Is Anthropic Making the Biggest Mistake in AI History? | E2258

Summary

This episode tackles one of the most controversial AI stories of the moment: Anthropic’s defiance of the federal government, which Erik Voorhees of VeniceAI characterizes as bravery rather than mistake-making. The discussion features Erik Voorhees and Logan Allin of Fin Capital debating open vs. closed source AI, privacy in an age of omnipresent AI, and where agentic AI meets the financial system.

Logan Allin shares insights on how Fin Capital uses personality types and AI to assess investments, and reveals that LAUNCH’s token usage is trending higher than payroll. The episode also features two demos: David Kaufman shows off Siteline, which helps companies understand their agent traffic, and George Pickett demonstrates OpenClaw Studio, designed to make OpenClaw as user-friendly as possible. The conversation wraps with discussions about MicroStrategy’s layered business model, which AI tools are most indispensable, and Jason’s discovery of the Grok button on threads.

Highlights

Why Erik Thinks Anthropic’s Defiance Is Bravery

“Why Erik thinks Anthropic’s defiance of the federal government is bravery!” — Jason Calacanis, 29:50

yt-dlp --download-sections "*29:50-31:33" "https://www.youtube.com/watch?v=uegOFJffAU0" --force-keyframes-at-cuts --merge-output-format mp4 -o "anthropic-defiance-bravery.mp4"

Open vs Closed Source AI

“Logan’s thoughts on open vs. closed source AI and what it means for the future.” — Logan Allin, 11:11

yt-dlp --download-sections "*11:11-14:19" "https://www.youtube.com/watch?v=uegOFJffAU0" --force-keyframes-at-cuts --merge-output-format mp4 -o "open-vs-closed-source-ai.mp4"

How to Maintain Privacy in AI

“How to maintain privacy in a world with omnipresent AI.” — Erik Voorhees, 14:19

yt-dlp --download-sections "*14:19-19:24" "https://www.youtube.com/watch?v=uegOFJffAU0" --force-keyframes-at-cuts --merge-output-format mp4 -o "privacy-in-ai-world.mp4"

AI Personality-Based Investment Assessment

“How Fin Capital uses personality types and AI to assess investments.” — Logan Allin, 24:28

yt-dlp --download-sections "*24:28-29:50" "https://www.youtube.com/watch?v=uegOFJffAU0" --force-keyframes-at-cuts --merge-output-format mp4 -o "fin-capital-ai-investment.mp4"

OpenClaw Studio Demo

“OpenClaw Studio makes OpenClaw as user-friendly as possible.” — George Pickett, 56:48

yt-dlp --download-sections "*56:48-1:09:14" "https://www.youtube.com/watch?v=uegOFJffAU0" --force-keyframes-at-cuts --merge-output-format mp4 -o "openclaw-studio-demo.mp4"

Key Points

• Logan’s Circle thesis (4:01) - Logan Allin’s original investment thesis in Circle and crypto infrastructure
• Open vs. closed source (11:11) - Debate on the merits of open and closed source AI models
• Privacy in AI (14:19) - Erik Voorhees on maintaining privacy with omnipresent AI
• LAUNCH token usage (21:42) - Token spending trending higher than payroll at Jason’s fund
• AI personality assessment (24:28) - Fin Capital uses personality types and AI to evaluate investments
• Anthropic’s defiance (29:50) - Erik characterizes Anthropic’s stance against the government as bravery
• Siteline demo (42:03) - David Kaufman shows how Siteline helps companies understand agent traffic
• OpenClaw Studio demo (56:48) - George Pickett makes OpenClaw more user-friendly
• MicroStrategy layers (1:09:14) - Analysis of MicroStrategy’s complex business model
• Most indispensable AI tools (1:11:06) - Discussion of which AI tools have become essential
• Grok button on threads (1:14:15) - Jason’s discovery and use of the Grok AI button on social media threads

Mentions

Companies

• Anthropic (29:50) - AI company defying federal government, debated as brave or mistaken
• VeniceAI (0:00) - Erik Voorhees’ privacy-focused AI company
• Fin Capital (4:01) - Logan Allin’s fintech-focused investment firm
• Siteline (42:03) - David Kaufman’s company for understanding agent traffic
• OpenClaw Studio (56:48) - George Pickett’s user-friendly OpenClaw interface
• MicroStrategy (1:09:14) - Bitcoin-focused company with layered business model
• Circle (4:01) - Crypto infrastructure company Logan invested in early
• Grok/xAI (1:14:15) - Elon Musk’s AI, integrated into social media

Products & Technologies

• OpenClaw (56:48) - Open-source AI agent framework
• Grok (1:14:15) - xAI’s AI assistant, available via button on threads

People

• Erik Voorhees (0:00) - Founder of VeniceAI, crypto and privacy advocate
• Logan Allin (4:01) - Managing Partner of Fin Capital
• David Kaufman (42:03) - Founder of Siteline
• George Pickett (56:48) - Creator of OpenClaw Studio

Surprising Quotes

“Why Erik thinks Anthropic’s defiance of the federal government is bravery, not a mistake.” — Erik Voorhees, 29:50

“LAUNCH’s token usage is trending higher than payroll — we’re spending more on AI than on people.” — Jason Calacanis, 21:42

“The layers and layers of MicroStrategy’s business model are like a Russian nesting doll of financial engineering.” — Jason Calacanis, 1:09:14

Transcript

Jason Calacanis: 0:00 All right everybody, welcome back to Twist. It’s March 4th, 2024.

Lon Harris: 0:05 Yeah, wow, I know. Still feels weird. Still sounds weird.

Jason Calacanis: 0:09 But it is AO 39. 39. 39 days after we first started talking about OpenCloud here. It’s been a full one month and 1/3, and it is still sweeping through the internet. It’s become the number one, number one open source project in history, based on people following it, you know, stars, likes.

Lon Harris: 0:30 Yeah, GitHub. GitHub stars exceeded React now, the language.

Jason Calacanis: 0:35 So we have a full docket for you today. Lots of demos, lots of discourse, the investment opportunity, the efficiency opportunity, the entrepreneurial opportunity, all of that. And when the world is this chaotic, Lon, you know what I like to do? I like to go over here to my Plaud pin.

Jason Calacanis: 2:43 We’ve got two guests. Well first joining us from Fin Capital, he’s the founder and managing partner, give it up for Logan Allin joining us. Also an investor in Circle.

Logan Allin: 2:54 Welcome back, Jason.

Jason Calacanis: 2:56 And I’m told here by the docket, crypto, not the community app. So just keep it straight everybody. Not the community… Circle, Jeremy Allaire’s company.

Lon Harris: 3:02 Correct. The stablecoin interest, not the SO community one.

Jason Calacanis: 3:07 I’ve known him for so long. I just interviewed him at Davos. I don’t know if you caught the interview I did with him, but I did an All-In interview with him. And stablecoins, you were so early on that. How’s that as an investment going?

Logan Allin: 3:17 It has been good. I mean, obviously, the company went public. It’s back over, above 100. You’ve got more clarity on the legal framework, right? That was a big catalyst for the stock, which is banks and TradFi actually being able to adopt stables. We’ve got some noise around how stables and deposits are actually going to work together. And you’re seeing that play out on socials now, vis-a-vis the Clarity Act. And hopefully that comes to fruition here in the near term, albeit the country’s got a lot on the plate right now. So we’re big fans of stablecoin, big fans of Jeremy Allaire, and think there’s a lot to come in terms of treasury money movement, yield, and so forth in that world.

Jason Calacanis: 3:54 When did you invest and what was your original thesis?

Logan Allin: 3:57 So we actually invested, Fin invested from pre-seed to pre-IPO. So on this one, we got a little late to the party. We were pre-IPO and then picked up some secondary and were very aggressive about our secondary purchasing. And the thesis was back in the post-FTX Circle was being SPAC’d world, and we were saying, look, we think stablecoins are the way money will move digitally. Obviously, we didn’t have agentic commerce or anything along those lines, but we felt like digital money was going to be important to that future as well. But the baseline thesis was TradFi’s going to adopt this. We’re going to need to move money cheaper, faster, better. Everything from credit card settlements to sending wires and remittances are going to be far more efficient on these rails. And that was our thesis. And you know, we view that as having played out but still early innings.

Jason Calacanis: 4:51 You know my thesis, Lon?

Lon Harris: 4:53 What’s that?

Jason Calacanis: 4:54 Look to what the criminals are doing. Do they adopt the technology? And we did have that with other stablecoins that were offshore.

Logan Allin: 4:59 I think you’re talking about Tether.

Jason Calacanis: 5:01 Well, I don’t want to dig anybody, but it was, you know, human trafficking, allegedly, some of these things, allegedly.

Lon Harris: 5:08 A lot of that dark web stuff you hear about, yeah.

Jason Calacanis: 5:11 Then you have the next group of people, which is maybe they’re discrete, not illegal, but they’re moving money around for discrete things like wagering. Right. So when it came on my radar was after, you know, the Tether truthers going wild on social media, Coffeezilla, all this stuff, tracking it, was when I saw people trying to settle, you know, wagering on sports or poker in this. And then you start to know it becomes legit when people who do a lot of transactions and are actually looking at the number of basis points, etc., start adopting it. Which is what happened with Venmo, Cash App, PayPal. People who cared about the 5%, 3%, whatever Visa was charging. People who are sensitive to that. Your gardener, your handyman, whatever it happens to be.

Lon Harris: 6:00 They’re like, ‘Hey, can you do me a favor, send it on PayPal friends and family or Venmo friends and family?’

George Pickett: 6:04 Yeah, so I don’t have to pay fees.

Lon Harris: 6:06 That’s when you know something’s starting to get there. And stablecoins have just started to hit that, and I think we’ll start to see when your gardener or your handyman or handy…

Jason Calacanis: 6:14 Person.

Lon Harris: 6:15 Handyperson. I mean, it sounds so weird. That’s when you know it’s going to be right there.

Jason Calacanis: 6:19 And I think we’re right at that moment in time. Let’s introduce our next guest.

Lon Harris: 6:23 All right, up next we’ve got Erik Voorhees. He’s the founder of Venice AI. Now they’ve been in the news a lot recently, Jason. They’re an AI chat platform that doesn’t log conversations, no safety restrictions, no censorship guardrails, but also briefly they were the sort of default model being used by OpenCola. OpenCola has since sort of gone back and now they don’t want to play favorites, they’re sort of neutral Switzerland, but for a short time, Venice AI was the default OpenCola model.

Jason Calacanis: 6:53 Welcome to the program. Erik, what’s the vision here for what you’re doing, and are you one of these like libertarian privacy nuts who just is making a point? Like, and I love these guys. This is my favorite, you know, this is my favorite, what are the things between superheroes and supervillains? There’s like the…

David Kaufman: 7:13 Anti-heroes?

Lon Harris: 7:14 The folks between, they’re sort of not anti-heroes but they kind of like Ahsoka, they’re not Jedi, they’re not Sith, they kind of chart their own path.

Logan Allin: 7:19 Right. Maverick.

Jason Calacanis: 7:21 I find the libertarians who don’t want to be part of the two different parties, or Ahsoka who doesn’t want to be a Jedi but uses… is a force wielder. I feel like, Erik, you might fall into this, which is chaotic good. Yeah?

Erik Voorhees: 7:34 Yeah, I think that’s fair. But I, first, I’m just curious how the stablecoin investments are doing so well. Like, I must be doing something wrong because all of mine have stayed at a dollar.

Jason Calacanis: 7:46 Heyo! Heyo! We were talking about investing in the corporate company but you are correct, they are pinned to a dollar, and that’s what they’re supposed to do. He’s going to be here three more nights, so take care of your waitresses.

Logan Allin: 7:58 I will say this was also very confusing to me in the early stablecoin era. I was like, ‘Wait, how do you make anything? It’s just… it’s almost a dollar.’

Jason Calacanis: 8:05 It’s the idea. Tell me who you are and why you’re doing what you’re doing. I want to get to first principles here, yeah.

Erik Voorhees: 8:11 I am of the crypto world. I am currently a tourist in the AI world. And I’ve been building Venice for the last two years. Venice essentially is meant to be like a private, uncensored version of ChatGPT. So when I started it, I realized that like the principles that I thought were important from the crypto world, namely like user sovereignty, the right to privacy, free speech, lack of censorship, these were entirely absent in AI. And realized that I should be the change that I wish to see in the world, and so started Venice.

Jason Calacanis: 8:46 How did you wind up getting so early into OpenCola? I’m curious.

Erik Voorhees: 8:50 I mean, we’re an AI company, so like we pay attention. And we’re an open source, you know, like we love open source companies and projects and so we were aware of it very early. Um, we made a few pull requests. Quest? Like right when it came out. And Venice was added as a model provider within the flow of setup. So basically if you want your Claude to run on private inference and not have like every detail about your life and business going to Sam Altman, then we were a good alternative. So that’s how we ended up there. And we in the company we’ve been huge users of Claudes and it’s been quite an amazing phenomenon. I love that you’ve been covering it so much.

Jason Calacanis: 10:53 Logan, what do you think about this open-source movement? You know you had so many people trying to capture the agent opportunity, the co-worker opportunity, and then you have just coming around the bend from the outside this dark horse, this open Claude project and it is the one that took the crown while the king was looking down the jester stole the thorny crown. Like this is it, right? Like how did this happen? And what impact do you think it has on the marketplace as investors?

Logan Allin: 11:26 Yeah, so we actually passed on the proprietary LLMs which I’ll call a, you know, anti-portfolio at this point. Uh, but our view was that open source was going to meaningfully catch up and close the gap around intelligence, speed, and price, right? And you’ve seen compression and the closing that gap, uh, in particular with Claude as well as, you know, players like a Gemini, like a Meta, etc. investing heavily into providing these capabilities out into the market. And then you’ve got all of the China-based players as well emerging in our… Our view is that now you’re seeing Claude and certainly OpenAI start to produce at the app layer. They always said, ‘look, we’re going to be the chassis. You guys go build on top of it and we’ll stick to our knitting.’ Now you’re seeing them need to monetize on application and suddenly, whenever Claude announces an application, it, you know, quote-unquote ‘destroys’ or cooks some company. And I think that’s obviously overborne and our view is that the AI-first companies that are producing revenue through a product, less so on the OpEx reduction, I think that’s more short-lived and more transient. Like you have to be AI-first, you have to be producing revenue through that product, and it has to have domain differentiation, workflow and data integration in order to be sticky and to have moats around it, otherwise you’re very quickly going to get out-competed.

Jason Calacanis: 12:52 Yeah. And the fact that small organizations are adopting a technology, that to me is always one of my early signals when people who are tinkerers, hackers, DIY people, Erik, they start tinkering, I take notice. I get curious. And then when I see startups adopt a technology, I get really curious. And then when both do it, I become obsessed. Or I should say, maybe the third factor is when I start tinkering, I get obsessed. Those three circles have now happened for me. So Erik, how do you think about open source, this open-source project, and then also crypto and privacy, how is this all going to come together and what are your circles when you’re evaluating something might break out, your early signaling, your gut?

Erik Voorhees: 13:44 Like these tools are obviously so powerful. They are, they are magical in many ways. And yet people have been led into this paradigm where like everything they think and do is being absorbed into several central repositories where it will live forever. That to me is just so obviously dangerous. And I find it kind of ironic that many of these frontier labs when they say they care so much about AI safety are kind of missing like the most near and present danger, which is the, you know, mass surveillance of all humanity. To me that seems like a danger and so yeah, I’m glad that there are alternatives.

Jason Calacanis: 14:22 Erik, I am shocked right now. You’re telling me that these entrepreneurs running the largest and fastest growing startups in history are saying one thing, but doing another. The thing they’re saying is they care about safety. But then they are doing things that are not safe and that are the basic blocking and tackling, which is: there’s no need to take my queries and ingest them into your learning system if I don’t want them, and I certainly don’t want you keeping my history and giving it to the police. You could easily have a button or a checkbox, Logan, that makes this very easy. for consumers to not be involved, and there are in fact companies like Brave, like DuckDuckGo, NordVPN, other VPNs, open-source projects like the Tor Project that are designed from the bottom up to be privacy first. So, what’s your take there on is this a missed opportunity for those large language models to incorporate this kind of thinking, that Erik…

Logan Allin: 15:21 I think so. I mean, I was just learning about Venice in coming into this discussion and I think what Eric’s doing is super interesting and obviously, you know, Tim Cook and Apple have highlighted this in terms of, you know, we want to maintain privacy and we want users to have control over their data. I think, you know, largely privacy is very hard to come by these days and I think we should all assume our data is everywhere. But I do think, first principles-wise, building with that in mind, Jason, is absolutely right, particularly if you’re going to touch end consumers. We’re more enterprise-driven, and so we’re looking at companies that are helping encrypt, manage PII, obviously manage users’ use of AI. Our view is that this crossing-the-chasm moment for the enterprise and adoption of AI actually is going to be driven predominantly by cybersecurity. And so we announced an investment recently with Ashton and Sound Ventures in Witness AI, which is focusing on how do employees use AI in terms of tools we give them, but also employees bringing AI to work. Everybody would talk about, you know, bring your device to work type dynamics. Now, people are bringing their own AI to work, and that has huge implications for what proprietary data is being added, what documents are getting added, and so forth. And until you can control that experience in the enterprise, you’re going to continue to have meaningful lack of adoption.

Lon Harris: 16:01 Erik, I want to jump in here and ask you about your—you have a token. Venice AI has a token, dollar sign VVV. So, explain to us like how the token interacts with the AI product and like how people are using the token sort of as credits on Venice AI.

Erik Voorhees: 16:42 I’m a crypto guy, so of course there’s going to be a token involved.

Lon Harris: 16:50 Naturally.

Erik Voorhees: 16:52 In our case, the token allows you to get access to the API for free. So basically, if you want, if you’re a human or an agent and you want marginally free inference, like free at the margin where you’re not paying per call, you can hold the token for that. So it’s a different pricing model, and we think like whenever you can reduce friction, you find interesting new use cases, and so that was the impetus behind it.

Lon Harris: 17:27 And there’s some degree of like volatility. Like I’m reading here from our notes, like after it was made the designated model for OpenBlock, your token surged 35%. So are you—is that, are you keeping that in mind? Did you guys—are running this AI company, like what’s the price on the token and how that’s—is that impacting decision-making day to day?

Erik Voorhees: 17:46 Being in crypto since 2011, like if I cared too much about the price movements, I probably would have like had to shoot myself.

Lon Harris: 17:52 Yeah, fair enough.

Erik Voorhees: 17:54 So you’ve got to like, you know, keep a little, keep a little distance and not get distracted by the noise, and there’s a lot of noise. Ultimately, what we’re trying to do is like build a… Super useful platform for humans and agents to interact with all the leading AI models in a more private way. And the token needs to be able to support and make that mission more feasible.

Jason Calacanis: 18:11 And is it using the infrastructure of BitTensor? I’m not sure if you mentioned that. But this BitTensor and Tau projects I think are super fascinating. I don’t know if you guys are keeping up on them, but I invested and helped co-found this fund that Mark Jeffries is running that is exploring all these subnets. And the idea is to do distributed computing and provide you know an LLM or resources or storage, whatever it happens to be, and then have a flywheel way to make money. Almost like, I think it’s kind of like the next Bitcoin in this way because it has an application tied to it. Have you been following it at all Erik?

Erik Voorhees: 18:52 Uh yeah, I’m generally familiar with BitTensor. We don’t have anything to do with BitTensor. We are not a decentralized protocol. We are a centralized company that offers inference both API and app for normal humans. Um so yeah like our our token is meant to enable people to access inference for free and uh if that is appealing to you then it’s worth looking at.

Jason Calacanis: 19:04 Okay, explain the economics of how tokens for free. Because you had me at free. You had my attention at free.

Erik Voorhees: 20:47 Um, yeah. Basically, we have a subtoken called DM. DMs are created out of VVV and DMs, each one that you have gives you $1 per day of credit. So it’s like a renewing resource. You can think of it like a perpetuity. So you can value DM as a perpetuity.

Logan Allin: 21:00 on inference. And, uh, if you hold these you get a dollar per per day of credit in the in the web or the API. Uh, and, uh, Ben is is willing to offer that even though it’s a pure liability to us because we can control the supply of those. So you can think of it um, you can think of it as like a a mechanism that we we can influence and we can, uh, constrain, but anyone that has these things it’s very valuable to them. And it’s been interesting to see like how the world thinks such a thing should be priced.

Jason Calacanis: 21:31 That’s fascinating. I have, uh, hit a wall, uh, or a big challenge, which is we were trending towards with our open Claude, about $500 a day in tokens.

George Pickett: 21:45 This is somewhat my fault, I apologize. I’ve been teaching GAF a lot about YouTube descriptions.

Jason Calacanis: 21:48 Yeah, and it… because it wants to please you Open Claude, Logan, it will not use the most efficient query. It will do more than it’s asked of, uh, and then we got Kimmi up and running on my Mac Studio and we have one in the cloud and we’re starting to play with that, but it’s not very good at complex tasks. Sometimes it, you know, bottoms out, but it does some things very well. So like and that’s what we’re trying to figure out is when to route it. We may start intelligently routing thing saying this is a Kimmi task, this is a Claude task by putting like a keyword at the end of a sentence. Like use Kimmi or use KKK. Oh, don’t use that. Use KI-

Lon Harris: 22:29 Pick a different one.

Erik Voorhees: 22:31 I know you’ve got VVV, but that one doesn’t work.

Jason Calacanis: 22:35 Yeah, you know what, that one doesn’t work. Let’s keep that one off the table. Um, but I I think that might be the future is trying to intelligently route jobs. Uh, but I’m super fascinated by any way to get these costs down while we wait for the next step function down. But I am living Logan in Jevons paradox now because I couldn’t get anybody to use our enterprise ChatGPT. They were like barely using it. Then I get Open Claude and I’m pointing it to our Claude Opus and I can’t get people to stop using it. This is like a very strange thing as a business owner. It was trending to be and being more than our payroll.

Logan Allin: 23:22 Wow.

Jason Calacanis: 23:24 Over to you, Logan.

Logan Allin: 23:25 I’m not sure what that says about what you pay people, but we’ll leave that as a problem. I mean, market! Market!

Jason Calacanis: 23:29 Yeah, exactly.

Logan Allin: 23:30 Um, I think so for us, as, as we’ve shared on, on past shows, we built a platform called lighthouse.ai and, and that has been our operating system. And I think as we all know, the tools for GPs generally in the market are great. And so you kind of have to build it and now the experimentation with Open Claude, we use Claude as a firm universally, it’s pretty incredible. And there’s been some posts on X, um, and I saw Alexis Ohanian retweet this this morning, and I totally agree with this, which is, um, the future The future of venture firms is very senior, it’s almost like the old school Benchmark model, right? Bunch of very senior partners who have deep relationships and can really manage through and then a huge, hundreds of agents that you’re, that you’re operating and those are looking for stealth founders, those are monitoring your current portfolio from a financial metric and team growth, etc. perspective, those are looking at new themes, doing deep research, etc. Like, there’s just so many permutations of that and so I just think venture firms are going to continue to get leaner and probably more effective overall. And we built a Gen AI model we call Founder DNA score which was a mix of research I’d done at Stanford with Ilya Strebulaev and it’s so highly predictive now of early stage founders in the model we built and the back-testing we’ve done, that we’ve come to really rely on it from an investment committee perspective, particularly in our early stage deals. Obviously, less signal as a company’s get mature for more of the late stage work we do, but we’re really excited and been adopting AI since 2018, when it was back in predictive and we named a head of AI in 2021. And now you’re seeing I think, you know, largely every VC and PE firm name a head of AI, but I’d like to think we were one of the first.

Jason Calacanis: 25:13 What is in that formula? What are the variables? And how are they weighted? I’m assuming this is like open research at Stanford?

Logan Allin: 25:25 It was. So Ilya Strebulaev who wrote the Venture Mindset is now more well known for that book, but back then he was doing research and highly recommend following him on LinkedIn. He puts out a lot of studies on is there correlation between where you went to school, where you’re building, the degrees you have and so forth. So we actually use about 40 weighted criteria and we’ve adjusted those weights using Gen AI and putting more and more data through that, which is by the way where a huge amount of our expense comes from in just training this model and then deploying these agents. And I would say for us the big factors are repeat founder, average age of our CEOs is 45. If you were repeat founder, were you CEO, how long were you CEO?

Jason Calacanis: 26:08 Ah, whoa, that’s an interesting one I’ve never heard. So did you stay in the CEO slot is predictive of future success? So not removed is another way of saying it.

Logan Allin: 26:19 Yes. Yes, did you stay in the slot? I mean if you, if you stayed in the spot in the seat and then handed the baton to somebody to take it public or, you know, you transitioned to another role there might be some signal there but generally speaking if you were in the CEO seat and you exited and this is the other big one, right? What was your track record on the exit? And there’s actually, you would think that the best CEOs are the ones that had the biggest prior outcomes, it’s actually just the fact that you had an exit whatsoever is highly predictive of future success and the ones that had smaller exits or were aqua-hired who are still hungry and want to build for more and have more to prove are actually more successful in our models in the future. So, we’ve done a lot of this work historically and continue to run data sets like Y Combinator, CB Insights, Stack Ranks, you know, fintechs and crypto companies and software companies, and we run all that data through and then we do backtesting on our own founders. We’ve also integrated personality tests, which I think is really interesting and you can get AI-driven personality archetypes now off of just public data.

Erik Voorhees: 27:27 ENTJ.

Logan Allin: 27:28 Right, exactly, ENTJ.

Jason Calacanis: 27:29 Every CEO, Erik, wants to be ENTJ. It’s not the only model. There’s INTJ as well. You got the Zacks, you got the Sergeys, the Larrys are INTJs and you know some people Bill Gates, Steve Jobs, more ENTJ, a little bit more extroverted in their approach. One thing I want to get both of your takes on is what happened with Anthropic’s revenue ramp. This is unbelievable. A bunch of information leaked. Lon, give it to us on this leaked tweet.

Lon Harris: 27:58 Yeah, well Bloomberg, it’s from a Bloomberg article that was published on Tuesday and then a lot of people are now tweeting about it. Anthropic on track to generate annual revenue of almost 2 billion. That’s a projection based on their current performance, more than doubling their run rate from just late last year. They recently surpassed 1.9 billion in run rate revenue, up from 900 million at the end of 2023 and roughly 1.4 billion just a few weeks ago, Jason. That’s a sign of how strong the adoption has been for coding tools like Claude Code and the other Anthropic projects, and it’s also a sign that maybe this Pentagon deal is sort of fueling growth from the public side.

Jason Calacanis: 28:46 Absolutely. Any public battle for an emerging technology company drives consumer adoption, business adoption. I think there’s two things mixed in here, Erik: OpenAI and the battle with the government. And then there’s, you know, some number of folks who are like, why are we even supporting what’s going on with OpenAI versus Anthropic? We should go with the peacenik one, the one that doesn’t want to create murder bots. What’s your take on that whole beef back and forth? And then Logan, I’ll go to you in terms of the revenue ramp.

Logan Allin: 29:23 I mean first the revenue ramp clearly happened before the controversy, so it’ll be curious or interesting to see what happens after it. I wonder if the incredible growth of that revenue is what gave Dario a little more confidence in taking the position that he took. Personally, I found it brave. It’s very rare that any company is willing to do anything against what the federal government says. Like I’ve come from a world where we continually battle regulatory agencies and it’s a total nightmare and most companies just comply.

Erik Voorhees: 30:00 just comply. Easier to bend the knee. You always bend the knee, your shareholders want you to bend the knee, like you always comply with the government because the government’s right and that’s how order works and it’s a bunch of bullshit and so whenever anyone is willing to stand up to it, um, it’s inspiring frankly. So I, I liked seeing that and um, I’m curious like what effect it’ll actually have on, on their business, you know, good or bad.

Jason Calacanis: 30:25 See Eric, this is such a good insight because if Dario had shown up for all those Trump CEO meetings, if Dario had given five million dollars to the Trump PAC if like the, like I think Greg did, uh, Greg Brockman gave five million to Trump’s thing. Uh, hugely unpopular I’m sure inside of OpenAI to do that. Um, even Tim Cook went and gave a gold bar and came to the Melania, uh, screening, which I think was more painful probably. You know, you have to play the game on the field is the common thinking and I heard Logan on CNBC or Fox News, one of the two, an investor in Anthropic said hey listen, Dario’s got to play ball, like you have to win the deal, play the game on the field. I think this is a lesson for him as CEO. Your take Logan on taking a principled stand and winning the crowd, right? Win the crowd, win your freedom. Gladiator style, he got the, he got the crowd to use the product. They became number one in the App Store versus bend the knee. Where do you stand on this Logan?

Logan Allin: 32:47 I think some of the thinking behind the negotiation was a little flawed in the sense that if you’re going to work with the military, you know, and and they’re out killing bad guys, you probably have You have to accept that your product’s going to be used to go kill bad guys. So like going into that and saying, well, actually, you know, we want you to use it in the following ways is a bit odd. I think Keith on X used the analogy of that’s like giving, you know, the Department of War a Mac and telling them how they have to use it, right? I think that that was a little short-sighted in that regard. They could have been probably more balanced. But yeah, short-term upswing. And I think it goes back to the productization of these apps on top of Claude, Claude Code being number one for them. I read somewhere that that’s about 2.5 billion of the current revenue, right? So Claude Code’s the fastest growing product, but every new product and app that they spin up is generating stickiness, and it’s because of their enterprise-first orientation. And that enterprise, those enterprise relationships are over 80% of their revenue today from the numbers we’ve seen, and OpenAI, that number is less than 5%, right? And so I think all those things are incredibly valuable. And then we talked about it earlier, which is the usage-based pricing model, right? Going to value-based pricing is such a huge aspect of the asymmetric growth, but it’s been astonishing but I think well-deserved overall.

Erik Voorhees: 33:49 I got to push back on Logan a little bit here.

Jason Calacanis: 33:50 Do it. Yeah. This is an important debate.

Erik Voorhees: 33:51 And yeah, I mean Anthropic wasn’t trying to tell the military how they can go kill bad guys. That’s a bit disingenuous. They had two prerequisites, two terms in their terms of use that were relevant here. One was that they did not want their technology used for autonomous weapons, meaning weapons where there was no human in the loop. Murder bots. Yeah, call them what you want, like autonomous weapons, because they don’t know that their technology is safe enough to use in that kind of capacity yet. I think that’s an eminently responsible way to treat things. And two, their other requirement was you can’t use it on wholesale surveillance of Americans. Basically all they’re saying is like, we want you to abide by the Fourth Amendment. That was not acceptable to the military. And so the military not only is like canceling the contract, but vilifying them and saying that other companies are not allowed to work with them either. It’s that latter part, I think, that was quite gross. If the military doesn’t want to work with them on those terms, that’s perfectly fine. But yeah.

Jason Calacanis: 35:22 I think it’s a good nuance, Erik. You’re 100% correct that the framing was as a provider of technology, this is new technology. These are not batteries that have been around for 50 years and we know the circumstance under which they overheat and you shouldn’t have them, you know, under the airplane in storage and you should follow these precautions with an electric car. This was, hey, we’ve invented this new battery, this new computer, and we as the creators of it don’t believe it’s ready for prime time. We believe that it will accidentally…

David Kaufman: 36:00 murder someone. Second, hey, we live in the United States, we hear about the Fourth Amendment, we don’t want this being used internally in the United States to create a mass surveillance state because this technology can do something that the camera at the 7-Eleven can’t because it doesn’t have a brain and it’s not connected to every other camera on planet Earth. There’s something between I think your two positions, which a reasonable group, I think you’re both right, and a reasonable administration would have said, let’s handle this behind closed doors. But instead, this is a performative group of individuals and their core belief inside this administration is to never apologize, never be reasonable, and in fact attack, attack, attack. So when somebody doesn’t get in the alignment line, correct me if I’m wrong, Lon, the operating system of Trump, as outlined in that movie about him, the documentary, was you’re supposed to attack.

Lon Harris: 36:59 Yes. And Dario is the one internet CEO of that scale company, I believe perhaps the only one that has not bent the knee.

Logan Allin: 37:12 Yeah, I think a lot of it actually comes down to approach and tone more than the actual lines in the deal or the contract. I don’t know if you guys saw, Washington Post quoted a Pentagon official who said they proposed to Dario during this meeting a hypothetical. They said, would Anthropic let the military use Claude to shoot down a nuclear armed intercontinental ballistic missile aimed at the US? And apparently, according to this Pentagon source, Dario’s answer was, well, you could call us and we’d work it out. Which they did not like. So to me it feels like if that answer had been 10% more political, we might not be in this situation.

Lon Harris: 37:48 Logan, in all honesty, there was only one answer to would you save American lives from a nuclear holocaust. The answer is yes, give it a shot.

Logan Allin: 37:59 Yeah, absolutely. That was a self-fail.

David Kaufman: 38:04 Or he could have said, if that’s technology you want to work on, I think we should sit down and actually build it for you because we want to make sure it works when it does happen because we all don’t want to die in a fiery nuclear ball. But what do you think, Logan, if you were an investor in the company and Dario said what could I do better?

Logan Allin: 38:18 I think Dario has tended to politicize this a bit and has set a little bit of a trap for himself, which is we want AI to be safe, I’m concerned about what AI is doing to society and it’s going to eliminate jobs and all these things. And he’s kind of talking against his book in order to support his book in some ways and he got caught in a little bit of a trap here in the sense that he tried to politicize in a conversation where frankly, as you said, these guys are not going to stand for that and they’re going to then showcase why they’re right and they’re going to pound their chest and I think he walked right into that unfortunately to the detriment of the company from a long-term perspective because as Palantir’s demonstrated, the government could be their largest customer.

Jason Calacanis: 39:00 Hmm. Erik, you want to, uh, give us your final thought on this? And Lon, I’m going to let… I see… I see the steam coming out of Lon there. He’s got… I think you have one thing you want to add, so Erik then Lon, then we’ll move on to our next topic. This is a fascinating one.

Erik Voorhees: 39:13 Yeah, I’ll… I’ll just say that the whole thing is interesting. I think there’s going to be more of these conversations and I’m curious… I’m curious what Anthropic would have done if Kamala was in charge. Um, I’m not… I’m not confident… I’m not confident that they would have had the same resistance if their people were running things. Um, so I don’t know that it’s a matter of principle on which this was done, maybe, hopefully. But um, in any case, it’s… it’s refreshing to see business leaders stand up to the federal government in any capacity ever. I think the… the United States needs much more of that.

Lon Harris: 39:48 One of the great things about America is that we can have major debates and still come together as Americans in good faith. It’s one of the things that’s been lost in our current political climate where these two sides never think, “Hey, maybe we could go have dinner together as rivals, uh, enjoy a nice steak, and then, uh, you know, smoke a cigar, have a scotch, debate it all, and then come to what’s in the best interest of all Americans.”

Jason Calacanis: 40:18 Yeah. I think we… we kind of had that… there was this moment… like I feel like that was Joe Biden’s sort of 2020 pitch was like “Hey folks, let’s… let’s stop all… all the fighting, let’s come together, let’s all be friends.” And… and we just… we’ve now fully rejected that. We’re like “No, we didn’t… we didn’t actually want that, we don’t like that. We like… we like Donald Trump’s uh, winner-take-all style.” And I think America sort of turned away from that and it’s… it’s a little bit of a shame. Should Anthropic be forced to comply when choosing to save the US? That’s an interesting question.

Lon Harris: 40:46 I don’t think that they should be forced to comply unless we’re under attack, in which case the War Powers thing comes in and, you know, if you’re Ford, you’re making tanks now, not Model Ts. It’s just the way it is.

Erik Voorhees: 40:56 Right. Yeah, I mean… that is… they’re sort of a very classic American idea.

Logan Allin: 40:58 Obviously business should not be forced to comply. This isn’t Soviet Russia.

Lon Harris: 41:04 Except after Pearl Harbor or Hitler invades Paris or Poland.

Erik Voorhees: 41:11 Fair enough.

Lon Harris: 41:12 Should we… should we take a look at this Polymarket, Jason?

Jason Calacanis: 41:14 Oh, I love a good Polymarket.

Lon Harris: 41:15 Speaking of things I don’t think we should do… um, Polymarket is asking will the CEO of Anthropic, Dario Amodei, be arrested? Uh, only 5% chance. Uh, so the smart money apparently still on not being arrested for saying the government can’t use his technology to make killbots or spy on all of us.

Jason Calacanis: 41:35 I know we have some demos in the wings, too. You want to see? We’ve got two demos. You want to take a look at these?

George Pickett: 41:40 Logan does, ‘cause Logan and I want to place some bets. That’s what we do for a living.

Jason Calacanis: 41:43 All right, let’s bring them up. We’re… we’re yum-yumskies.

George Pickett: 41:46 We like to place bets, Logan and I. And if you’ve got something good, we’re in.

Lon Harris: 41:49 Who knows? We’ve got two eager founders. First up, we’re going to bring up David Kaufman. He’s the founder of Sightline. This is a Launch portfolio, Jason, so you did already… you’ve already thrown a little… a little slice.

Jason Calacanis: 41:59 Oh, I got my slice.

Lon Harris: 42:00 Part of LF4…

Jason Calacanis: 42:00 Sightline tracks agent behavior on websites. It’s sort of like Google Analytics for your agents. So David, welcome to the program.

David Kaufman: 42:08 Yeah, thanks guys. Happy to be here.

Jason Calacanis: 42:10 David, show us what you’re working on.

David Kaufman: 42:12 Yeah, of course. As mentioned, we are Sightline, we are an analytics tool focused on agent visits coming to your site and how to capitalize these, what they’re doing, where they’re getting blocked. As you all know, people are searching more with agents themselves or in agentic ways, Chat-GPT, obviously, it does searches in query fan outs, but often times it will go visit sites in real time. And the reality is, most websites are just not prepared for the agentic web. They’ve set up fancy UI that, you know, is nice for humans, but often times will get in the way, and so these agents get blocked, they get confused, maybe they’re spinning up and trying to look at screenshots, but the reality is most people don’t even see this traffic because a lot of them run server side only, and so they won’t show up in traditional analytics tools. And so our goal is to give people visibility on the agents that are coming to their site, and then figure out how to capitalize on it, figure out what they’re looking for, what they’re interested in. Right now, a lot of what agents are doing is reporting back to the humans that delegated them, but we feel that over time this will shift and become more and more autonomous, and they’ll start taking actions, they’ll submit demo requests, they’ll start doing bookings, there’s agentic commerce in the wings, and so we want businesses to be prepared for this new customer source and be able to, you know, present themselves in the best light and drive growth from it.

Jason Calacanis: 43:42 Logan, your honest opinion as an investor.

Logan Allin: 43:45 Yeah, so where, who is the ICP here? Like who are you finding that you’re selling into predominantly and how, what channels are you using?

David Kaufman: 43:52 Yeah, so at the moment a lot of the ICP, it’s mostly SaaS companies and often times a more technical profile. So we work with some authentication tools, B2B SaaS. Really this is because they, their customers, so procurement internally, for example, are already adopting agentic search in a lot of ways. So when you’re evaluating bunch of options for your company, it’s more than just Google, it’s, you know, customized as far as what is your tech stack, what are your requirements, what you’ve been doing in the past. And we found that resonates the best just because their customers are already there, and they’re a bit more ambitious and ahead of the game on thinking about this new paradigm because it is fairly new as far as doing growth marketing not only for humans, but also for agents.

Logan Allin: 44:41 Super interesting. Yeah, I was reading the other day that a massive part of Meta’s revenue these days is AI monetization driven, and they’re at a sixty plus billion dollar run rate on revenue as a result. And it feels like, you know, how do I understand my revenue? Attribution and how to understand the analytics from that growth marketing is is incredibly important particularly as you as you grow monetization there is that is that kind of what you guys are finding?

David Kaufman: 45:08 Yeah, and I think one unique challenge to this new world is that attribution question is even tougher and there’s a lot of SEO GEO tools out there that kind of say like, hey, are you on ChatGPT? Are you appearing in an answer? But it’s really hard to track that fully through let’s say the purchase funnel because often there’s people that are using ChatGPT to do research procurement, but then the ultimate last step is to come in and navigate with Google and still come through that way. So we’re trying to get a good feel on what parts of your site are being visited. There’s different types of agents too. Some that are directly triggered by humans and then you can do the whole funnel from I’m being ingested, I’m being cited as an answer, I’m being mentioned in the answer and then finally humans come on through and I can actually, you know, tie this back to new customers or orders or what have you.

Logan Allin: 46:03 That’s awesome. Yeah, I mean I think agentic search and agentic commerce are the future. So having analytics around that makes a ton of sense. I really like what you’re building.

George Pickett: 46:08 This is fantastic. I… I’ve already sent it to my… Claude actually to implement and leverage. We absolutely need this. So… yeah, we’ve already ingested it.

Jason Calacanis: 46:25 What do you guys think of Erik… I don’t know if you’ve seen this but our friends over at Cloudflare, and I talked to the CEO when I was in Davos about it, they have a really interesting concept which is pay-per-crawl. If you want to come in instead of just robots dot txt, so Erik and David and Logan will go in that order, what do you think of the pay-per-crawl concept?

Erik Voorhees: 46:48 Obvious and inevitable. That’s very smart.

Jason Calacanis: 46:51 David, your thoughts on the pay-per-crawl, I’m sure you’re watching Cloudflare closely on this?

David Kaufman: 46:55 Yeah, watching Cloudflare closely on this and actually we integrate with Cloudflare in order to see the visits. I think it makes sense from, you know, a publisher’s standpoint, author’s standpoint, someone that needs to capitalize and monetize their IP. From a business standpoint, you know, it’s kind of the opposite. They’re telling us we want to encourage as many bots visits as possible, we’ll even pay to have more of them. because they want as much exposure as as they can and they want to be ingested, they want to be recommended etc. So I think it depends on the context. Obviously if you’re a journalist, I think there’s some upside there on protecting the IP and what you’re producing, but from a business standpoint they kind of want to open the floodgates.

Logan Allin: 47:34 And I prefer more conversion orientation so like, you know, we’re seeing agentic in the checkout now, right? So pay-per-crawl, I think you’re gonna have it different levels of that. So, you know, did you land on the page, how did you interact, how long did you spend? And and it feels like at least with Sipeline, David correct me if I’m wrong, what you’re trying to build is attribution and and kind of removing the black box orientation around that so people can actually monetize against those types of metrics.

Jason Calacanis: 47:58 I’ve got a really simple solution here Lon. This is… What if I could set my crawl to be… this website is crawlable however you damn please for $500 a year, put in your credit card, do what you will. And if you were a Substack and you had a thousand people or even a hundred people pay 500, it could be material. This could just be a very simple solution. I know this because Microsoft went to Harper Business, which produced AngelTheBook.com, my book, and they said, “Would you accept to me,” $2500, my agent called, “$2500 for the next five years or something, 500 a year, to put it into whatever Microsoft is up to with, you know, having a book in there?”

Lon Harris: 48:28 Sure.

Jason Calacanis: 48:29 And I was like, “If you sure… and are there ten other people who would pay 2500?” This is inspiring to write another book because it’s two swings at bat, Lon. One, I’m included in the crawl and maybe even featured, so it gets me prominence. Two, if they link to the upsell, great. And if I got a little revenue stream coming in here, maybe it’s time for two more books.

Lon Harris: 48:59 Right.

Jason Calacanis: 49:00 Yeah, I think it makes a lot of sense.

Lon Harris: 49:02 I mean, there are a lot of these platforms and companies and even authors or media sites where they feel like their data is the value of the company and if they sell it at any price and just let everybody crawl their stuff… like, that’s why Reddit, I feel like, is so hostile to all of our agents because the meat that’s on Reddit is what makes it valuable and if we can all just crawl it unlimited. But for a lot of different kinds of resources, I think that makes total sense.

Jason Calacanis: 49:36 For David, I’m going to be hosting Launch Festival again. Launch Festival is a conference I created because, Lon, I couldn’t afford to come to conferences as a founder. And they wanted me to pay $20,000 to demo back in the day to show my product. I couldn’t afford that either. So I said, “If I ever make it, I’m going to create Launch Festival.”

Lon Harris: 49:38 I remember.

Jason Calacanis: 49:54 And it was originally called TechCrunch 50, and then Mike and I broke up. I went to do Launch Fest. I’m starting it up again Monday the 16th and 17th of March, two weeks. 400 seats. It’s like 50 investors, 250 founders. And I have 50 seats for sale for people who want to come to lunch, thousand bucks, and that underwrites the founders who are coming. And Logan, I would love if you would come. Are you going to be in San Francisco, are you in the Bay Area?

Logan Allin: 50:19 I am going to be in. I was just going to say, I’m in town, so I would love to join.

Jason Calacanis: 50:25 Can I get you to judge a panel with me and give some feedback to founders?

Logan Allin: 50:29 That’d be fantastic. Looking forward to that.

Jason Calacanis: 50:32 I appreciate you, I appreciate you for doing that. I know time is limited. And if anybody wants to go, opencall@launch.co if you want to be involved or just on the website, you can figure out. It’s free for founders, by the way. The way I do it, Logan, this is my brilliant move. It’s 20 bucks for a founder to reserve their seat. When they pick up their badge, we give them the 20 bucks back. Then we don’t have, you know, drive-bys. It just creates a… just enough friction. David, are you coming? Are you in the Bay Area?

David Kaufman: 50:54 I’d love to. I’m currently in Telluride, but I am…

Jason Calacanis: 51:00 Oh, Telluride. Please.

David Kaufman: 51:00 Lucky you. Telluride, great skiing. First time here.

Jason Calacanis: 51:05 Ever tell you the story when I almost died going to Telluride airport? This is I almost went on the do-not-fly list. Wow. This is before I had kids. I’m on a little puddle jumper from Los Angeles. They pull up the Telluride airport video landing there. Most dangerous airport in America. You can find it on YouTube to my team. Telluride airport, dangerous. This thing is on the Telluride airport, David, you’ve flown into it, yeah?

David Kaufman: 51:26 Flown into Montrose, but yeah, I’m familiar with it, yeah.

Jason Calacanis: 51:30 This is the worst possible airport you ever go to and it’s on a- it’s in a box canyon type situation and, um, when you land at this thing, it’s got a runway that goes up and down. Why? Because it’s so short you want to slow down. It’s also windswept and I’m coming in on this in a prop plane, but it’s a commercial, but one of those small like Eagle kind of regional jets. And as you can see, this is like, um, a little peninsula type situation on three sides is a cliff. We come in, there’s crosswinds, the plane goes sideways down the runway. I think if this thing hits the tires hit, we’re flipping and rolling, we’re dead. We miss the runway lines.

Lon Harris: 52:14 Entirely?

Jason Calacanis: 52:16 We miss it.

Lon Harris: 52:17 It’s like an action movie.

Jason Calacanis: 52:21 And so, but he’s already powered down the plane, so we fly off the end of the runway. We haven’t touched it and we start falling down into the town of Telluride. He powers the thing up. This plane starts shaking like crazy. We finally go back up in the air. Um, and I kid you not, um, we go back up. The guy says on the thing, the plane is bouncing around. He says ‘I’m going to try one more time’. Yeah. Everybody on the plane goes crazy. I tell the flight attendant. I’s like ‘Listen, you try this again, the pilot’s going to get beat up by all these passengers.’ They say ‘Hey, we’re going to Durango.’ All of a sudden, 10 minutes later, we’re going to land in Durango. We go to Durango. My wife said ‘did you threaten them?’ I said ‘no, no, I just warned them like listen, you’re going to have a mutiny come down out of here.’ We land. Two police officers get in, they block the aisle, they let the- they let the- the pilots off the plane who don’t make contact with the passengers. They leave. Boom, they’re gone. They don’t want to confront the passengers. And then I’m like, oh my god, we’re going to get arrested. I’m going to get arrested.

Lon Harris: 53:24 Yeah, you threatened a mutiny.

Jason Calacanis: 53:26 They just let everybody off. We had to drive an hour and a half in, but David, please, for the love of God, will you promise me you’ll never go into that airport because I’m an investor in the cap table here and I need returns. We need DPI, right Logan?

Logan Allin: 53:35 Key man risk and DPI, yes, both are important.

George Pickett: 53:38 The only worse- I’ve flown into that airport, didn’t have as bad of an experience. Only worse landing in the entire world in my opinion is St. Barts. Um, and that is brutal. Yeah.

Erik Voorhees: 53:46 Wow, that is spicy.

George Pickett: 53:48 Yeah, and by the way, the plane is like 30 years older than whatever you were in.

Erik Voorhees: 53:54 Is that the one where it goes right over the beach and like it’s terrifying? Yeah, I think I’ve seen videos.

Jason Calacanis: 54:00 Oh my God. You have to nosedive in. Is basically what it is. I know a lot of billionaires who go there, and I had one billionaire who, you know, company goes public this 20 years ago, makes a bunch of money. He’s a cowboy. He’s a pilot. All this kind of nonsense. Here we go. And he starts talking about like, this landing and how intense it is. And he says, yeah, you know, my new plan is, I was talking to my friend in the CIA who’s now on my security, and they are able to open the doors on a G650 and you can parachute out. They figured out how to do this on a private plane and they slow down. And so I’m going to start skydiving to my, to my Christmas vacation. Oh my God. Never. And I said to him, I said, hey buddy, I know that you’re a billionaire now, and I know you have unlimited resources. Let me tell you about Eagle Computer and toxic wealth. Eagle Computer goes public in the 80s. On the way to the IPO party, the guy who’s the CEO buys a Ferrari. He flips it, kills himself and the CFO, I believe, tragically. Wow. And now they sit everybody down. You can look up Eagle Computer on Wikipedia and the CEO dying. And ever since then, in Silicon Valley, they sit people down when they make their money and they say, we know you can buy all the toys now. Toxic wealth. Don’t kill yourself. You’ve got kids.

Lon Harris: 55:11 Yeah, he died in an auto crash hours after his company offered stock to the public for the first time, making him a multi-millionaire. He was only 39.

Jason Calacanis: 55:18 Do you guys know this story? Does anybody but me know this story?

Logan Allin: 55:21 I’ve never heard - I’ve never heard of this.

Erik Voorhees: 55:22 You’ve heard this one, Logan? Eric? I’ve heard that story because I grew up in the Valley in the 90s, and there are other unfortunate examples about it. As a result, like, GPs, depending on the LP base you have, will literally have keyman clauses that prevent you from doing things. Like an NBA contract, you can’t skydive, you can’t go pick up, play pickup basketball. You know, GPs have similar, albeit, you know, for different reasons, can’t get on a helicopter, can’t go skydiving, can’t do any of these things.

Logan Allin: 55:54 No scuba.

Erik Voorhees: 55:56 Scuba is not for you. Yeah, so I agree with it. I’m highly risk-averse. Pickleball is about as extreme as it gets for me.

Jason Calacanis: 56:01 Oof, some twisted ankles in that game. All right, David, this is great. What are you in, Launch Fund 4? I don’t know, what did I do? Did I put in like a little 25, 25k check? What did I do? I can’t remember.

David Kaufman: 56:13 125 at full launch. Yeah.

Jason Calacanis: 56:16 Oh, okay. Great. I need to get in on this. Maybe we need to do another round that me and you and Logan can take this offline. All right, let’s drop David off. Great job. Thanks, David. Great job. We’ve got one more demo. We’ve got to welcome George Pickett. He’s the founder of OpenClaw Studio. They make OpenClaw user-friendly for non-technical users. We’ve heard from a few of these kinds of concepts. OpenClaw can be, as I’ve recently experienced, very intimidating when you’re brand new, if you are non-technical. And so OpenClaw Studio is one of these projects that’s kind of easing people in, getting them started with agents without having to do all of the complex setup. He was also a core maintainer of OpenClaw. Met Peter Steinberger, self-taught programmer who helped bring us OpenClaw in the first place. So thank you, George, and welcome to the program.

George Pickett: 56:59 Awesome. Thanks for having me. And great to meet you, the world’s greatest moderator right here.

Jason Calacanis: 57:03 Oh, that’s not for me to say, but you know. It’s for George to say.

George Pickett: 57:08 It is for George to say. Thank you, George.

Jason Calacanis: 57:11 Show us what you built.

George Pickett: 57:13 Sure thing. So one difference with what I’m building here is that in its current iteration, you actually have to have Open Claw installed on your computer. So this is for people who already have Open Claw. Basically, you can run NPX Open Claw Studio with your existing Open Claw setup and you get a really nice UI to view your Open Claw. We’re turning it into a SaaS product. I’m working with another one of the contributors from Open Claw. And yeah, basically the idea here is that this platform provides guardrails, it’s opinionated, and it does some version control behind the scenes. But basically, yeah, my thesis is that basically all of the benefits of AI agents have been accumulating to technical people and software engineers, so I want to make it really easy to chat with your agents and have some controls around them.

Jason Calacanis: 58:13 Very nice. So explain to us what we’re seeing here. This is the interface we’re seeing. You’ve got 10 agents, Bill Nye, SF Events, Open Claw use cases, latest AI news, growth engine, YouTube channel, all these things that might be interesting. Now, they’re all agents? Each one of those is an agent running on Open Claw?

George Pickett: 58:31 Yeah, exactly. So each of these is a separate agent. Usually, this would be like a separate Telegram chat for any individual agent. But what you can do here is each of these has their own separate memory. You can have different tools enabled, different permissions enabled. And so in the, yeah, like I tend to have different projects inside of different agents and scope them like that. And inside of the settings here, you can set up the capabilities, so you can make it, you can kind of control the permissions like when it’s going to ask for your approval or not. You can configure skills on an agent-by-agent basis. You can set up cron jobs in like a pretty nice little UI here. So instead of chatting with your agent and telling it, ‘Hey, every morning at 7 a.m. remind me to do this,’ or whatever you want, your repeating, I call it time automation, you can kind of set it up in a really nice, easy-to-use kind of like more understandable way.

Erik Voorhees: 59:29 So this is like, this is like Wix, Wix for Open Claw, right George? You’re making it easier for all of us non-techies, non-coders to build. I guess where’s the, where’s the use case for our being PolyMarket, because that would be like my first, that would be my first use case.

Logan Allin: 59:44 PolyMarket Sharp!

Jason Calacanis: 59:47 PolyMarket Sharp! I was just thinking the same thing because you know, Logan, so many people are doing repetitive work that if you did the PolyMarket Sharp and you said to me… If I have one agent doing this and all we do is focus on the PolyMarket sharp, uh, I’d be like, that’s worth $99 for me to throw into my, you know, agent cluster. Uh, so many different ways to win here, but yeah.

Logan Allin: 1:00:15 Totally. Well, if I gave you my PolyMarket strategy, uh, you know, you guys would be taking my arbitrage, so I’m not going to give you that part.

Jason Calacanis: 1:00:22 Totally fair. But your YouTube optimization you might be okay with. What do you think of, you know, all the work that has to be done on an open source project to make it more accessible?

Erik Voorhees: 1:00:33 Uh, well, yeah, it’s cool to see this George, good to meet you. I actually had already come across this project and I had literally told my Claude to install it and set it up. Uh, the Claude failed and I hadn’t been able to use it, but I don’t think it was the fault. The Claude has failed in many ways, so I have to have a stern talk with it, uh, later this afternoon. But, uh, yeah, cool to see this, uh, this demoed. Uh, obviously useful. I mean, like, multi-agent orchestration is the whole point and value of these softwares to make it a little bit easier, or, you know, materially easier here, I think, to give you some visualization on the agents and to personify them a bit. Very smart, um, and, uh, I’m now renewed with the initiative to, uh, make sure the Claude will get this thing installed so I can try it.

George Pickett: 1:01:26 Totally. Yeah, and I mean the way I see it going is that you probably have in the tens of millions of engineers who are using agents. I think in the next two years, we’re going to see probably a billion people on earth using agents. But I really don’t think it’s going to be WhatsApp, Telegram, it’s going to be some kind of interface. And I think if the answer, if the answer was Perplexity Computer or Claude co-work, you would see an explosion of hype. But no one is using co-work, no one is using Perplexity Computer. They’re confusing, they’re coming from this way of interacting with agents from this really developer mindset, you know, they’re showing you tool calls. I do in the current version allow you to see tool calls just because I’m a nerd and I like to see like, okay, what’s it, what’s it actually doing? But in the SaaS version of this, we’re going to, we’re getting rid of tool calls, thinking traces. Um, we have a, a really cool continual learning angle where it’s going to be learning more about you and retraining. Um, which I’m really excited about. I’m working with someone else who’s a core contributor of Open Claw who so far doesn’t want to be named, but he’s just an awesome, awesome back-end DevOps guy and has some really cool ideas around this continual learning stuff.

Jason Calacanis: 1:02:37 George, is this a, uh, is this a startup or are you doing an open source project? What’s your aspiration here?

George Pickett: 1:02:43 Yeah, totally. So right now it’s, it’s totally open source. This project is just Open Claw Studio on GitHub. Um, I’ve been working on it for about five weeks. Um, 1.4 thousand stars, 1.4k stars.

Jason Calacanis: 1:02:56 Nice. Great start.

George Pickett: 1:02:58 Yeah, yeah, no, it’s, it’s been going well. Uh, mostly all through Twitter. …and I’ve gone since I started working on this gone from like 1300 followers on Twitter to almost 5k, it’s been super awesome. But yeah, we’re working on basically a lot of these OpenCloz startups are, you know, they’re wrapping the deployment for you and handling that so you don’t have to deal with API keys, you know, virtual machines, all this stuff. I think that’s table stakes so there’s no reason why why someone should have to work like, you know, deal with AWS. But I think combining that with a really nice UI where you don’t have to think about tokens, you can log in with Google, um just providing a really simple way and we’re we’re not calling it OpenCloz, the SaaS product is like it’s it’s totally separate from the brand of OpenCloz.

Jason Calacanis: 1:03:46 Love it. Logan, your best advice for open source projects, tinkerers who want to then build a SaaS product and build a company, what should they be thinking about?

Logan Allin: 1:03:54 Who you’re building for, George. Is this something where you’re going to individual consumers and saying hey, you know, we want this to be open source and available to all of you or you trying to back into like a Dropbox style, let’s go to the developers who are then going to, you know, be in these side of these enterprises, tell everybody inside of the enterprise and and we’re going to back in effectively into enterprise sales? Is that kind of your mindset from an impact perspective? I think having a view of who you’re building for, how you’re going to monetize and and position, and then the third piece is like, you know, this freemium open source model and then, you know, where do we go from there in terms of um the the premium models and how do we scale this from a product roadmap perspective? Are the three things I would think about.

George Pickett: 1:04:40 I’d say on the open source side, I’m going to continue maintaining um OpenCloz Studio on the open source side because number one, I just find it fun to build for open source and kind of get that feedback. On the target market, I have a couple um different thesis. One, one is that there are a lot of people out there who are already really great with ChatGPT, they’re good at prompting, they’re kind of power users at ChatGPT but they’re kind of too scared to go set up a virtual machine. I think those people are the perfect kind of like, you know, um maybe like a technical product manager at a startup, um some like ops person, people who really get AI and are using it a lot but aren’t yet using agentic AI. I think that’s the sweet spot. Um I’m giving it to my mom. I set my mom up with OpenCloz and she’s doing genealogy, all kinds of random stuff so I’m curious I’m curious to see people who have never used OpenCloz, see what they do with it. So the very our first like closed alpha is just giving it to just giving it to people who have never used um OpenCloz before and seeing what they do. But yeah, my prediction is that it’s going to be people who are semi-technical but not programmers who are going to want something more than WhatsApp or Telegram um and and not have to deal with API keys.

Logan Allin: 1:05:57 Do you think like the other part of this is just restricting what OpenCloz has as a… access to, right, on your box and in your files and what APIs you’re integrating with it. I think that’s been one of the barriers to entry in the enterprise and certainly for us, like, you know, do we want to… do I want to just give it access to, you know, our desktop and let it run? I think that component would be really helpful if it’s not already built in just to help people understand the security components of this.

David Kaufman: 1:06:24 Totally, that’s a huge part of what we’re doing and the SAS… like in this OpenPipe Studio, you can do anything you can in OpenPipe. In the SAS version of this, it’s going to be a lot more restricted. It’s not going to be able to do as much, but you’re going to… it’s going to be safe. Like, you kind of can’t screw it up. Um, so it’ll be nice.

Logan Allin: 1:06:42 Idiot-proofing the data is super important. We’d be users, so I’d love to learn more. Thank you, George.

George Pickett: 1:06:47 We’re thinking about this like going from Linux to Mac. It’s like Linux allows you to do anything and kind of shoot yourself in the foot. The Mac version, it’s like can’t quite do as much, but it’s really easy to use and fun to play with. So that’s kind of how we’re thinking about it.

Jason Calacanis: 1:07:00 Well done, George, you’re my guest at Launch Festival if you can join us launchfestival.com, March 16th and 17th in San Francisco, right in the city. All right, we’ll drop George off. Well done. And Lon, we have a lot of Notey questions.

Lon Harris: 1:07:14 We do.

Jason Calacanis: 1:07:15 What’s Notey? Everybody asks.

Lon Harris: 1:07:16 That’s the Notification Gang. Those are the people who have notifications turned on on YouTube. You subscribe to the channel, but you hit the bell. Then you go live, and the Noties are the most loyal fans.

Jason Calacanis: 1:07:26 Exactly, top 1% of fans who come and ask us questions inside the live streams on X, on YouTube, but YouTube seems to be a good place to do it.

Lon Harris: 1:07:31 Exactly.

Jason Calacanis: 1:07:32 Uh, what does Erik think of increased Bitcoin exposure to TradFi? Will players like Jane Street win? Is Do Kwon going to win their lawsuit? Erik, go. Erik, you speak for all crypto. Criminal, unindicted, legit, everything in between. You are the spokesperson.

Erik Voorhees: 1:07:52 Crypto people love having one centralized spokesperson for all their opinions, so I will happily take this mantle. Um, I mean, the whole point, the whole point of Bitcoin was to take over the global financial system. So you don’t do that without the financial system merging with it. Uh, I’m glad to see TradFi adopting Bitcoin. Um, I don’t trust all those firms. They don’t deserve all the trust, but it’s inevitable and, uh, it’s a sign of Bitcoin, you know, ascending to dominance. So I, uh, ultimately like it. Um, as for the Do Kwon question, I’m not up to speed enough to answer that one.

Jason Calacanis: 1:08:27 He was on and I… I asked him like three times to explain staking, and I’m like, ‘I, you know, I deploy millions of dollars a year as an investor. I’ve got lots of equities. I’m deep in finance. I can’t understand this. One more time, explain it to me.’ And he couldn’t. And then he got picked up on a tarmac.

Lon Harris: 1:08:45 Wow, you’re absolutely right. He was on episode 1251 back in July of 2021. Well, you know, I’ll have to track that. We’ll have to track that down.

Jason Calacanis: 1:08:59 Oh my lord. You should have the Twist archivist put some clips from that one out because it was a banger. Well, let me ask you this then, Erik.

Lon Harris: 1:09:00 Um thoughts on MicroStrategy, you know basically buying I don’t know if he’s up to seven or eight percent of the of the non-dead float, you know the active float, you know not the Satoshi wallets that have been dead since inception but what what has that done to the the game, the religion, the project known as Bitcoin in your mind?

Erik Voorhees: 1:09:23 He is someone who has a a thesis and has gone like full bore on it, which I respect. Um to acquire as much Bitcoin as possible is uh a very smart strategy. And um I don’t know all the nuances of like how the financialization of of strategy itself works but the thesis the thesis is valid. The thesis is valid.

Jason Calacanis: 1:09:49 Yeah I don’t know Logan, I’m not sure that anybody understands. Oh boy. There’s layers and layers of ways to participate in the game known as MicroStrategy. Your thoughts, Logan.

Logan Allin: 1:10:03 Yeah, so I’m a full disclosure, a holder of MSTU. So I’ve enjoyed in the yield. Uh and right now just for for the record, it it is yielding 285 percent uh and so not financial advice uh but I did play it that way um because I did not want to have direct exposure or ownership because to your points, I have no clue as to as to how he’s he’s running this business and and it feels like, you know, it feels a a little a little pyramid-y so to speak in terms of how he’s floating these assets and recycling the proceeds and buying more and now he’s investing in these ETFs and raising new ETFs to finance it. So I think, you know, BTC hodl, you got to be up and to the right for this thing to work, otherwise you’re in a lot of trouble.

Jason Calacanis: 1:10:51 Yeah, it’s a you know Lon, when I look at any startup that I’ve been successful in, I’ve been able to explain the business model to my daughters whether they’re 10, 12, 16. I say oh yeah, Uber takes 20 percent of the ride and they match people together and Robinhood, you know, makes money this way and you know Thumbtack makes money this way, Calm charges for a subscription if you need meditation software and you want to have a nice interface, it’s a great way to do it. And um yeah, when I when I can’t explain it or understand it, I tend to move on to something I can understand. And when I see multiple layers of creative financing or community-based EBITDA, like new definitions, these are the red flags for me. WeWork of course had community EBITDA, Logan. A new way to understand profitability is the value it has to the community. Hmm. I haven’t seen any other company take that on as their key metric. For everybody, what AI tool do you find absolutely indispensable? Which one are you using every single day? That’s the question.

Logan Allin: 1:11:57 I use Claude every day particularly for things like… Two big use cases for me, just in saving time and so forth. One is board materials. So just being able to dump board materials into Claude, getting a nice summary, it’s a great way to prep before board meetings and to at least, you know, see the forest for the trees. And then two is data rooms and decks from startups, because we get thousands of them per week and it’s like, I want to make sure we’re doing justice to them. And as you guys all know, VCs are sitting there scrolling, you know, DocSends, and we’re probably not getting as much substance out of it versus a Claude. And then the third one, which, you know, I was at a dinner last night, I will effectively take the list of dinner participants who I’m going to go meet, and if I’m going to a dinner party, whether it’s personal or professional, and I’ll put all their names in Claude and I’ll say, hey, I would love the bios and backgrounds on all these people. And I think the richness of the conversation and the engagement, initially feels a little creepy to them, they’re like, whoa, how did you know where I went to school and why did you do all this diligence? And then I just come up front, I’m like, look, I put all your names in Claude before this. I wanted to learn who I was meeting, and I wanted to have deeper, richer conversations. I think that makes sense. And actually, it’s resonated. So I think, you know, there’s the double-edged sword of AI, but thus far, I’m seeing a lot of positivity, and Claude is the number one tool I use.

Jason Calacanis: 1:13:15 Eric, what are you addicted to? What do you use on the daily?

Erik Voorhees: 1:13:17 I mean, this sounds talking my own book, of course, but like I literally use Venice every day. But as a specific example, to upload PDFs of like legal contracts and medical records, like those two things, which prior, you know, you would need specialists to review, and sometimes you do still want specialists, for sure. But to do cursory reviews and to get a gist of things with AI is, you know, 10 seconds now, and can save literally thousands of dollars. And often will give you better answers than professionals will do. I think that’s an incredible use case that everyone should be trying, whether you’re using that on Venice or any other platform. I mean, that’s…

Logan Allin: 1:14:02 Pretty much table stakes feature right now.

Jason Calacanis: 1:14:04 Yeah. My job, as you know, Lon, is to stay super informed and have conversations. Obviously, those lead to a lot of investment. And one of the things I love is the Grok button on Threads, because a lot of times I don’t understand the context. And then it takes, you know, whatever, one to five minutes to understand what ‘clavicular’ is and what…

Lon Harris: 1:14:24 Clavicular… Clavicular… I believe it’s clavicular. It’s spelled like clavicular. I haven’t heard anyone pronounce it.

Jason Calacanis: 1:14:30 And gooning and looksmaxxing, all this stuff. Like, I could wait, you know, a hundred days for the New York Times to do the ‘trend’s officially over so we’re covering it’ story, or I can talk to somebody to explain it to me, or I just hit that Grok button. Boom. I’m up to speed. The second thing I do, and I’ll just talk about use cases because I think you can use this anyway you like, but I had my Open Claude installed on the Mac Studio, a way to rip videos. And I am a bookmark guy. So fashion, health, inspirational quotes, ideas for… our content here for the podcast, health, all this stuff. Like, I have a bookmark list on TikTok, I got a bookmark list on X, I got bookmark list on Instagram, and then I write my own stuff down in browser bookmarks. I’m a bookmark guy. So now what I do is I just shift the link to the video and my Open Claude downloads the video, puts it into a Google Drive for me, then puts it on a Notion page for me and gives me background on it and I say ‘You’re a curator’. When I send you this shoe—like I was looking at shoes, I want to get a new pair of shoes—and you know I send it some shoe stuff and it was like ‘Okay, yeah, these shoes, you have Crockett & Jones shoes, you have Danner boots, these are other boot brands, this is the history of the boot, this is why they’re special’. I was going down the pen rabbit hole and I bought this new pen, this Zebra G-750. You know, it’s an eight dollar pen, but it’s really quite nice. And I went down the pen thing and I said ‘Hey, put this into my, you know, gadgets and things I like’. So now it’s become like my life curator of the bookmarks and in just one zip, you know, two clicks I can get it and save it for all time because then I don’t have to go find it and what if like the bookmark—what if that TikTok or Instagram goes away? I own it now. I have it on my computer for all time in my memory, which I would never take the time to do that. Second, I don’t know if you guys have a spouse or a significant other—you got a significant other, Erik? Or Logan?—and you watch the stream— you stream a TV show together and then you gotta find the next show to watch, or the next movie to watch? So I just sat there with my wife and I think I did this on Gemini too because it’s very good for, you know, consumery stuff. And I said to my wife ‘Hey, we really enjoyed watching this show, we really enjoyed Get Him to the Greek, we like these stoner films, we like these comedies, we really enjoyed White Lotus, we really enjoyed this show. Give me other shows that we might like. We like Battlestar Galactica.’ So we were going through all the shows that we love to watch over the last 25 years together and it started giving us ideas back and then it has it in memory. I mean this— um, check-in report, CIR. So I say to my replicant and I cc somebody on my team, ‘uh, do a CIR for Marcus.’ While I’m talking to Marcus. And it says, ‘okay, here’s Marcus’s schedule today and tomorrow. Here’s all the emails he sent yesterday and today, ranked by order and with a little summary of the emails. And then here’s all the Slack conversations he’s been involved in. Here’s his Zoom meetings and his, you know, Google Calendar.’ So it gives a snapshot. And then I say, ‘okay, looking at your time management, I can tell you these four things are not super important. Why don’t we deprecate those? And then the two things I just gave you, when you complained you have no time, now you- you swap those things in and out.’

Erik Voorhees: 1:18:06 He’s doing my bit! That used to be my bit!

Jason Calacanis: 1:18:08 It used to be your bit, but I don’t— I can’t bother you all the time. This has led to more joy in my life and this is my key thought. What do rich people have that everybody wants? You mentioned Logan, you know, understanding all the people at the party, that’s called a chief of staff. Chief of staff would do that for you. Huma Abedin would whisper into Hillary Clinton and Bill Clinton’s, you know, ear like ‘Hey, Jason Calacanis is here, he donated this amount’—

Logan Allin: 1:18:32 Like Gary, like Gary from Veep.

Jason Calacanis: 1:18:34 Okay sure, Gary from Veep. I don’t know that reference but anyway you got that chief of staff. Then you have everyone’s personal driver, Uber. Then you have your second vacation home, that’s Airbnb. The chief of staff… such a really interesting thing for everybody to have. And the other one is executive coach. So I started using my Replika to executive coach my team. So I tell it give a— what did I call that acronym? So an executive coach, and I said, anytime you’re looking for executive coaching, run the executive coach CIR. So this is a skill, Erik, that I’m going to refine over time for my team members, which is: you have an executive coach, you have your CIR, use your EC CIR to just manage yourself better and be a- a better executive and- and improve your skill stack. So anyway, those were mine.

Erik Voorhees: 1:18:59 Just what does a rich person have that the rest of us don’t as a framework.

Jason Calacanis: 1:19:03 Yeah.

Erik Voorhees: 1:19:04 Those are awesome. Yeah. Well, I’m building on yours, right? You- the other one you said like getting through all those documents, that was called an associate at a venture firm, and you mentioned earlier to go full circle, ‘Hey, maybe the benchmark thing is better. No assistants, no analysts pairing, just use your, you know, associate replicant to do it.’

Jason Calacanis: 1:19:25 All right, Lon, you got any before- as we wrap here?

Lon Harris: 1:19:27 Oh, uh, I mean, the one- we- we’ll wrap. I do want to give a shout-out to NotebookLM. It was one of the first AI applications that I ever got really obsessed with. I still love it. I still use it for research all the time. And just today we- we had this in the news, we didn’t get to it, they’ve added video overview. So you know how you could put documents into NotebookLM and it’ll make a podcast for you? Now it’ll make an immersive video presentation for you based on your documents.

Jason Calacanis: 1:19:51 You know what this is great for? Your kids. So, uh, I have been working on this when I- I have a, um, bulldog, Minty the Bulldog, that I tell stories to my 10-year-olds for the last three years, and I will come up with ideas for Minty the Bulldog stories. Uh, he’s a superhero bulldog. And just, you know, for me to be able to visualize it would be like, oh my god, I’m- I’m programming my kid’s, uh, you know, education through Minty the Bulldog lessons. We’ll see you all next time on Twist Friday.