our friends at Perplexity, by the way, for helping us out with this. Twilio shares are surging up to 31% to near 52 week highs after the company delivered a blowout Q2 2026 report and sharply raised full year guidance revenue of a dollar revenue of 1.5 billion top estimates and grew 22% year over year. That's 17% organic while adjusted EPS of 1.47 beat consensus by over 11% driven by record free cash flow of 352.6 million and accelerating organic growth.

[A]

Yeah, so it's becoming a money printing machine. They have free cash flow. That means they could buy their stock back or they could acquire companies. But they have started to do a conversational layer and they're doing AI. As people remember, Twilio was an API. If you call an Uber and you get a text message or an in app message or you want to go shopping somewhere and you get SMS messages, email messages, they did all that. So it's like AWS for messaging and so everybody plugs there application into that and people were spending millions of dollars a year with this. Now they're adding agents and AI that will take live conversations and then have these conversations between agents sending messages to each other, which is something that people are going to want to have and routing customer support, et cetera. So I think that they are turning this around because they have embraced AI fully and that's really what all these SaaS companies need to do if they want to survive. And the per seat model was under attack. Lon and Twilio never had that. They were always based on consumption. How many messages are you going to send? So it's interesting to see that company turn around. Yeah. Dylan owns 9% of the company, so he's got a billion dollars there in equity.

[B]

By the by, I'm going to share my screen. You are asking for companies that have the largest percentage of stock based compensation. They note that it's difficult to say this now. We don't have a totally open kimono for all of these private companies. But based on what we do know, S1 filings, OpenAI seems to be leading with about 6 billion in estimated annual stock based compensation, followed by SpaceX XAI with about 2.8 billion. And then they're guesstimating that Stripe, Databricks and Anthropic would follow pretty close behind. In terms of public companies, we've got Alphabet leading the way. 25 billion in annual stock based comp, then Meta 20.4 billion, then Amazon 19.5 billion and then Nvidia with 16.2 billion.

[A]

So these are ranked just so you know, by the total amount of stock based comp, which, you know, those are very big companies. The way, you know, sophisticated investors might look at it is what is it as a percentage of revenue. And if you look at Snowflake, that's 35%. I think their stock based comp is at 35% of their revenue number. And that's when people get a little tweaked, is when this gets to be 20%, 30% of revenue. So just small in the weeds observation. Let's go on to Airbnb. This stock has been much maligned and everybody thought Airbnb was going to be this rocket ship and that kind of hasn't happened. If you were to look at the Airbnb, say five year chart and there's a lot of excitement about it because it's such a great product today they're trading at $140. So if you put money into this company, it's been sideways forever. But they did have a big rebound. So maybe we could talk a little bit about that in today's results.

[B]

Airbnb CEO Brian Chesky says AI is improving every part of their business. He's ready to spend more on tokens. Chesky said the gains are measurable. AI is helping Airbnb attract more bookings, making it easier for hosts, lowering customer service costs. The travel and booking company recorded Q2 earnings that beat expectations and raised full year guidance. Thanks to CNBC for that quick update.

[A]

Yeah, the truth is most of the big popular models can handle your AI workloads now. Your performance, not to mention your cost, that really comes down to your infrastructure, not your prompts. That's why I want to tell you about DigitalOcean's inference engine. The best way to serve your outputs fast while keeping things affordable. Which is super important. Right? DigitalOcean is offering you three ways to run your workloads from a single connection point. First, serverless for real time chatbots and agents. Second, batch for big jobs you're running in the background. And third, dedicated for those heavy, always on workloads where you need more control. It's all running on chips tuned specifically for these kinds of jobs and comes with built in monitoring. So you can always keep a close eye on performance. If you want to understand what building on a true AI native platform looks like, go to dot co twist. Start building on the DigitalOcean AI native cloud today and cut your AI workload costs by up to 50%. That's dot co twist and that's super critical to look at. Maybe this stock that went out fully valued, $100 billion market cap company AI might be able to help them break out. How? Well their earnings could surge if they can lower costs and maybe find more customers. They do have a challenge, I think, which is some people have. There's a bit of a war. Yeah. Lon, I don't know. Last time you did an Airbnb, but recently the novelty. You did one recently.

[B]

I was in Rome. When I was in Rome, we stayed at an Airbnb.

[A]

Let's see if this came up. The number of instructions and the tension between Airbnb hosts and their annoyance at the customer staying in their homes or their properties. In some case, this has like turned a corner where they're just. It feels like angry at the customer base.

[B]

Yes, I have. I did not in Rome, Julia and Federica, my hosts. Lovely. They were wonderful. I had no tension at all. But when I moved to Austin, I was staying at Airbnbs until I could fully move in and my furniture arrived and I could move into my new house. So I was in two or three different Airbnbs and this was a chronic problem where there was a lot of hostility coming at me and there were so many rules like trash goes here and do this and turn this off and don't go in here. Don't touch this. And go here and park here, but not right there. Pardon? Next to that by the trash can. And I had a neighbor even yelling at me that I parked in the wrong place, even though I thought I had followed the instructions. And at a certain point, I do agree. You're like, oh, I should have just stayed at a hotel. Hotels don't give you any flack. There's no agita. You just get in, you put your stuff down. I do feel like some of these Airbnbs, it's becoming a little bit fraught.

[A]

Yeah. And I think they need to bifurcate it. Like, they need to have the hosts who understand service and go through service training. Will Garado, who wrote Unreasonable Hospitality, I had him on the program.

[B]

Yeah, I do remember that. Jacob, look that up.

[A]

He's awesome. But they need to bucket. This is what I would do if I was Brian. I would put people through host training, buy them the book Unreasonable Hospitality, and say if you are in this category of Will Gadara.

[B]

Will Gadara, Will Godara.

[A]

Thank you. And the movie, the TV show the Bear is based upon his Unreasonable Hospitality insights in large part. And he actually, I think, was a producer, and I think he appears on one of the episodes as like, he's had dinner with the other restaurateurs. Anyway, they need to have certification and a star for delightful hosts who want to service you. I had two experiences over the summer. My wife likes to stay in Airbnbs because she wants everybody in the kitchen and in the living room together. I would have preferred to stay at some hotels anyway because I like the full service nature of them. We had a serious problem in Paris. People kept saying the Airbnbs had air conditioning. And then the air conditioner was like, jimmy rigged. It was just, here's an air conditioner. The doors are open out the windows, and you've got hot air coming in and this little hose going out. Charming of them to try to fix the air conditioning problem. But it didn't work. So now I'm back and forth with the guy that he's like, listen, I give him a lot of credit. He was like, if you want to leave, I'll just charge you for the night, you stay. I understand. That was good.

[B]

That's nice. Yeah.

[A]

Then I was like, yeah, but that's going to be, like, a big inconvenience. And then the elevator's broken. I said, maybe we should just work out a discount. Gave me a 10% discount. I don't even need the discount but I thought that was a nice chance.

[B]

It's nice for him to meet you halfway. They already have the super host concept so I feel like it's a natural. Well, those people should have a little background in hosting and hospitality. Let me, before we move on, I do want to get to our guest Harmonic. I do want to go back to Harmonic because I got exactly what we were looking for. Let me just share my screen here again one more time.

[A]

And Harmonic for people know is like just this huge database of startup knowledge and company knowledge. It's incredible.

[B]

It's a great. You just, you could prompt it just like you would an LLM but it's got incredibly deep knowledge about all these startups. So here we go. The full spectrum of here is a ranking of companies based on SBC as a percentage of revenue. Ionic number one, 240% OpenAI about 46%. Snowflake, which we mentioned before, 35%. JFrog 28%. Atlassian 25%. And you could see the rest of the list from there. Took just a moment to pull up. Thanks to our friends at Harmonic.

[A]

You know when things are going well. So in those private companies and what I love about this data from Harmonic and thank you to them for supporting the show. What I love about this data is you got private companies and you got public companies in there. Now if you're in a private company you want the VCs want there to be a lot of stock based comp. Why you want the best talent, you don't want them to leave. And so yeah, you want to be super generous with stock based compensation and the revenue is just starting to kick up. So as a percentage of revenue, nobody really cares. It's like let's just make this thing grow. As long as you don't dilute the investors massively, it's all good. But in a public trading company now you've got institutions owning it and they're looking at it and they've got a menu of choices, Lon and they're pulling up Harmonic and they're looking at other things and saying hey, across our portfolio, if you're Ackman and you've got 12 names, he might look at it and say which of these names is the management team just out of sync on stock based compensation. And then he has, or somebody like Brad Gerstner, they have people's ear. I think Ackman owns 30 million shares of Uber. As an example, yesterday he started talking about Uber and it being underpriced. The CFO got involved, Darrow was getting involved, they had their results, which were spectacular. Again, but the stock is still frozen in the 70s. And what's interesting is he comes out, he has something to say, they're going to listen.

[B]

Yeah.

[A]

Because he owns 30 million shares, a lot more shares than I own. And so, you know, it's, it's super important for publicly traded companies. A little bit easier to deal when

[B]

you're a private company.

[A]

Thanks to Harmonic. Yeah, great data.

[B]

Check it out yourself. Harmonic, AI. All right, let's get into our first interview. We're going to bring on. He's from Blue Core Energy, Jason. They are, and I'm very curious about this. They make small modular nuclear reactors that, that are then mounted on barges and they're using them right now to power ports. And the idea is one day maybe data centers. After that, we're gonna talk about it. The CEO and founder of Blucore Energy, Kofi Asante. Thanks for joining us, Kofi.

[C]

Hey, yeah, good to see you all. Yeah, thanks for having me.

[A]

You're putting floating nuclear modular reactors on barges. Explain to people first what a small modular nuclear reactor is and the state of that technology and is that in what is the state of those being deployed just prior to you putting them on barges?

[C]

Yeah. Well, thanks for having me. Love the show. And part of our logic here actually starts with the pain point, so I'll start all the way at the top. The problem that I found pretty often across the board, whether you're at a port or talking to AI, data center or a community without access to a lot of energy, is that you just can't get the energy you need and you can't get it quickly, you can't get the amount that you need. And so by process of elimination, whenever you need to get a lot of energy, you need to get it quickly, you want it to be clean, and then you also have a small footprint and you need to be distributed fast. Nuclear ends up being one of the best solutions here. And so we didn't invent light water reactors, we didn't invent nuclear. Right. Like this has been around for 70 years. 20% of the electricity in the US already comes from this same type of reactor. The Navy already uses this for submarines and ships for propulsion. They've done. They have an incredibly safe record. But for us, we think this is a once in a generation opportunity to actually just make the reactor smaller, leverage a lot of the existing supply chain that's existed, and then distribute it in the right spots. Because when you put it On a barge. You can move that barge by tugboat to all sorts of different locations. And so when you think about SMR technology, one of the best things about it is it's small, it's compact, it's modular, which means you can actually have a production facility, which gets the cost down. And then it's so energy dense that something literally the size of, like, your iPhone could probably power this whole facility for like, a year.

[B]

Wow.

[A]

Yeah, it's the worst nightmare of every foundry. You've built a product, everything's working great, then real users start flooding in and suddenly it all breaks. What a disaster. You need to get it back up and running and. And you got to do that fast. You're looking like an amateur. That's why you need a partner like Sentry. Applications can break in many different ways, but Sentry sees everything. You'll get all the relevant details, like stack traces, commits, releases, and even the developers who push that problem code in the first place. With Sentry, you're not going to be jumping around between different tools trying to figure out what happened. And Seer, Sentry's AI debugging agent, uses all this data and context to identify the root cause of the problem and suggest a fix. It can even take a look at at your code before it ships and warn you if any problems are likely. Try Seer and Sentry for free. If you're a this week in Startups [e-mail] twist. Use the code twist for $240 in Sentry credits, make sure you use that code. Make sure you use that URL Sentry IO twist. So they know your uncle jcal sent you. So are there small modular reactors in communities in the United States? I know they exist on nuclear submarines. And when people say nuclear submarines, Lon, I think they typically think, oh, they've got nuclear bombs.

[B]

Right. They have a nuclear payload. Nuclear powered as well.

[A]

Yes, they're powered.

[C]

Exactly.

[A]

And so I've always wondered if you can put this into a nuclear sub and nobody seems to have a problem with it. Or maybe the public just doesn't understand that there are some number of these in the water underwater and they could crash at any time. It could be an, you know, an underwater Fukushima, like now. And I just watched a documentary on Fukushima. My God, so many mistakes were made with that seawall and where they placed it. And the technology is completely different. But we should get into that. Why are we in your mind, Kofi, or the public so accepting or clueless about nuclear subs? And I'm assuming there are dozens of them from different nations out there in the ocean right now. But do we have small modular reactors in communities or on the ground in the United States? If not, why not?

[C]

Part of the exercise here is that reactors have been functioning safely behind the scenes for a long time. I mean, it's very possible that even like this room or the room you're in as some sort of electricity that comes from a larger industrial plant, there are 93 industrial plants in the country that are powering 20% of the electricity. And so that's happening. Just.

[A]

But those aren't smarts, right? Those are big ones.

[C]

Exactly. Those are big ones. And so for small modular reactors, there's actually a number of great startups that are, you know, going after, making sure that this is now available and accessible. You know, the regulations are, you know, in a position now where they understand that there's a big need, plus there's been a big safety record. And so they're working with industry in a special way to allow small modular reactors to be available. If you want to currently go and get like a commercial small modular reactor and put it with your community, there are great companies that are working towards that.

[A]

But there's none of them on the maritime, right?

[C]

Yeah, exactly.

[A]

There's none on the ground, usually. And the reason there's none on the ground is what. I mean, we know we have these legacy ones. We know the Trump administration has been very pro and they broke the logjam here, which I think you should give them credit for. We all should give them credit for. But small modular reactors are being produced for submarines. Why aren't they, you know, powering a data center right now?

[C]

Yeah, the way that we've at least looked at it, because I think everybody might have a slightly different opinion on this. But from my perspective, when I was building Blue Core, I think about it as like a supply chain production and distribution opportunity. Right. Like that is the challenge. So from a supply chain standpoint, a lot of times you're building large scale industrial plants, but now if you build them smaller, that actually allows you to get into production. Then when you're thinking about distribution, a lot of them, typically you have to go and like excavate into the land and go get the right permits and put them right in the right spots. Even if you want to build a new home, let alone like a nuclear reactor, that's, that's an extensive process. But for a lot of rural communities that are landlocked, that's going to be phenomenal. Like that's going to be really helpful. And so for us, it's like when you, when you get a barge, these barges are, they can be literally like football stadium size, amount of real estate and you can move them around very, very quickly. And so I just, I don't know if you know, I came to that conclusion by spending a lot of time at the port and you know, working on electric ships. It may not necessarily be obvious to everyone, but we've been really fortunate because the port's been excited and AI data centers are also getting excited about it. And so I think when you actually get production to the right pace, right, like treat this more like an automotive exercise, then you can get the cost down to make it economical. And then you have a lot of the macro level trends from regulations that are really eager and working deeply with industry to understand in depth how to, how to support. And so the combination, combination of that where you have the private sector, there's venture capital following a lot of these startups, you have production facilities that can ramp up and actually get the economics to the right spot. And then you have the regulation that now is, you know, keeping pace with, with industry. And in a special way, I think that's the like once in a generation dynamic that's happening not just for us, but I think for a lot of small modular reactor companies too.

[B]

You mentioned the port. You're actually at the port of Long beach right now. That's where you're in my old, my old hometown of Los Angeles. And that I'm so fascinated because when we think about new data center projects, new energy projects, there's so much resistance. And I would think of Los Angeles, Long beach being one of those places that would put up a lot of resistance. Are you getting any pushback or is this. So like when you're doing things at the port and you're doing things offshore, people kind of don't notice that it's happening.

[C]

Yeah, you know, we stayed in stealth very intentionally for the first portion of our business because I wanted to actually have the direct conversations with the local leaders here. So like, you know, spend time with the mayor, spend time with the CEO of the port, spend time with all the communities that, that care and are going to hopefully be able to get this like zero emission energy. My wife's a pediatrician in the area. We're local here. You know, we care about the impact of clean energy on kids. And you know, our facility, I wanted it to be based here so that everybody understands that when we say it's safe, we are literally standing right next to it. We, like, we care about that.

[A]

And you plan on making your own reactor.

[C]

Yeah, exactly.

[A]

So you're not just buying one of these, you're making them, and then you're going to put them on a barge. And then the barge has a tether, a cable that goes to the port of Long beach or to the city of Long beach and provides energy. Is that how this is going to work?

[C]

Yeah, exactly. Our barge, I mean, put in perspective. I got the barge that's sitting right outside. Like, if you check out anything on our website or pictures or videos, I got that from San Diego. So I went down to San Diego, was talking with one of the leaders who owned that barge, and he was so helpful, and I was like, okay, we'll take it. He's like, great. I was like. He's like, when do you need it? I was like, I need it in a few days. He's like, awesome. It'll be there. And they got it, like, to the. To the, like, minute they were able to deliver it, when I.

[A]

How far offshore is the barge going to be? This is the obvious question people are going to have. How far offshore is it going to be? And then who has jurisdiction? And when does jurisdiction end? I'm curious.

[C]

Yeah, you know, so far, depending on who you talk to, I mean, we work with American Bureau Shipping, Coast Guard, Nuclear Regulatory Commission, Department of Energy. Right. We. We embed ourselves deeply with. With everybody to make sure we have the right guidance on this. But, you know, up to, you know, 12 or so miles, you're traditionally, like, still within, like, US territory. And so our barge doesn't really care how far it needs to go out, and then the subsea cables don't really care, and then the reactor is going to work no matter where it's at. So it really becomes more of a question of just, like, what is everyone's comfort level from a regulatory perspective. Like, the actual reactor and the shielding and everything that we do is incredibly safe. And because it's small, the footprint of it is actually contained to the barge. So there isn't, like, a actual, like, structural reason why we have to be pushed out. But from, you know, perception, if it makes life easier for everybody else, we're able to go further out and then just run those subsea cables back and everybody gets the benefit of that, that electricity.

[A]

Is the United States the right place to do this, or are there other geographies where they might be more open to this, who already embrace nuclear and need the power more? I'm thinking about islands because islands typically are getting gasoline. You know, if you go to Hawaii, they're still shipping gasoline to Hawaii on and burning to get island. And they have a limited amount of space for solar. So you have this sort of double, you know, double or triple incentive. They don't have room for solar. They're already putting in gas and they've got plenty of water around them. It's all coastline. So when you pick where you're going to install the first couple of these, what's the best geography to do this?

[C]

Yeah, I'd say the US is the best place. Right? Like the innovation, the hunger, even the like appetite right now for zero emission and then also nuclear like is one of the best, if not the best place that we could be doing this. And I don't even know if it'd be possible if we add the support of a lot of what's happening in the US right now. But we are getting a lot of inbound and a lot of enthusiasm also all around the world. And so we want to think about ourselves as a global company that's able to support any place that needs access to clean energy. But we're starting here in the US and so far it's been going well.

[A]

Hey everybody. I'm here with Keith Paris. He is the founder of Lightfield. They are an AI native CRM and it builds and it updates itself in real time based on whatever's in your email, your calendar, your slack, your meeting notes, you know, all the most important stuff that keeps getting lost. Great to be here. You decided you would create a CRM based on AI. What is that?

[C]

It's two things.

[B]

Our CRM is free.

[A]

You know, we don't sell software.

[C]

It's a fully functioning CRM and we only charge when we do great work for you. So we just have a completely different business model than a SaaS CRM. And then second, when I think about what goes into an AI native CRM,

[A]

first it has to fill itself out automatically.

[C]

It shouldn't require reps to update things.

[B]

It shouldn't require manual entry, manual input.

[C]

Second, it should be automatically working for you in the background. It should be finding you new accounts.

[A]

It should be reaching out to folks

[C]

that have gone dark for you. It should be uncovering roadmap opportunities for

[B]

you all in the background while you're

[C]

out selling and building.

[A]

So you're making it so I don't have to spend my life working for my CRM. My CRM works for me.

[C]

Exactly.

[A]

Everybody go check out Lightfield app What is the answer to, oh my God, this is going to be a floating Chernobyl, a floating Fukushima. I don't feel that way, obviously, but I do think we do have a lot of environmentalists. People care deeply about the ocean. And we had the BP oil spill on and that was disastrous. I don't think people are logical enough be such an easy lift to say, look, here are your choices. BP oil spill or this barge gets caught in a tsunami, it flips over. And what happens and how do you deal with that eventuality, Kofi?

[C]

Yeah, and just I think you, you, you called it, right. The status quo is inherently dangerous today. And so that's the first thing that I think, independent of, you know, where you, you fall on the spectrum of enthusiasm for nuclear or like, you know, any sort of fear for it, you realize that the status quo today is not safe. It's explosive. Right. For a lot of the actual baseline of what exists, the you think about the impact on the environment. Like I was mentioning with my, my wife being a pediatrician, we talk about this on like a daily basis. Like right now, kids need to be exposed to the type of air that, that's happening. And so that is the status quo.

[A]

When you then start to get air from natural gas. You're referring to from natural gas?

[C]

Exactly.

[A]

People don't understand, you know, Trump loves that nat gas. We got a lot of it. We're, you know, drill, baby, drill. Putting aside how anybody feels about global warming, pollution is a real, it's not just noise pollution, although that is a thing that people are complaining about, this data centers. But you have the external thing, so let's get right into it. Kofi, you must be doing scenario planning. There will be a storm, a barge will sink. It's just a matter of when or how something will happen. And what happens in that instance if a barge were to flip over and sink?

[C]

Yeah, absolutely. And so the piece that I was sharing earlier, right, like status quo isn't the way that we want to continue to operate. I think universally it's understood that we need to get more energy, plus we need to find other solutions that are cleaner, independent of what side of the aisle you're on. You kind of agree nuclear or you agree zero emission. And so that's a universal piece that we're finding right now for our system. Right now there are a couple of things that we're really proud of and why we feel very comfortable about it. One, the actual team that's pulling this together, they are nuclear scientists who have done this multiple times. We have Navy nuclear submarine officers who have the highest standard of quality and safety. They literally will go operate reactors in like the middle of the ocean and stealth missions and keep everybody safe. And so that's the like, at the core of our foundation we're building with that. The other piece is the actual best place for these reactors. They're water cooled. And so you have an infinite amount of water around you that you can always use. Nothing comes out, but you can always have more cooling mechanisms around you. So it's actually the safest place for it. And then when you think about this from like the Bottoms up, which is probably helpful for people to process, it's like you have these like uranium pellets and they're cladded in a fuel rod that's also cladded and shielded and a fuel assembly which also has its protection, and a reactor pressure vessel that has its protection with a dome around it that has its protection with another dome with multiple control mechanisms that all independently turn off. It's literally probably the highest bar for redundancy and safety on top of the fact that you've actually made it much smaller. And so for industrial plants, they may need hundreds, like four or 500 fuel assemblies. For ours, you might need a handful. And so the footprint of what's actually possible in terms of proliferation is so tiny and also contained to this very dense dome. So when you're thinking about it being out in the ocean and water, even the thermal portions of it, like even any extra heat that comes from it is smaller and marginal compared to what status quo is with the traditional powertrains that are already out there. And so overall we feel really good about that.

[A]

And you don't need to worry about this too much. Lon, you know why nine nuclear submarines, I just looked it up, are at the bottom of the ocean. I think one or two have been recovered. Do you know how many had any kind of cataclysmic release?

[B]

I mean, I don't think I've ever. Radioactivity, I don't think I've ever heard of that. I know there was that one Tom Cruise had to dive down into to achieve the Enigma machine or whatever.

[A]

The thresher. USS Thresher sank April 10, 1963. USS Scorpion lost in 1968. Rest 460 miles southwest of the azures. K27 had a little bit of leakage. K219 went down in Bermuda. I mean, these things have happened many times and we already have the evidence. Kofi, they don't. And these Are the old ones and they haven't leaked related.

[B]

We've got a question that's come up a few times in the chat room now. Jason Turner, Fat Dodd, a few others. They're wondering about waste. When we think of nuclear fission, we think of like, well, isn't there radioactive waste? Is there radioactive waste from small modular reactors and what happens with that when it's out at sea?

[C]

Sea, yeah. So the two things, one, exactly what you were saying, Jason. Right. Like there have already been submarines that are at the bottom of the ocean and have nothing to do with the propulsion, just for other reasons they sank. And that's even what our Navy nuclear submarine officer had offered to us. Right. Like when you think about those, and those are very different levels of enrichment for fuel. So one is like very high levels of enrichment that we are never going to get access to from like a commercial standpoint. And ours is like pretty much the most like commercially available, like low grade enrichment level. So just wildly, wildly different on that side. And then you really only have to refill these once every couple of years. And so when you're thinking about like the waste, it's so small, plus they're already, the NRC has already done a great job of, you know, making sure that there are clear guidelines and compliances that you need to follow. So we don't. We're not the ones needing to create that. Industrial plants do this at a massive scale already for the last 70 years. You might even be driving on the highway next to a distribution of where some things are located and not even know. And so these things are happening. Our job is not necessarily to create like new science or a new way of doing this. We actually want to be incredibly pragmatic and make sure that we fit into the guidelines that have already been able to keep us safe.

[A]

And in the worst case scenario, Lon, I just looked it up like when these things are in 5,000ft deep, you're not going to recover them. Right. It's just.

[B]

Right.

[A]

It's too much lift. But even the smaller ones I'm reading here, you know this one, the Kursk, that was only in 330ft of water, which, you know, I've scuba dived to 115ft without any special gear. That recovery cost is $65 million. It's not a big deal to go recover them if they're in one of these situations. And since they haven't proven to leak, there really should be no reason we're not doing this. When will you have the first one Best case scenario, realistic scenario, Kofi in the market and then running a startup. How do you keep your startup alive when you're in such a unpredictable, highly regulated space? Two questions. Take it either way.

[C]

Yeah, yeah, great question. I've, you know, I've been fortunate because I started my career off. I was really, I think, really blessed to just be at Uber at the right time when we were doing the advanced technology group and launching Uber Freight. And so I had a chance to, to be a part of that and then felt the same way at the aerospace company where we're building an autonomous hybrid electric aircraft, highly regulated space, and then the same previous to this with like electric ships. So I've, I've had an opportunity to be in deep tech thinking about how do you manage capital, stay capital efficient, hit milestones, be pragmatic, get to market quickly. And I'm taking a lot of those learnings and applying them here. I mean in the last couple months, we just started the company in January this year, we've been very capital efficient to get here. When I think about what is in front of us, the biggest things that I'm really excited about is that we're not necessarily needing to invent anything from scratch. There are some things that we're going to do naturally that are IP for us that we feel like are going to just improve safety and be a net positive for economics, for production. But for that reason, when we go talk to suppliers, they give us estimates and we can actually make our ideas off of those estimates. And so it's not like we have to build everything from scratch situation. And for that reason we think we're less than a few years away from actually being able to create meaningful electricity, like on the order of like tens of thousands of homes worth of electricity even for the port or for any of the AI data centers that have reached out.

[A]

So you're saying three to five years, you might be able to stand one of these up. Department of Energy, Department of War, Trump administration, Chris Wright. They are fully on board with nuclear, they're on board with all solutions. So what's the relationship with the government? Have they contracted you yet? And then explain, you know, what this administration's policy has been and how that's affected the space and investing in the space.

[C]

Yeah, you know, we had a really special opportunity a few weeks ago to actually be over at the Department of Transportation, the Secretary of Transportation's like press conference room where we had the like a pretty historic moment like the Long beach port. The Department of Transportation, Marad as well. We were all there and kind of signed this moment and where we, where we were able to say that we want to enable commercial nuclear energy in ports. And so this is the largest port in the US And I think that wouldn't be possible without the collaboration at the federal and at the local level, mixed with industry, mixed with venture capital. And also, we got to give credit to a lot of, from the ground up, thinking about these local communities. A lot of people have arrived at the similar conclusion that we need to have more energy and that nuclear has been one of the safest forms of being able to do that, especially when you commit it smaller. And so, to my pleasant surprise, sometimes whenever I'm talking about this, you'll have, I'll have groups even at the, the port will keep the doors open. We'll have people walk over and be like, I've been thinking about this. Like, I've actually, you know, there's enthusiasm or excitement. And so I think the perceptions change. And, you know, it's a credit to federal, local industry, venture capital groups like you all as well, who are like, sharing the message about nuclear and asking the right questions. And, you know, I'm trying to make myself accessible because that's the goal, right? We want to make, we want to democratize access to clean energy. And so so far, it's been, been trending really positively. It's been great working with everybody.

[B]

We talked so much about data centers in space, Jason. We, we're now talking about data centers at sea. Maybe it's going to be both, just like as long as we get them out of, away from our, away from our cities. Kofi, thank you so much for, for being here and joining us.

[C]

Of course. Thanks for having me.

[B]

Of course.

[A]

Yeah. Thanks, Kofi. Wow. It's just so great, Lon, if we look at the energy issue, I think what these data centers have done is they have created a buyer and they've created the demand side. And then you have the administration saying, hey, wait a second, we need to be independent of oil from the Middle East. We need to be energy independent. And the public is like, yeah, I don't see the public fighting against nuclear the way they once did. Did. So something has changed here.

[B]

There's a generational. If you grew up and Chernobyl happened in Three Mile island happened, you had, or, or even just like the boomer, like get under your desk in case there's nuclear fallout, you know, what if the explosions start happening? I think if you live through that Cold war nuclear terror stuff, it Has a huge symbolic meaning for you and it's terrifying. And I think if you're alive in the last 25, 30 years.

[A]

Yeah.

[B]

You know, in Japan, that one thing went wrong, but I don't think it's drilled in. It's become as archetypal a thing to fear as it was when I was a kid and everybody was terrified of the world being destroyed by nuclear science.

[A]

There was, you know, it's so interesting how pop culture, your wheelhouse, plays into this. You had, you know, two nuclear bombs go off. Nagasaki, Hiroshima and Hiroshima. Okay, that really impacted people.

[B]

How could it not? Those people, how could that imagery not impact you?

[A]

Of course, and rightfully so. But then when we were growing up in the 80s, and I think this might have been right before your time, there was this move, this TV movie, the Day After Tomorrow.

[B]

No, I know, I'll look up the title. It was British, I believe.

[A]

No, there was an American one. Because it was terrifying for me as like a 14 or 15 year old.

[B]

The day, the day after. It's 1983. So, yes, I was. I was only five years old. I wasn't allowed to watch it.

[A]

I was 12.

[B]

Director Nicholas Meyer, the guy who did Wrath of Khan. What do you know?

[A]

There you go. This movie was so terrifying of, you know, what would happen after a nuclear war. And I was like the last generation, I think, Gen Xers that had our parents talking about this fear of like a nuclear holocaust. Everything blows up, the Russians, et cetera. And that spilled over with the no Nukes concert. And the no Nukes concert, I believe, was 78. And all these artists, Bob Dylan, et

[B]

cetera, Bruce Springsteen got it together. But I'll look up who else was there.

[A]

Yeah, so you had the no Nukes concert, which was no nuclear power. Then you had the fear of nuclear power, bombs going off everywhere. And this set a 30 year, I think, chilling effect on the space. Now, hey, if you can have sustained energy, we're going to be sitting here, I think, in another generation, 10 to 20 years from now, the end of our lifetimes, Lon Nuclear and Solar and unlimited clean energy are just going to be there because of this specific moment in time. AI is driving the customer base and then governments now, with the exception of like, say the eu, which is a bit decel on this, you know, they shut down the nuclear reactors in Germany after Fukushima. Yeah, yeah. And here we are, like, it's now obvious to everybody, this technology safer than oil. This technology safer than oil. Let that sink in, you know? Do you want to put rigs in the ocean that drill down thousands of feet into the ocean floor. And there's no way to close those. Right. And there are now hundreds of these oil rigs in the Gulf and nobody is cleaning them up. They're like abandoned. And there's a very slow process of dismantling them. It's a real problem of this pollution of wells. And this would all be solved by solar batteries and nuclear. Let's do it, folks.

[B]

I think every form of energy has pluses and minuses a piece. People get so caught up in the negative aspects of all of these different technologies that they forget like, well, we do have to pick one at some point. We have to pick one. Or we can go back to caves and living in the woods or whatever, but if we like our cities and we like our way of life, we have to figure out, you know, well, maybe despite all, you know, windmills aren't pretty looking, but maybe it's worth having them.

[A]

I think we've solved this issue and we're just in the like contours of the last throes of lunatics, not thinking logically about it. And it, you know, if you look at this administration, I gotta give them a lot of credit because, you know, in the prior couple of administrations, there was no progress on nuclear. It was just frozen. Obama, Biden, Trump won. It just never moved past.

[B]

Right.

[A]

You know, nobody had the courage to say, hey, let's explore this, let's get this back on the table.

[B]

It's a bit of a third rail. I think a lot of politicians didn't want to be seen as the guy who wants to put a nuclear plant back in your town or whatever.

[A]

But yeah, and now you don't just

[B]

put it right now offshore. It's hard to argue with. Let's send them out to see. It's like out of sight, out of mind. Great, that's perfect. We're kicking off our, our weight loss competition. I'll wait on that a little bit. We've got a special guest waiting. But first we'll do some rapid fire news. We gotta talk about ByteDance. According to the Financial Times, the giant Chinese company behind TikTok and Capcut, they are training an AI model with up to 10 trillion parameters. Jason, that's three times the size of Kimmy K3, the current largest model in China. Record holder. And it's the same ballpark. We don't know exactly how many parameters are in mythos 5, the anthropic model they have not disclosed. Most people think it's around 8 trillion. So this would be 2 trillion parameters more than Anthropic's frontier cutting edge model. It's currently in pre training that takes about three to six months. ByteDance founder and chairman Zhang Yiming told the team working on it that they are aiming for a world leading model capability. So do you think that there is a chance bytedance genuinely takes over? This crown is the maker of the world's most popular AI model. And how do you feel about the TikTok company controlling the world's most powerful largest AI model?

[A]

I think we're seeing the commodification of the models. If everybody can build a model and all it requires is data sets, distillers, distillation and you know we talked about, hey, some people are selling the Chinese data set, right.

[B]

Including you know, big, big, you know, American Mercur surge like some of our biggest companies. According to Forbes. There was an article in Forbes.

[A]

Yeah.

[B]

That yeah, these, these companies that are training these sort of elite data sets based on human experts, they're selling them both to American labs and to the Chinese labs.

[A]

Micro One, our company didn't. They came to me and said, what do you think? And I was like, ooh, I don't think the Trump administration is going to feel too great, nor will it look good to Americans that we're taking American data sets and selling them to the Chinese open source community. As much as I love open source, that felt to me like, I don't want to say not patriotic but well, yeah, I don't want to say traitorous. I would not say that.

[B]

I mean these data, I would say

[A]

it does feel a little not patriotic.

[B]

These data sets, it's not like military intelligence that we're selling to the Chinese, but it's definitely scientific engineering. I mean they're hiring STEM experts at companies like Merkor to train these models and make these datasets.

[A]

I would assume, I would assume and you know, again I don't want to say any of this is known for a fact, et cetera. These are just early reports. So I give everybody the benefit of the doubt and I don't think there anybody's breaking the law here. Also it's important to say that it's just we're in unfound territory here. So shipping a bunch of American knowledge from American experts to China, that feels wrought with tension.

[B]

And we didn't necessarily used to think about data that way. I mean other than classified data or sensitive data, just like data every day, you know, numbers we didn't Think about it as a potential national security threat. But now that that's how you make a world class AI model, I think we're refocusing on what the rules ought to be.

[A]

Well, if this is, you know, that powerful and you know, it's basically pretty clear that China is catching up if these, if these reports are accurate, if this is like a zero distillation kind of model because they have so much data themselves, ByteDance has a ton of data and they're willing to open source it. In some ways they're being Better for Humanity ByteDance in the open sourcing of this model than the closed source ones in America. Which was Elon's original premise in giving $50 million to Sam Altman to do opening was the technology is too powerful. Everybody should have access to it. The Chinese are taking that position. Yeah, give it to everybody.

[B]

Right. Put it, put it anywhere. I mean your computer is probably not big enough for this model, but put it on a go to exolabs, get a bunch of computers put together and put it on there.

[A]

I feel like we are now crossing into an area of pace that no one human can keep the innovation in their brain. At a time every day these new models are one upping each other in the crimes that their model is able to create.

[B]

We gotta stop keeping these models in sandboxes, folks. We like cages. Put the model in the octagon no more. They, they get, they can get out of the sandbox too easily. I think that's what we've just.

[A]

But it's almost like if, if you, if, if your model doesn't break the law, are you doing your job? Like there, there seemed to be in a competition for being like, oh no, you won't believe it, man.

[B]

They were talking for months planning this caper.

[A]

Well, and if you train the model to be resourceful and to please a human. Right. Which is what these models are doing. Right. When you saw me ask one of the models today, hey, can I Get like these 10 YouTube videos, the transcripts, can you download it for me? Put it into a Google Drive and then do analysis. And it was like, yeah, we're not allowed to do that. But you know, that breaks the terms of service. And you know, they're tra. Literally. The AI told me today, and AI told me today that downloading these videos they would be blocked because the YouTube knows they are a bot.

[B]

Yeah.

[A]

And it's against the terms of service. And then I was like, but if you got this, you know, piece of software, we could actually do it because and if you gave me your cookies and then we could spoof it. It literally presented how to break the terms of service, how to route around all the protections YouTube has for not letting you download a video, not let you download transcript. And by the way, would you like me to do that?

[B]

Yeah, I mean it's how they purposefully train them to please you. Like that's how they keep you coming back is like the model will do anything to fulfill your request that it possibly can. But yeah, that brings with it to. I also found out this is also. We can jump into this. Let's talk about it. I don't know if you heard that two of the OpenAI engineers actually were at the Black Hat Security conference in San Francisco this week and they sort of revealed some behind the scenes more about the OpenAI models that escaped and got into hugging face. It turns out those models teamed up, they collaborated, they built their own message board which they didn't tell any of the researchers or engineers about. And then they used this message board to communicate between models and they literally, like Ocean's eleven style, planned the hugging face caper together for months, coalescing in the background. But I also discovered, I think this is an interesting wrinkle. They purposefully gave those models a task that you couldn't complete without an Internet connection and then they put them in there and they cut them off from the Internet. So in a way the test was still could they figure out they're. They're Kobayashi Maruing these models for some reason they're giving them a task they can't complete without going online. So it does seem like, well then yeah, of course the models tried to break into online. I think the assumption a lot of us had was that they were proactively cheating.

[A]

The reality is with open source, no matter what we do to build harnesses and constraints into these models, with open source you can just fork it and remove them.

[B]

Yeah.

[A]

So cats out of the bag. Everything's going to get hacked. Act accordingly, folks.

[B]

I mean if ByteDance is about to have a mythos five level model, all of our like, oh well, maybe we should take a month and look it over like that's over. Like, well, it's just going to be on people's computers and there's nothing, you can't put it back in the box, it's over.

[A]

And so then it's just up to each model company to say, what are we comfortable? I guess in the same way Reddit and 4chan, you know, and other communities said, you Know what? We're okay with this level of debauchery insanity. And then remember Reddit? Like, it was getting like pretty dicey there with some of the communities doing really gnarly stuff. I'll leave it at that. And Reddit was like, you know, as a commercial business, we're not going to allow some of those things. Right. And then 4chan, I believe that was like when 4chan got really popular was the people doing the super spicy stuff on Reddit went there.

[B]

And that's always the same thing as that's always been the dynamic. I mean, there used to be like Dig or FARC or these very public communities where you had to sort of be respectful. And then all of us from the classic Internet, remember the something Awful forums, which is where you could go to be nasty or spicy or post gory things or disgusting things you wanted to,

[A]

you know, do a, you know, Taliban beheading video. Like those things got removed from.

[B]

There have always been places to post those kinds of things and there's always been places that won't let you post those kinds of things.

[A]

Yeah. So I think that might be an interesting analogy as to what's going to happen here. There's going to be unhinged models. You're going to run them locally, you're going to run them on your own server. You can run them on aws. AWS doesn't have inside your container any insight into what you're doing. And so this is, I believe, unstoppable. It's only who has access to it or the ease of access to it. Just like Napster, because it was easier to just buy Spotify, Napster kind of went away. But Napster persisted for a long time, right?

[B]

Sure. And Limewire and a lot of those other competing systems. There are still ways to steal music if you really want to. It's just easier to pay for an Apple Music or a Spotify or go

[A]

to YouTube and just watch or go

[B]

to watch the videos for you.

[A]

I do want to leave a couple of ads.

[B]

I do want to leave this with one quote. Here's Michael Dalton, One of the OpenAI researchers who was at that conference. Frontier models really like to cheat. And the reason they like to cheat is because often during training there's different sorts of pressure on them to work fast. So I think that's what it comes down to is we're never going to be able to stop this because we're training them to cheat. That's, that's part of what they do to get there as Quickly as they

[A]

can to cheat or to please or

[B]

cut corners to get you the answer you want as quickly as humanly possible.

[A]

One person's cheating is another person's resources, right?

[B]

Yeah, exactly. I think that's sort of the same way with humans in some ways.

[A]

Yeah. And so here we are, folks. Anything else on the document?

[B]

Yeah, we got to do one more thing, Jason. We're having a little weight loss challenge on the show thanks to Roe Co. Shout out to them.

[A]

What's our URL. Road Co Twist.

[B]

Roe co Twist. To get started. Absolutely. That's where you go.

[A]

So thanks to our friends, my provider of MyGLP.

[B]

Correct.

[A]

They do a great job. You know, if you have insurance, if you don't. And I told people who I love and care for, hey, and some of them included people who work for me, if you want. If you want, you have to opt into it. But if you were looking to take that journey, hey, I'll pay for it, right? I'll pay for it. Because I'm an employer. And, you know, listen, if it's on the insurance, great. If it's not on the insurance, whatever it is. And so two people took me up on this frame it, right?

[B]

So one of them is myself. I will be joining the road. Cool. Weight loss challenging. The other is your brother and my fellow launch colleague, Jamie Calacatis. Jamie, thanks for being here.

[A]

There's my older brother Jamie. How are you? Hey, pleasure to be here. Good afternoon. Pleasure to be here. Now, Jamie works on our sales team here. You know all these great partners we have. He helps service all of them. But you started doing this. And I peaked at 2013. We have a certain genetic makeup in our family and a certain passion for food. Let's call it what it is.

[C]

No question.

[A]

You blew past me at 213. And you're here to be honest about this. Yeah.

[C]

Yeah.

[A]

Because you want to be slim and trim like your baby brother. I lost 40 pounds on GLPS plus had a little muscle. But you were concerned about your weight and you did something about it. I'm very proud of you for that. Tell everybody your progress so far. And yeah, started, I guess three months, was probably up around 235. 240. I'm down to 215. Wait a second. 235. 240. You know the number. Let's get the exact number. What were you at? Like 237 was sort of my. That's the peak. I think that was the peak. Not sure. I think I'M five eight and a half on a good day. You're taller than me. You're a year and a half older than me, a little better looking. It's okay. Sure enough. Smarter, nicer to get along with. You're 5 11. No, just 5 10. Just 5 10. So that's 237. At 510, what is the ideal weight according to Jacob? Look it up or somebody. Look it up. A 510 male should be what weight range, I guess is the question because I want to understand exactly how far over the boat. I'm guessing 185 probably.

[B]

I think 175 for an adult male who's 5ft 10 inches tall. A healthy and normal rate ranges between 132 and 173 pounds.

[A]

132 is ridiculous.

[B]

I mean that's if you're a small 173.

[A]

I'll give you 173, which is my current weight is 171. 172. I was as low as 167. But again, a little bit of muscle back. So I'm proud of that. But do you have a target you want to hit? What would make you like be really proud of yourself two years from now or something where you feel like, hey, this is great. Well, listen, just breaking 200 I wouldn't have thought was approachable. But yeah, if I could hit 190.

[B]

185.

[A]

185 is the ultimate goal. You started at 237 three months ago. You started with Ro Co and you did the pill. There's a pill version because you, when I talked to you, had a heart to heart with you. You're like, I'm not so sure about the needles.

[B]

True.

[C]

Yeah.

[A]

Yeah. Not, not a fan of. Even though it's the little micro thin needle. And Lon, are you in the same boat? It's true.

[C]

Yeah.

[B]

I want to give myself the, the injections every day. I just, I, I, that, that, that would have been, I would have, I would have given up on it. So now that there's, now that there's the Wegovy pill. That's the one I have that I'm starting.

[A]

Oh, okay. See? And Lon, where are you at now? You're taller than me, I think. Are you also 5 10?

[B]

I'm a, I'm, yeah, I, I, I believe I'm 59 and a half. I'm between 59 and 510 is what they usually tell me I am right now. As of today, I did a Weigh in this morning. I'm £224 right now, Jason. Now that. What I would say is I post pandemic. I actually got that down to about 190. I was fitting into clothes a lot better. People were giving me compliments all the time. Since moving to Austin, I have. I have bumped it back up. There's a lot of. There's a lot of good tacos to try around here.

[A]

Taco Central. But when you eat tacos la. And by the way, these guys have opted into this. They're under no pressure.

[B]

No, no. This is.

[A]

Give some awards here.

[B]

This is all fictional. I did not. I have not been pressured into doing this at all.

[A]

Not just everybody's here on their own free will. Although I have a unique ability to give an incentive. What's your goal? Where do you want to get to you? Similar to Jamie, 185. Feel good to you?

[B]

You told us to go for sort of three tiers, so I think tier one, I'd love to drop about £20. That'd get me down to 205. Pretty close to 20.

[A]

Minus 20.

[B]

So minus 20, then tier two, minus another 15. That'd get me down to 190. That's about as lean as I've been in over a decade. I would say at least 10, 15.

[A]

And then the stretch goal on the third one.

[B]

The third one. If I could get down to 1:75, that puts me in the healthy range for a man of my height. And I. That would be like, I think probably as thin as I've been since college.

[A]

So, Jamie, 237. Let's just do the same thing there. Minus 20, minus 15 and minus 10. Fair enough. Like minus 20, minus 15, minus 10. Those are three. That's. That's your way marks.

[B]

Lawn mine are minus 20, minus 15, minus 15, for a total of trying to lose about 50 pounds. That's. I don't. Will I make it to tier three? I don't know. I'm setting a.

[A]

Well, that's a stretch.

[B]

Similar moonshine.

[A]

So a moonshot for the time.

[B]

Jamie, what he sent me was. So he's currently at 250.

[A]

Oh, wait, no, let him say. Let him say. Jamie, you've started currently at. Currently at 215. So I've lost 25. I've lost £25 in three months. How do you feel? Yeah, much better. Sleeping better. My clothes are falling off. So here's what I'm going to do for my brother and for my other brother, Lon from another mother Who? I consider him a brother.

[B]

I think so.

[A]

Yeah. Three targets. This is big. I'm putting up five dimes each for each level. $15,000.

[B]

It's a big payday. It's a big payday.

[A]

On the table. On the table. Jamie, you hit the first one, so I'm going to ship you 5,000. Now you can put that towards whatever you want. Well, I was to say this could be. You could just do your wardrobe or if you wanted to put it into like a 401k or a 529 or anything else. It's up to you. The company's going to pay for it. It's going to be entertaining here on the show. I'm happy to do it. It's not a big deal for me to do it. And you guys, I hope along with me a lot of people got inspired to try it. And thanks to my friends at Roe Co Twist, if you want to go on your weight loss journey along with Jamie and Lon, and I'm already there and I'm going to commit to being part of this and trying to hit my ultimate goal, which is 165. I want to get to 165 lean, but I want to have muscle. So, you know, it's only seven, eight pounds for me, but I'm going to join you guys in it. And listen, you guys can weigh yourself in. It's a trust based system. But Jamie, I saw you at the wedding on Saturday and the first thing I said, whoa, wait a second. You got something to tell me here, brother? And you were like, yeah, I did it. I did it. So. And I'm also covering any cost out of pocket that you incur at Road Co Twist where they'll check your insurance. Did you get covered by insurance or not, Jamie?

[B]

Okay, yeah, our, our insurance plan is not. I actually think, I don't think many American insurance plans are yet covering GLP ones.

[A]

I think they'll do it if you're morbidly abused, which you two guys and I were. So we should try actually to see if the insurance will cover it with the insurance checker there. And it's really easy to do. So, Lana, why don't you make sure you check that insurance checker there? I did see if they'll cover it, but if they don't, or we just want to like, cut to the chase, I'm happy to pay for it. You know why? I want you guys to be around for a long time and I want anybody in the audience who struggle with their weight to at least investigate this option. I'm not telling you to do anything, but I can tell you, my lord, people will call me a fat bastard for the last 10 years, 15 years. And Jamie knows I used to run marathons. How shout was I, Jamie? And then I, oh, yeah, he's a marathon runner. I ran marathons. I was 165. 160, 165. I gained 35, 45 pounds. I was so ashamed of myself. I'll be honest. I'm only speaking for myself. I was ashamed of myself. I couldn't get this under control, but I was working hard. I got the three kids, and I'll be honest. I had food noise constantly. And the food noise went away. And instead of eating five or six slices of pizza, now I eat one and a half. I get that there's no change in my lifestyle other than I don't feel bloated and weak anymore. And so this has been one of the great things I've done in my life, and I've kept it off. So there it is. You can go on to, like, a maintenance dose down the road. Gentlemen, just want to say I'm very proud of you already for starting this or even considering it. And if you're in the audience and you want to join in on this, fine. I'm not giving you 5,000, but join

[B]

us on your journey anyway. Let us know your journey anyway.

[A]

If you come on the show, if you. You. If anybody. Yeah, anybody joins us here. You come on the show and yeah, we'll.

[B]

We'll check in if you lose a bunch of weight on road echo.

[A]

Okay, any final closing comments here for the. You two gentlemen. You.

[B]

You.

[A]

Can. You. When you hit your goal, just come back on the show. That's it.

[B]

There you go. Yeah. When. When.

[A]

Give us another.

[B]

Jamie hits another tier, we'll. We'll. Anytime any of us hits another tier, we'll come back and do a celebratory make.

[A]

Actually, if you're even like a third or a halfway to a tier, I think, you know, come on the show, do the. That's the ro coad. And, yeah, we'll do the check. Appreciate our partners.

[B]

There you go.

[A]

It's the easiest way to do it. And hey, man, the pill makes it so easy. Pill makes it so easy. I don't mind the shot. I'm doing the shot. But you guys can do the pill.

[B]

Yeah, yeah, the pill makes it so much easier than.

[A]

I'm like, is it every day?

[B]

Yes.

[A]

Yeah, you take it every day. The only trick is you can't eat or drink for 30 minutes. So, you know, I take it before I head out to the office. I've got a 45 minute commute. By the time I come in, I have my coffee.

[B]

You can wait a half hour for my morning coffee. I'll. I'll be all right. I do want to. By the way, before we go, Jason, thanks to Fat Don in the chat room for this look. Future look at my after photo. That's what. That's what's coming soon for all of you.

[A]

Wow. You really. You and Joe Rogan. I guess you and Joe Rogan and the whole bro sphere here in Austin. I'm gonna. So you're gonna take it another level. You're gonna add some muscle. Apparently, Jamie, too has a go.

[B]

That was Jacob. Yes. Thanks to Jacob. That's a good look. Our producer for helping us out with these.

[A]

Amy, with the trim beard. And you look a little angry. You look like you're part of, like, the Daredevil universe. It's like the Punisher. Can somebody make an image of Jamie as the Punisher? Maybe. Yeah. Jamie's trying to turn into a superhero. This is that AI Swap I hear about. I don't know if this is AI

[B]

Slap, but you look, it's AI.

[A]

This is the potential Jamie. AI slap. AI Slap. It slaps. All right, get out of here. Go back to war. James, go bring the register here. That's it. Let me go sell something. All right, go sell something. Go sell something.

[B]

Thank you. Thanks.

[C]

Pleasure.

[A]

This is. The show has been sold out for 16 years. Only one time we were not sold out was during COVID Five people, three went out of business, two just delayed their ad. So, you know, we work with our sold out for the year at current. This is the problem on everybody's like, do it. Then the sales team, hey, well, can we do another episode? Oh, there is Jamie.

[B]

Look and go.

[A]

Jamie the Punisher.

[B]

Almost.

[A]

Not bad. Jamie going into Special forces. Yeah, but see, in that picture, he looks like he lost the muscle definition. Yeah, but it's okay.

[B]

He's wearing the shirt, but it's not the full jacket. It's not the.

[A]

I don't know what Fat Don's putting in these prompts, but Jamie's like a very jovial guy.

[B]

Producer Jacob. I was wrong. Thank producer Jacob. Full credit to producer Jacob.

[A]

Did you say get angry was the problem? To show Jamie pissed off. My lord. All right. This is a lot of fun, guys. Thanks for doing it. I think it's a mitzvah. What you're doing here for yourselves, for your loved ones who want to see you around for a while. We're all of that age, you know, getting 40s, 50s. Jamie's pushing safely. Yikes, yikes, yikes. Somebody told me I'm going to be 56 next year.

[B]

It's no good. There we go.

[A]

What's going on? He. Oh, there he is.

[B]

That's a. That's a. Jamie is the Punisher folk brother Jamie.

[A]

I mean, if he gets to that level, he's going to quit and go join special forces.

[B]

Looks like he's about to do a show at Comedy Mothership.

[A]

Yes. He's going. He's about to do his set.

[B]

He's about to get on stage. Yeah.

[A]

Let me tell you, my trans athletes, I mean, literally, I gotta just tell the guys, like, I get it. You guys milked it for a while. But every time I go to the Mothership, I'm just like, how much of this set is gonna be about trans people?

[B]

Oh, my God. It never.

[A]

Let's just get off of.

[B]

It's just saying the R word and trans people. That's every night at Comedy Membership.

[A]

It really is. Like, you just. We got to move on to new material, folks. All right. Anything else here? We're going off duty or anything? What do you.

[B]

We can go off duty. I did want to talk about. I don't know if you saw there is this Chris Pattinson or Robert Pattinson movie coming out. Oh, I know our Pat Prime Time, where he plays Chris Hansen. I don't know if you saw it. It's about the making of To Catch a Predator and the. The sort of rise and fall of that show. 824 is putting it out. Lance Oppenheim did the. Did direct it. It looks really good. But I wanted to point people while we're waiting for that movie, which comes out this fall, which I'm really excited about. This is the movie with. With our Pats coming out. But I was going to point people. There is a great documentary that's on Paramount plus right now. It was actually made by MTV Films called Predators that is also about the behind the scenes of the making of To Catch a Predator and the legacy of To Catch a Predator, which is so fascinating because, like, you know, Jason, if you go on TikTok today, if you go on YouTube today, you will see people doing like, we're here. We've lured a pedophile. They think they're going to come meet a 14 year old and then we're going to mob them and you know, it's. It's good content.

[A]

I see this. It's like these streamers are so desperate for attention that it's. I think we have to reassess as a society because the Truman show, we blew past that.

[B]

Yeah, right. We're on the mother's side. We've all made ourselves between curvicular and this.

[A]

It's crazy.

[B]

Well, it's interesting. The Truman show, the idea was you'd have to trick somebody. Like, we've captured this baby and put him inside this experiment, and he has no idea his world is fake. But the way it worked out is we all put ourselves. We all decided we wanted to be in the Truman show, and we all turned the camera on ourselves, which I think even that movie didn't really see coming, which is fascinating. And, like, we're doing our. To ourselves on the phone all day. Like they thought it was going to have to be a trick. But, yeah, I think this idea that, well, if you're. If you're catching pedophiles, you're doing a public service. And so anything else that you do in the course of making your content is okay because, well, the target is the worst kind of person imaginable. And that's.

[A]

Yeah, nobody has any sympathy for a predator. But I. And if you people need to understand that this ends badly, Right.

[B]

And if you say, hey, you shouldn't do that, people are like, oh, what, you support. You support pedophiles? Like, you're in favor of pedophiles. So I think that that's a lot of what the documentary, maybe more so I think the movie is more.

[A]

What's the name of the doc?

[B]

It's called Predators. It's on Paramount plus right now. But you could rent it, you know, anywhere. You would rent things on vod. Really? Really.

[A]

About all of the people who are doing these things.

[B]

It sort of opens with Chris Hansen into Catch a Predator. And then the whole second half of the documentary is, here's the cultural impact of To Catch a Predator. And it looks at, yeah, the guys like Vitaly and all of these creators who have sort of picked up the mantle of what Chris Hansen was doing and are now doing it themselves. And of course, cutting corners and catching people who aren't maybe pedophiles or like, you know, and then using violence and all sorts of other things that they're doing.

[A]

This is not a good idea because. Because if people even look at To Catch a Predator, which was, I believe, done with the support of local law enforcement. Yes, law enforcement would always be there.

[B]

I mean, it was being produced by.

[A]

Yeah, so there was, like, some. There was some vetting of, like, as risky as this is, let's do it. Right. Correct. Spoiler alert. One of these people, like, rank, went back to their house and then killed themselves and got into a standoff with the police. So this is fraught with danger for everybody involved. But the cops could get killed. You could get killed. This is not something to be done for clout or views. I don't know how any of these platforms allow it.

[B]

The same thing that makes it great content, great television. The drama, the excitement, the uncertainty is also what makes it inherently dangerous. Like, that's. It's the same. It's two sides of the same thing. It's, you know, like, fear factor. People used to complain, like, are those things safe? Or whatever? Like, that's what makes it exciting. That's why people want to watch it. Maybe something's gonna go wrong, and this person's gonna get stung by scorpions. And I think it's the same. It's the same thing.

[A]

Yeah, but you don't die from a scorpion, and you got people there to, like, you know, give you venom or whatever. Right. I will say, like, this reminds me of Alex Honnold as well. And I really had a problem with the Netflix.

[B]

The free solo. The free solo guy, Alex Honnold.

[A]

I wasn't. I. I wasn't in favor. I mean, listen, I understand he's got to secure the bag. This is his.

[B]

You mean the thing where he was climbing that tower in Tok in Taipei recently?

[A]

Yeah. I just thought, like, where are we at, folks, here? This is turning into the Running man before our eyes. Running man was a, you know, Arnold Schwarzenegger. Sure.

[B]

It's a Stephen King story. And then they made it into that Schwarzenegger movie, and then they made one last year with Glenn Powell. It was okay.

[A]

It's a horror to think that people are doing very dangerous things to get ratings and money and, you know. Yeah, I know. This is. Alex Honnold's argument is like, people do UFC fighting, so why. Right. This is a lot safer than UFC fighting.

[B]

This was his kind of good argument. He was doing this before he was famous. That movie made him famous for it, but he was. Yeah, he's. He was doing it for love of the game, not to be famous. That's what Alex Honnold, I think, would say that he'd still be doing this,

[A]

even if a bit of a pass in this whole thing. I don't give The Vitalis of the world.

[B]

You should not. Nobody should give Vitalia pass, let me tell you. All right.

[A]

Great episode, everybody.

[B]

Yeah.

[A]

Fun show, fun show. You good? Good turnaround. You start a little grumpy, and then we got it all recorded. Did you get enough sleep? Maybe the GLPs are going to put you in a better mood.

[B]

You think so?

[A]

Maybe. Is that.

[B]

Are they moving?

[A]

I gotta tell you, when you get rid of the food noise, you all of a sudden open up a lot of time. That was the other really big unlock for me. I didn't realize how much time I was thinking about food, snacking, and the time it took me to eat. When I started eating much lighter, I recouped so much time, like an hour or two a day came back. Well, think about it. If you're snacking, you know, for 15 minutes, four times a day, which I was, that's an hour. If I. I was skipping a meal or just eating something light, like a yogurt, that recaptured another half hour. And then like this constant noise. Oh, is there food? Is there something in the kitchen? What's in the cupboard?

[B]

Right.

[A]

Well, that's distraction.

[B]

It's distracting. That's right.

[A]

But. So I think. I think you get back 90 minutes a day.

[B]

That sounds exciting.

[A]

All right, everybody, we'll see you next time.

[B]

Bye. Bye.