Loading summary
Transcript201 lines
- [00:07]
Sam Bankman-Fried
You've had a dynamic where money's become freer than free. If you talk about a Fed just gone nuts.
- [00:13]
Host/Announcer
All.
- [00:14]
Sam Bankman-Fried
All the central banks going nuts. So it's all acting like safe haven. I believe that in a world where central bankers are tripping over themselves to devalue their currency, Bitcoin wins. In the world of fiat currencies, Bitcoin is the victor. I mean, that's part of the bull case for bitcoin. If you're not paying attention, you probably should be. Probably should be. Probably should be dc.
- [00:37]
Connor Brown
It's easy to get to from Philadelphia. I think I want to be here a lot more.
- [00:40]
Sam Bankman-Fried
We need you here, Marty. Do you. We need you here in the trenches. Yeah. We've got pub key now. We've got the office set up. We need more Marty Bent. We need rips all the time in dc.
- [00:50]
Connor Brown
I prefer them in person, as you know.
- [00:52]
Sam Bankman-Fried
Yeah.
- [00:52]
Connor Brown
We've only ever recorded in person.
- [00:54]
Sam Bankman-Fried
Yes.
- [00:54]
Connor Brown
For the third time now, as is tradition, we didn't wait, what was it, seven, eight years between the second and third, like we did with the first and second.
- [01:01]
Sam Bankman-Fried
Right. We're going to keep these recurring basis. A lot's going to be happening in dc. DC is going to be, you know, the new up and coming important city, I think.
- [01:09]
Connor Brown
Yeah.
- [01:09]
Sam Bankman-Fried
Everyone likes to talk about Austin and Miami, but this is where a lot of stuff is going to happen over the next five, ten years with all the technology changes. Like DC is going to be the true city to be in, I think, the coming years.
- [01:25]
Connor Brown
That's why I told you this last time I was here. I've been to the city more times in the last six months than I was in the first 34 years of my life. Yeah.
- [01:33]
Sam Bankman-Fried
I mean, I'm not saying it's a great thing. I would prefer it not be the place that the magic is happening and then. But just with the way things are headed. Yeah. I do unfortunately think it's important that we're here.
- [01:45]
Connor Brown
Yes. Couple things on the agenda today. We're going to talk about AI. Not only the policy implications around AI, but actually implementing it in our businesses, particularly at BPI and what you guys have been up to. I think it's very fitting that we're meeting today because you guys launched the Q2 report of what BPI has been up to this year. So I think starting there and going through all the policy efforts that you guys have been pushing forward and the wins and maybe losses or neutral victories you guys may have.
- [02:18]
Sam Bankman-Fried
Yeah. Gosh, so much stuff going on But I think at the corner of it, it's just this technology continues to improve, whether it be both with Bitcoin and AI, and that is just speeding up the pace of change. We're seeing it in our organization. We're seeing it in what we're being tasked with, different meetings and stuff. And so it's been a lot. So yesterday we released what we did over just the past three months, and it's crazy. I was reading through it and I just couldn't believe how much we got accomplished in three months. And people say AI is going to replace all our jobs and stuff. And everyone I know is working harder than they ever have before because suddenly all of these stray ideas that you're like, oh, I don't have the people or time to do that. You can actually just roll up your sleeves and do it with like an afternoon. And so, yeah, it has been sort of overwhelming. We've been doing a lot of new international work. We've been doing lots of new research that we were able to build out much deeper than we had been previously. So, yeah, it's been a whirlwind, really, since I joined. I mean, we last spoke when I first joined bpi. And just the change in the organization since then has been crazy, crazy. And we're really excited for what it means for the next five years of what we'd be able to build here in dc. Yeah.
- [03:46]
Connor Brown
Well, let's start internationally.
- [03:47]
Sam Bankman-Fried
Yeah.
- [03:47]
Connor Brown
Taiwan. How did that happen?
- [03:50]
Sam Bankman-Fried
So Taiwan, this is a perfect example. We had an idea for a research paper. One of our fellow, we said, hey, Jake, you have a background in DoD. Would you be interested in doing a paper on Taiwan's cash reserve, why diversifying into Bitcoin would make sense for them? And he was excited for it. Taiwan has a really interesting history. I didn't know this, but Taiwan exists entirely because they use their gold as a strategic resource to flee the country, basically. So the only reason Taiwan exists is because they airlifted all their gold outside of mainland China and were able to fund a government in exile. And so we thought, oh, that'd be interesting. We have the research expertise. We'll just do a paper sort of thought experiment. And then we get an email, like, two weeks after we released it with photos from Taiwan's legislature. And they're like, your paper was amazing. It is from Member of Congress Legislator Koh. He emails us photos of him delivering our paper to the Premier of the Legislative Yuan, the head of the Central bank. And we were just like, oh, my gosh, this is Crazy. And then a few weeks after that, we got invited to go and brief them for a week. So another thing that's been keeping us busy is lots of international travel.
- [05:13]
Connor Brown
So what can you share about what was shared in the briefing or the information that was exchanged?
- [05:20]
Sam Bankman-Fried
I can say that Taiwan, every person we talk to is really, really cognizant of the geopolitical situation they're in. It feels like the weight of the world is on them and they're bringing it up all the time of. They know that they've always been in this tenuous position between the United States and China and that every day they feel like the stakes are higher because the strategic importance of Taiwan goes up every day as the compute build out accelerates and there's more and more value on the island. And it just leaves them in this, like, strange psychological position where everyone is thinking about it constantly, like, what does this mean for us? What does it mean for the people of Taiwan? For a long time we had this principle of the silicon shield that America would defend Taiwan's national interests. It would defend them in a military escalation because of how valuable their chips are. In many ways, that's an even stronger case today than it was 20 years ago. But at the same time, it can also be seen as a reason that China is even more likely to intervene because they're now not just have this historical rift, but they have a strategic element here of China is trying to figure out what it means. Could they actually pull off a blockade, all that. And everyone in Taiwan knows that people on both sides are trying to figure out what their plan is. And that just came through across every conversation we had. We met with people from both the different parties there. The DPP and the KMT are the two ruling political parties there. The DPP currently has the majority. And the KMT as an opposition party is leaning more towards some form of reconciliation or, you know, some sort of lessening of sort of hostilities with China. And it's, it's this really, really interesting political dynamic. If you look at all the other policy issues for the parties, they pretty much agree on, I don't know, 90% of policy issues.
- [07:39]
Connor Brown
That's what I was going to ask. Like, how is the partisan polarity over there?
- [07:43]
Sam Bankman-Fried
It's, it's very, very strong. But it's just on this issue of reconciliation, what side do we end up? What side do we lean towards? And I think that both of them are just coming to grips with it. It's like, it's like what everything else we do. It's so hard to predict the future right now because of AI, the strategic questions wrapped up in that. And Taiwan feels it probably stronger than anywhere else on earth. I mean, it's just in the daily conversation all the time. So what I can share is we met at length with their head regulator for their Financial Services Commission and the Central bank of Taiwan. And it was my first time meeting people from a foreign central bank. Super interesting. I went over there not really knowing what to expect. So you write a paper on bitcoin. I've met many central bankers over the years working in policy, and I think everyone is always been extremely skeptical of bitcoin. You know, arms length. Oh, yeah. Nice to meet you. Pleasantries. This was the first time I walked into a briefing. They had read our report, cover to cover, all the different members of the Taiwanese central bank had taken extensive notes and had really, really good questions, actually, about the reasoning, the practicalities, without going into too much detail. I was surprised at their hospitality, their level of understanding, the quality of their questions. It was just a fantastic experience. We want to go back and. Yeah, so hopefully more updates on that soon. But, yeah, I think that this is something that is easy to forget about, is what a signal that the American administration has sent to the world. You know, no one is ever excited with the pace of change in D.C. or how fast or slow things go, but things like the executive order for the sbr, that is still being digested actively and everyone around the world is trying to figure out what it means, what to do with that. So it was a really, really cool experience. And it was my first time being there. So it was just all around, just fantastic.
- [10:13]
Connor Brown
Yeah, sbr, they're fighting over who actually gets control over the reserve right now. Right. It's between Commerce and Treasury.
- [10:21]
Sam Bankman-Fried
I mean, it was something that we kind of scratched our heads at a little bit when the EO first came out, which was they've jointly delegated authority between treasury and Commerce to execute the SBIR executive order. And I think they're just still trying to figure out exactly how do they do this, how do they build up the capacity? I mean, it's been sort of embarrassing for the administration a bit. You know, they had that scandal where they gave custody of the marshals funds to, you know, a contractor with shady origins whose son then ran off with some of the money. I mean, it's just, like comical. It really is just like a circus. So. So in true government fashion, it's always slower than you would expect. But I do, I Mean, what we've heard is, you know, there are lots of things happening that people are trying to get things stood up and there's just like an endless series of bureaucratic hurdles. But they say expect something soon. So two weeks.
- [11:30]
Connor Brown
Tm, we know that well here.
- [11:32]
Sam Bankman-Fried
Yeah, two weeks. Two weeks. Two weeks to flatten the curve, two weeks to execute the sbreo. I don't know.
- [11:39]
Connor Brown
Bring it back. International uk.
- [11:41]
Sam Bankman-Fried
Uk. Yeah. Another thing we went over there and I guess what we're seeing is what happened in D.C. and with the new administration has just want. All of these other governments are wanting to get the inside scoop from the people that work through the process of how is the US thinking about this and how can they be a fast follow? That was the. We went over to the UK and heard that term like a hundred times. The UK wants to be a fast follow. And it was honestly pretty surprising. So I went over there just sort of expecting that they would like throw me in jail if I was looking at Twitter. And instead, like, surprisingly positive. You know, they had just announced a cap on stablecoin balances. And then the week we went over there, they actually rescinded that. And they're still like many other governments trying to wrap their head around what all of this means. But again, surprised by the level of openness. We met with people in. We did a briefing in the House of Lords. A ton of fun. First time being in there, it was in the parliament building. So it like felt like a lot of pageantry involved and lots of pageantry. They have statues in there from like 1600s. I tried to take a photo. I got accosted by some security guard. Looked like the Hogwarts, like cafeteria. But yeah, surprisingly optimistic there too. You know, afterwards, the head of their is like a head of innovation at their fca put out a thing about the importance of bitcoin policy. I mean, we have to remember, like these people, they don't even understand the difference between bitcoin and shitcoin. They don't understand how bitcoin is different from the parade of horribles they hear about. And they are very interested in the story we have to tell. So I think the lessons learned is you have to tell the story in a way that relates to their history. You know, when we go over there, we talk about, hey, you know, Newton set you guys up on the gold standard and that paved the way for 200 years of British Empire. So those are the types of stakes that are at play here. And I think explaining in those like broader historical terms Same with Taiwan. You know, we talked a lot about their historical use of gold as funding government in exile relating to those longer term historical narratives for a country. Same here in America. Is the key to getting people to understand the policy implications.
- [14:24]
Connor Brown
Yeah, it's incredible work.
- [14:27]
Sam Bankman-Fried
It's exhausting. Thank you. I mean, we got a lot more to do. There's no shortage of work.
- [14:33]
Connor Brown
Well, I mean, you guys are the hot chick on the block. Not because of bitcoin, because of AI. I mean, bringing this back, bringing this back to Taiwan. Yeah, you guys, I think Sam spearheaded the report, basically unearthing that China has been influencing conversations around AI data centers here in the United States.
- [14:53]
Sam Bankman-Fried
Yeah, that started as like we were saying, with AI, you can just do so much more. So it started as sort of like, I don't know, a side bit. You know, people had mentioned to us, you know, bitcoin miners mentioned to us they've been seeing just unprecedented local opposition and that they've been doing this for years and they're used to some local concerns. Obviously, with any major infrastructure project, people are going to have questions. But a few months ago they said, you know, we've. We're seeing the same local people that have concerns. Typical stuff. But we're also seeing this new thing which is people from out of town, that while the local people will ask a reasonable question, like, okay, where is this going to be located? Is this going to create noise? Is this going to impact me in XYZ way? There would be these people from out of town that would really aggressively say, this is going to give everyone an aneurysm. Right. Or this is going to destroy all the bees in the area. I heard that no bees could ever exist within 50 miles of the data center. Is that true? Just like ludicrous things. And they were like, we've never seen this before. Can you guys see if there's anything going on? We said, yeah, we'll take a look at it. And we just got Sam Lyman that joined. So he came over from treasury and Sam has experience dealing with this stuff. So he used to work at Riot, right? He used to work at Riot. He's done the rounds in local communities in Texas working on this stuff. And he just kind of started pulling the thread and you start going through these different layers of nonprofits and all the different activist groups that are on the ground and mobilizing, and it just kind of shocked all of us. And then it just all sort of worked out in the virality of it because everyone's talking about data Centers now. And we were really the first to sort of put all the pieces together and say, well, actually if you look at all these different projects, we're seeing a common element which is the Party for Socialism and Liberation are functioning as at least important organizers in a lot of these debates. And they have sort of a dubious history. So yeah, it's been a whirlwind past couple of months since that story broke.
- [17:22]
Connor Brown
Who's the individual behind all of it who's very vocally anti.
- [17:26]
Sam Bankman-Fried
We don't know who is. Like a lot of. There's a lot of questions about the organization. What we know is that someone named Neville Singham Roy. Neville Roy Singham is a, you know, very wealthy philanthropist that lives at least part of the time in Shanghai, has said a lot of positive things that sort of align his views with those of China. And he is at least funding some nonprofits that are doing anti data center stuff. There's a lot of dark money involved. A lot of it is sort of unclear. And we're trying to piece it all together and understand exactly what's happening. But yeah, I encourage anyone that's interested in this, read our reports. We have two different foreign influence reports in the case for American AI that sort of go over exactly what this nonprofit structure looks like. Sort of all the different elements of foreign influence we've identified and we've sort of detailed case studies of what we're seeing on the ground. And one thing I do want to say is this is not to say in any way, shape or form that Americans themselves don't have legitimate concerns about this massive multi, multi billion dollar trillion dollar build out of compute.
- [18:53]
Connor Brown
Multi trillion. Now at this point, yeah, I mean
- [18:54]
Sam Bankman-Fried
it's like huge in its implications. These projects are larger than anything we've ever built before and are going to continue to just scale up over time. People have many legitimate questions about that. How is that going to change their way of life? I don't think that the narratives that anthropic or some of these effective altruist groups are putting out are compelling to the everyday American. And that is manifesting in a lot of concerns about this build out. So all of that, I think there are legitimate concerns and fears about AI and these data centers, their impact. But a proper accounting of this means you need to also think, okay, is there foreign influence? What does that look like? Because there are strategic questions at play. And we know that our adversaries are savvy. They understand how social media works. They understand that in modern global great power competition, you're usually not fighting with boots on the ground.
- [19:59]
Connor Brown
Fifth generation warfare. Yeah, the podcast I recorded this morning, we talked all about it.
- [20:03]
Sam Bankman-Fried
Yeah, yeah. I mean, it's just. It is the. The modern battlefield is on Twitter and Instagram and they are, I think that we're just seeing like the latest in. Well, that's why in the stock.
- [20:15]
Connor Brown
We should all be thankful that BPI exists. We try to do our part at tftc, but you have to engage in fifth generation warfare and that means going, pulling on the threads, writing the reports, getting it out there. And I mean, was that the most viral thing you guys have ever done? Funnily not related to bitcoin.
- [20:30]
Sam Bankman-Fried
Yeah, it was probably the most viral. I mean, we're on the front page of New York Times last week, so. But just the beginning, Just the beginning. I think there might have been. I don't know, I mean, the numbers on it have just been insane. It's one of those things you hit at the right time, it gets circulated by the right people. Marc Andreessen tells all his followers to read it, and then the Internet explodes.
- [20:50]
Connor Brown
Yeah, I mean, that's why I'm in town today is for the event tonight, which is talking about energy policy, data centers, the intersection and how we should think about these things, approach them. And I had a conversation last night. If you listen to the previous episode, you're getting this again. But I think it's important to reiterate to Connor. I was sitting down with somebody who was involved in one of the largest utility companies in the United States, and we were just talking shop about like, all right, how do you educate the masses about this? Because that's, I think, one of the biggest uphill battles we have. Bitcoiners, I think, know this more deeply than the AI data center people now, because we've been doing this for 10 years. But there is a lack of understanding of energy, the grid, how all of these things work, how data centers, bitcoin mining operations, how they interact and how they affect your ability to access cheap electricity. And that's why I'm happy that American bitcoin, Pubke and Pubkey are starting this, the series which would go for the next 10 months. But having conversations around these things and I think to your point about social media, it's like packaging it and getting it out there in a way which I think your report did very well.
- [22:04]
Sam Bankman-Fried
Yeah, I mean, it's so important to tell the story and I do think it is critical to bitcoin as well, is people. If we are going Down a path where people are hostile to just energy production generally that is not good for anyone. But, but really it's not surprising what we're seeing either because what we're seeing is a big tech executive gets up and says, well, AI is probably going to be as dangerous as pandemics and nuclear weapons combined. And it in the process is also going to take your job and the jobs of everyone, you know. And yeah, we're going to keep building it. Okay, how is this going to be compelling? This is not compelling for anyone. And I think it's just such a break from the American tradition of leaning into innovation and entrepreneurship and the optimism of building great things. That is what has made people inspired and want to come to America and want to build here. And what causes people to get up every day and build great companies is human achievement and greatness. And energy is at the bottom of all of that. And you know, I'm really excited. There's a lot of really, really great startups now that are getting into. I'm really glad to see their startups that are doing real hard tech, energy production breakthroughs and stuff like Valor Atomics or whatever. You know, I saw that on Twitter. It's been all over my feedback. But if we are not telling a compelling story about how technology benefits us with either bitcoin or open source AI
- [23:51]
Host/Announcer
or
- [23:54]
Sam Bankman-Fried
biotechnology, if we're not telling that compelling story, then it's going to manifest in hostility in a million different ways. All the way from the hyper local level of not letting any of this stuff get built, all the way to federal licensing regimes. KYC for compute surveillance of all your chat logs. It gets like very, very dystopian very quickly. When you start convincing everyone that this is this crazy danger that we have no precedent for how to handle. It's just not true. It's not how our society was built. We're a society that trusts people with dangerous things. We, a lot of people have guns. I mean in theory, America itself was founded on the idea that the people themselves can be trusted with dangerous tools, with self government. I mean this is like the core of what we believe is that the people can be trusted with things and
- [24:51]
Connor Brown
if they break that trust, they will be prosecuted accordingly.
- [24:54]
Sam Bankman-Fried
Yeah, exactly, exactly. We've dealt with dangerous things in the past.
- [25:01]
Connor Brown
What's up freaks? This episode is brought to you by Block's bitcoin ecosystem. You probably know Cash app, you probably know Square, you definitely know bitkey. If you've been listening to the show for the last couple of years, what matters most is that these are not disconnected random products. They are part of a broader push to make Bitcoin easier to buy, easier to hold and easier to use in the real world. Cash App helps people stack Bitcoin Bitkey helps people custody without turning self custody into a weekend engineering product. Square helps merchants run their businesses and increasingly plug into the Bitcoin native payments Rails Listen, this is the kind of ecosystem Bitcoin needs. We've been talking about this on the show for a while. We know that we need the different parts to line up to make Bitcoin. Everyday Money Block is doing just that. Check the episode description for the current Cash app, Bitkey and Square links. All right freaks, you know me. You know I don't take sponsor money from products I wouldn't use myself. So listen up. The Aven Bitcoin Visa card is one of the most interesting things I've seen in the bitcoin lending space in a long time. Here's the deal. You can get a line of credit up to a million dollars backed by your Bitcoin without selling a single sat. No games, no annual fees, no minimum draws. And your Bitcoin is custodied by Bitgo, which is one of the most trusted names in digital asset security. Ava never lends it out. There's no rehypothecation. You stay in control. And guess what? You can lock in a fixed rate for up to 10 years. That's 10 times longer than most lenders out there, right? Go interest only for up to five years. Rates start at 7.99% APR. For a product that lets you keep your stack and still access liquidity, it's hard to beat. I mean the duration in the rates is the best I've seen in the market to date. You also get 2% unlimited cash back every time you use the card. Spend fiat. Keep your Bitcoin the whole game. If you've been stacking for years and you need liquidity without triggering the taxable event, this is worth a serious look. Go to aven.combitcoin that's AVN.com bitcoin check it out. I think getting back to the AI narrative and it's one thing we've been working on at TFTC too because we've been leaning in and maybe we can spitball some ideas, compare notes to how we've been implementing this stuff. But I think I've been trying to do this more and it's not an attempt to brag, but I think people, the broader public has an outside looking in view of AI. And most of them haven't really dove deeper than downloading ChatGPT and interacting with the chatbot. I was actually relieved to see Secretary Bessant do an interview this morning. We were clipping it at TFTC but talking about this exact problem where the interview was asking him, Dario Emodi is telling us it's going to take all of our jobs. And you have this Secretary treasury has to be like, well, he's wrong about his business. If you actually look into the data, I believe that AI proliferating people learning how to use it is actually going to lead to thousands of flowers of individual small businesses springing up. You can do the job of 10 men, 100 men with one now. And if you're implementing this tech at your company, which we both are, it is very obvious that that is true. Like you said, I've never worked harder in my life because I can do more. But just diving into and really bringing like a real world example of how AI is not taking any of the jobs at bpi, it's extending your ability to do more jobs. Yeah. How are you guys implementing it?
- [28:32]
Sam Bankman-Fried
Yeah. So when Claude code came out and then Claude cowork, it just completely blew our minds. We had been using these tools a lot. I was already, I guess, like a power user of this in many ways. But the transition to agentic models that can actually do full workflows is just crazy. And so we just sat down early in January and had an all staff call where we said, here's what we're doing, we're going to really lean into this. We're going to become the best people in D.C. to know how to use these tools. And each of us spent, I don't know, I don't know how many hours, right? Each of us just started playing with the tools, right? We said, everyone think about these tools. Here's a brief overview. We did a presentation for like two hours for the team and said, here's generally what they can do. We want everyone on the team to spend a lot of time using these. Think about your existing workflows. Everyone has some sort of recurring task. They do think about how you can build these into recurring workflows. And then we want you to present to the team how you're going to use this. And the great thing about it is you can then encode those workflows into skills. And so that was a major unlock for the org because suddenly we just had our team's skill set, just started populating you know, and every. We had a new thing in Slack of announce your skills here and people can know how to use them. And so it became this force multiplier where if you needed a specific skill to be called, obviously you would. If you needed something for research, you talk to the head of research. But if you're doing something that needs some research, you can call the research skill and sort of like slot that in. And it was just like, it was an incredible process to see it happen because now we're six months past that and I would say our work product is better quality than ever before. We're getting a lot more traction on social media. It's just very clear to me that there are going to be two types of organizations and we're going to see the K shape divergence of the firm. There are going to be firms that start to punch well above their weight because they have leaned into gaining the expertise in these tools. And the best thing about the tools is if your team really understands not only how to use them, but how to improve them over time, give really good feedback to them. Make sure that the agents themselves are taking notes on your feedback. You can get it so that every error they present to you, you never see it again. And so now we're at a point where not only are we doing the things that we were doing six months ago better, but we're doing things that we could have never done before, like our AI research on Bitcoins. Preferring AI, sorry, AI agents, preferring Bitcoin. That was hugely viral. We were able to build out an entire website, much more interactive, much more multimedia stuff. We were able to do the scale of the research that wouldn't have been possible before. So I think that there are obviously growing pains. We are dealing with reducing slop all the time. It has become this constant sort of thorn in our side of like, how do we make sure there's no slop in the organization? Because, you know, there, there are some, some like, you know, stumbling blocks to it or things you have to learn how to manage.
- [32:25]
Connor Brown
But we deal with this too. Yeah, you can imagine.
- [32:28]
Sam Bankman-Fried
Yeah, but like the long view is like just so obvious that you're just going to have the adopters and the people that don't take the time. Yeah.
- [32:38]
Connor Brown
Does BPI have a company brain?
- [32:40]
Sam Bankman-Fried
A company brain? What do you consider a company brain?
- [32:45]
Connor Brown
A persistent memory system attached to a knowledge graph. So we have a company brain at TFTC. It's three layered. So we've leaned into harnesses. We used OpenClaw for six months. And then a month ago, after hearing about Hermes, so much like watching it from afar, I was like, all right, let me try Hermes. So we have our main agent that we use for a lot of tasks at TFTCs, and a VPS that I control. We're anthropomorphic. We're anthropomorphic, these machines. The clanker is in the VPs, and that's where everything sits. So we had the Openclaw, and I basically just went to the Openclaw agent that was running on Claude 4. 6, and I said, hey, I want to try out Hermes. Can you spin up a Hermes harness for me in your server and just let me know when that's set up and then tell me how to create a telegram, chat with that, with that agent, and I'm gonna go from there. And so, like, set it up and then connected it. I wanted the A B test, 4.6Opus 4.6 versus GPT5.5 at the time. Yeah. And so I had the Hermes harness set up, connected it to 5.5 as the brain, and played with it for a day. And almost immediately I was like, oh, this is better than what we've done. So I'm just gonna start using this as my daily driver. But because we had the company brain, so in that same server, we have like a three layered memory system. It's notes, it's like solemd, the agent md, the skill MD files, daily notes, wiki, LLMs. There's like flat markdown files. Then we have a QMD semantic search vector database, which, like, so if I give a prompt and it has keywords in it, it'll query that database that lives on the server without wasting any tokens, and it'll find context of us dealing with that term in the past. And then we have cogni C O, G, N, E, E, which is like an Obsidian vault, but it's agent first. And so that's the third layer which is more expressive. So you have basically terms and vaults of particular subjects. And QMD will query, like, every time we talked about it. Then it'll go to cogni, like, okay, we've talked about this in the past. How is it interrelational to every other thing that may relate to this topic? And so it gets a full context of how we've dealt with these certain terms. And it has temporal context to it, so it knows, like, when we've talked about them.
- [35:14]
Sam Bankman-Fried
That is amazing.
- [35:15]
Connor Brown
And then it'll go to the LLM like, okay, based off all this context, how we frame things so it has all of our transcripts, all of our newsletter. So it now knows how we frame things. And that has been the most ROI that we've gotten in terms of investing the time to build out A particular part of the stack is that brain.
- [35:35]
Sam Bankman-Fried
It's really interesting. So that sounds like you're really trying to get as close as possible to an employee who has been with the org since its inception. Yes, has seen everything. Because that's what we're doing every day when we're working there is like you're building institutional knowledge. That's very cool. We're a little bit more simplistic. We use just the standard like Claude Enterprise subscription. But we do have something similar to a brain in a sense that we have like a shared set of context files. So we've sort of curated a project file that then has subfolders for every different arm in the organization. And we have at the top a Claude MD file that gives a basic overview of what we are and all of context and all the different files within it. And then explain how it can search through those files and understand what we're doing. And then each one of the arms of the organization will have its own subfolder for comms or design or research or whatever. And then each of those we have the sub cloud MD files for each of those subfolders, sort of example workflows, example prior precedent and then cross references of. All right, here's our research workflow. We expect you to do this 20 step process for doing a research report and then once you get the user's approval or feedback, we sort of have encoded the entire lifecycle of the product. So we'll start with. All right, if you kick off the research skill, it'll start with ideation and then we have specific gates where the user provides feedback and then it'll continue working up the product and then give you another user feedback gate and it'll continue working up the product and then once it's finalizes the research portion, it'll then hand off to the comms portion or it'll hand off to design and then comms. And so we have it just as a sort of intricate folder and subfolder system with cloud MD files throughout.
- [37:43]
Connor Brown
And it seems like an assembly line.
- [37:45]
Sam Bankman-Fried
Yeah. What do you think of it? Yeah, exactly. Yeah.
- [37:48]
Host/Announcer
Well, that's the thing.
- [37:49]
Connor Brown
There's like a thousand ways to skin the cat for your purposes. Like that probably makes a lot of sense for like our purposes. This makes, that's, that's what's so fascinating. Is that the landscape is so wide now, everybody's still mapping the territory too. And so I just love hearing.
- [38:04]
Sam Bankman-Fried
And honestly, we could use both of these. We could probably use a database like that for just context about BPI and what are our positions and what are the papers we published that would probably be a helpful thing to know how to plug in. What I have noticed though is you do have to be really careful because you can almost like accidentally context poison yourself where you'll notice it'll keep making an error and then at some point you're like, oh, it's saying that because at some point in my feedback to left a note for itself that it reads every single time it gives an output and it should not be reading that note. And I've consistently found whenever I'm having some sort of capability problem, it's either problems with my prompting itself or problems with the context that I'm giving it. Like I'm giving it bad examples of prior work product or something that aren't up to the standard we need. And it's drawing on them for design inspiration or something. And so always thinking about what is the context for every single prompt and trying to. It's almost like pruning or something like a bonsai tree. You want it to constantly be improved. And you know, every time you use it, you're in some ways you could like accidentally poison the context for future conversations or improve the context for future conversations and teaching the employees that too of using these systems. Your initial implication, your initial, I don't know impulse would be if it makes a small error. Okay, I can just like fix that myself. It'll be a small tweak, but if you understand the right way to give it feedback, you can provide feedback in a way that what might just be an easy fix for you now is fixed forever if you just explain, hey, this needs to be fixed and keep a note for yourself to never make this problem again.
- [40:07]
Connor Brown
Make no mistakes.
- [40:08]
Sam Bankman-Fried
Make no mistakes forever going forward. But it is really great to see. It's also just really fun. Yeah, I find it to be like a much more fun way of working.
- [40:18]
Connor Brown
I do too. Are you voice to text or are you typing stuff?
- [40:25]
Sam Bankman-Fried
I am voice. Of course.
- [40:26]
Connor Brown
Yes.
- [40:26]
Sam Bankman-Fried
Yeah. Whisper flow.
- [40:27]
Connor Brown
I use Super Whisper.
- [40:28]
Sam Bankman-Fried
Oh my gosh. Local model. Great. I know. I mean we need to be better about this, right? But.
- [40:35]
Connor Brown
And I was just having this conversation with Thomas because I co write the newsletter now with my agent.
- [40:42]
Sam Bankman-Fried
Yeah.
- [40:43]
Connor Brown
And basically what I'll do actually.
- [40:45]
Sam Bankman-Fried
Does he have a name?
- [40:46]
Connor Brown
Martin. He's A more sophisticated Marty.
- [40:49]
Sam Bankman-Fried
That's amazing.
- [40:51]
Connor Brown
Okay, so Martin is trying to. I'm trying to make somebody recognizing that I don't have the capacity to store files in my brain like Martin does on a server. So he's more sophisticated, he's going to be better. But I was just talking about this. Thomas to Thomas. I was writing the newsletter here at PubKey earlier. And by writing the newsletter it's like, hey, basically my flow is I was showing you yesterday, I sent you a screenshot of the BPI Q2 report which we covered in the newsletter today. And basically yesterday when that came out, I sent a note to our bitcoin brief topic in our telegram chat and said, hey, I want to talk about this tomorrow when we go to write the newsletter. This is one of the things we're going to cover. So I'll do that and I'll do anywhere from five to ten links. And then morning of it's like, okay, pull up all the links, all the emails I forwarded to you about stuff I want to cover. And let's. I want you to. Based off of the engagement that we've gotten in the newsletter in the past, you have access to that data on the back end of our cms. Like, structure the newsletter in a way that will get the most engagement. But like, so we're filtering. I'm definitely covering these topics machine, you know, based off the topics we want to talk about, which is going to resonate the most with the audience and what they may deem most important. So like, we have a lead story and then signal stories. So then it comes back to me. It's like, okay, I think this should be the lead today and these should be the signal stories. And here's how I'm thinking about framing it. And then I'll be like, okay, I understand why you're framing it this way, but let's tweak this. That and I talk for 10 minutes. Okay, here's what I want to cover. Frame. And then it's like, okay, I think we're at a good spot. I like the lead, I like the signal stories. Let's go. Right, so that puts together a draft, gets in the back end of our go CMS and I read it twice and then give final edits. Like, okay, you messed up here. And then we hit send and everybody's like, oh, you're not writing anymore. You're like, lose that skill. And I thought that for the longest time. Like, I actually shied away from using Claude for writing for a while, but now I feel more comfortable for two reasons, one of which is the company brain is there. So it has that institutional knowledge of how we frame things, even how our opinions have evolved over time and like how we. The voice and the positions that we have on certain topics. And it's able to inject that based off of stuff that we've talked about on the show and written. When I was still manually writing in years past. The second thing is I had a conversation with somebody, you know, we're not going to dox him on the show. And he completely changed my mind because people were worried, like, we're going to use the art of writing. And I do still. I do write, I do type. I tell people. I'm like, ah, I don't type in. I do sometimes. I do still write handwritten notes. I like that art. It's an art form that should never die. But he described it to me. He's like, we're basically just returning to the era of the scribe. And he changed the way I prompt because the way he describes it to me is like when you're talking to these things, since it's an LLM with a matricy vector database in the back with points of association, you need to give it as much context as possible so it can eliminate points of association in that matricy to give you more narrow output. And so what you do is you pretend like you're Napoleon Bonaparte.
- [44:02]
Sam Bankman-Fried
You took it out of my mind. I was just about to say this.
- [44:04]
Connor Brown
Right, go ahead. It's like after a day of battle, you get back to your war tent and you have 10 scribes. One for munitions replenishments, one for war philosophy, like, ah. One for strategy, like, here's where we messed up today. And you just talk to. You spin up a bunch of sessions and each session is an individual scribe that you give a bunch of context to. And it forces you to think from first principles and really articulate what you're trying to get out there.
- [44:29]
Sam Bankman-Fried
So funny. Bitcoiners. We all just like come to the same conclusions independently. Have you read Napoleon's biography?
- [44:35]
Connor Brown
No.
- [44:36]
Sam Bankman-Fried
Oh my gosh. Okay. Andrew Roberts, Napoleon A Life. Best book I've read in a long time. Just incredible. I think about this all the time. Napoleon was like the most incredible human I've ever read about. He would do this, he would go back, he would triple dictate. He would be able to dictate faster than his scribes could write things down. And so he was one of the only people in recorded history that had this ability where he could do exactly what you said he could dictate. It's similar to a grandmaster chess player who goes and plays 20 different games simultaneously and comes back and he can remember the context for the moves he's planning. Napoleon would do the same thing with writing these really just incredibly detailed and important letters to all the different world leaders, all of his different troops. It's crazy, the breadth of what he was able to cover. But, yeah, he would go back to the tent, and after a day of the most intense fighting in documented history, he would then go and dictate a letter on the importance of building a local bridge in a rural municipality in France, and then follow that up with a letter on the need for a girls education academy, because there really isn't one in this certain area of Paris. And then he'd go on to another one and dictate something about, you know, preparing his army for invasions in Russia. I mean, it was just incredible how he's able to do this. And I do think that that is what these agents allow us to do, is multiply human agency. And that is the good timeline. The good timeline is that the best aspects of human greatness, achievement, and will and vitality are multiplied. And I feel that happening when I use this. But I do think it is something to be really cognizant of how you're using the tools. Like, if you're using it in this dictation sense where you're able. When I am using it, you know, and dictating to it, I can say so much more than I could possibly type because I can just talk forever. All of us can. Yap. Typing can be exhausting and slow.
- [46:53]
Connor Brown
I type slower than I think.
- [46:54]
Sam Bankman-Fried
Yeah, exactly. So you actually like. I feel like I lose some fidelity of my thoughts to commands when I write now. There are certainly many things that you have to really carefully phrase and write correctly, and that is in publications and things like that, that I think you should spend more time thinking about exactly what you're saying, phrasing it, refining it like a sculpture or something. But for just raw output and ability to impact the world, the dictation model is just incredible. It's awesome. Yeah.
- [47:32]
Connor Brown
Yeah.
- [47:33]
Host/Announcer
Sup, freaks? When you take Bitcoin seriously, you start with custody. You want to control your keys, avoid single points of failure, and make sure your savings cannot disappear because you or someone else screwed up. That is what Unchained has been focused on since 2016. Unchained is the leader in collaborative multi sig custody and Bitcoin. Financial services that keep you in control. They secure over $12 billion in Bitcoin for more than 12,000 clients. That means about one out of every 200 Bitcoin sits inside an unchained vault. Their model is simple. You hold two keys, they hold one key. It always takes two keys to move bitcoin, meaning their single key can't access
- [48:02]
Connor Brown
your bitcoin on its own.
- [48:03]
Host/Announcer
Just resilient shared custody that gives you institutional grade security while keeping you sovereign. Unchained also lets you trade straight from your vault, access bitcoin backed commercial loans, open a bitcoin IRA where you hold your own keys and set up personal business, trust or retirement vaults. They even offer inheritance solutions built for long term hodlers. Or opt for the highest level private client service with unchained signature and get a dedicated account manager, discounted trading fees, exclusive access to events and features, and much, much more. If you want a partner that helps you secure and grow your bitcoin without giving up control, go to unchained.com and use the code TFTC10 at checkout to get 10% off your Bitcoin multisig vault. That's TFTC10nchain.com freaks.
- [48:40]
Connor Brown
Look at me. I'm glowing. I've got like an angel's halo going around me. You know why that is? Is I feel good. I feel taken care of. I feel blessed, healthy, happy. That is because I'm a crowd health member. My family and I have been crowd health members for five years now. Literally this month, five years ago, we joined crowd health. We've had two babies, we've had multiple health events and we're never going back to health insurance. Crowd health is crowdfunded healthcare. So you, you sign up for crowd health, you pay a monthly fee, you help out with other people's bills and it's significantly cheaper than health insurance. We were on Cobra. It's a family of three. When I left my last job before I went full time with tftc, went on the crowd health. Now as a family of five, we pay, I believe $700 a month. It's significantly cheaper. They're going to negotiate prices lower for you. They've consistently negotiated health care prices as much as 50, 60, 80% in many cases. They help out with babies. You have a pregnancy, you pay the first $3,000 and the crowd covers. If you have a regular health event, you pay $500 and the crowd pays the rest. Go to join CrowdHealth.com sign up today. Use the code TFTC opt out of health insurance. I'm uninsured, baby, and I love it. Use the code tftc@joincrowdhealth.com you'll get $99 a month for the first three months that you're on the crowd health platform in the community. Bitcoiners, you found sovereign money. Now find sovereign health and sovereign healthcare.
- [50:02]
Sam Bankman-Fried
What.
- [50:03]
Connor Brown
What's the rate of adoption of AI tools outside of BPI in the city? Is it big or is it.
- [50:08]
Sam Bankman-Fried
I expected it lot higher than it is.
- [50:11]
Connor Brown
Well, I have. I have this theory that, like, bitcoiners are just like early adopters to all this stuff.
- [50:16]
Sam Bankman-Fried
This is what I was just thinking is like, every time I talk to a bitcoiner about this stuff, they have some crazy workflow because we just lean into new technology. And I mean, I've talked about this with like, Thorne and Galaxy and Skyler on the design stuff. Like every bitcoiner I know at all these different places in the industry or different types of jobs or skill sets, all of us have independently come up with crazy interesting ways to use this stuff. When I talk to people at traditional think tanks, this isn't even on their radar. There's just an institutional inertia for a lot of these more legacy institutions. I think this is true. Zach Shapiro, our head of policy, he's finding this at the law firm level.
- [51:04]
Connor Brown
He's beating the drum. I mean, he's going out BPI Policy Hour, I feel like, has been like, zach, lawyers who better start by coding, better start using this.
- [51:12]
Sam Bankman-Fried
Well, he wrote an article that went viral and then he from that had so many attorneys reach out to him that are at major law firms and they're just so frustrated. I mean, I dealt with this. It was frustrating for me in 2022, 2023, when I was working at a firm here in D.C. of just how hard it was to use any of the tools in the way that they're meant to be used. And that was when the tools weren't even that capable. You know, like, we couldn't even access it on our laptop. Now they have some mediocre tools, but it's all sort of like baked in stuff like Westlaw AI and stuff like that. And Zach has done a great job of just saying, no, you need to just integrate it top to bottom, and you will be able to provide professional services at 1/10 the cost and twice the quality. And I think that has really surprised people at what he's been able to pull off. And then from that, I've met lots of Other, it seems like what's happening is there's lots of senior associates or young partners that are breaking off from law firms. And I'm sure this is happening in other fields too, where they're realizing, hey, as a solo shop, I can do what a team of 20 would do and do it a better quality.
- [52:35]
Connor Brown
And keep all the money.
- [52:36]
Sam Bankman-Fried
And keep all the money. Yeah. Cool to see.
- [52:41]
Connor Brown
I mean, this dovetails into an important point that we definitely should discuss, which is like policy around AI as we're describing it, is incredibly powerful. And I think it's incredibly important to make sure that we're able to utilize these tools in the way that we know we can moving forward. And so like, policy around AI, whether it pertains to the models you can access and the energy debate around how those models are being trained and how the inference is being delivered is extremely important. And it seems pretty chaotic right now.
- [53:19]
Sam Bankman-Fried
It's very chaotic. I constantly think about what we did with nuclear energy, right. We invented basically unlimited energy and then we banned it and built enough weapons to blow up the world many times over. I mean, we, we basically did the absolute worst thing possible with that technology. And I'm really, really concerned we're going to do the same thing.
- [53:42]
Connor Brown
It took us, what, 50 years to get back to a semblance of sanity in nuclear policy.
- [53:47]
Sam Bankman-Fried
Yeah. And honestly, we're really far from it. Like, the NRC has done an unbelievable amount of damage on the nuclear energy, on the nuclear industry. And what's so funny is if you go back and you actually read the announcement of the nrc, all the different major nuclear industry heads were like, this is a great new day for America. I can't wait for America to build out nuclear under this new leadership of a nuclear regulatory commit. I mean, people were just like patting themselves on the back. The people in Congress are super happy with it. The industry loved it, and it killed all the progress. It just became filled with people like, if you listen to the guy who is the head of the NRC under Obama, he will openly get on stage and brag about how nuclear is the worst technology. We stopped it in every way we could. It's not going to save anyone. It's very funny. So point being, I'm really concerned that we're going to go to the same place with AI. The government, just the nature of the institution struggles so hard with understanding exponentials or understanding just reasoning about existential risk. I feel like, is this way to just like one shot the government? You know, you have someone like during COVID lockdowns. Right. You have someone like Trump who in theory is pro freedom, pro innovation, pro small businesses, et cetera. They said, okay, there's a risk here that this could be some sort of major catastrophe. We're just going to shut everything down indefinitely. Print a bunch of money, print infinite money. Like they just can't reason about when someone says, oh, this is existential, like just logic goes out the window. We saw it with nuclear, we saw it with the pandemic. We saw it. We're potentially seeing it now with AI. I do think that the administrative, the current administration is somewhat inoculated to restricting AI, but it is. We're still at the very early innings. How many people have died from AI? We haven't had a major catastrophe. I think that inevitably, like any technology, you invent ships, there's going to be shipwrecks. You invent AI, some people will use it as a tool for bad things or accidentally use it to cause some sort of harm. I guess my point is we are seeing pretty strong calls for heavy regulation of this technology before we've even seen any of those, you know, misuses or harms manifest.
- [56:38]
Connor Brown
We're on the other side, on the positive side. Explore the full potential of it.
- [56:42]
Sam Bankman-Fried
Yeah, or explore the full potential of it. And so I'm just worried that without someone explaining in clear terms the optimistic case for human flourishing with AI, how this is a complement to human agency, greatness, achievement and all of the sort of knock on benefits it can bring to individuals, localities, states that you're going to just go down a pretty dark path of you'll get surveilled, closed, permissioned AI fiefdoms. There are people in the administration that defend open source, but is that going to stand up over time? If we have open source, that's at the capability of fable or five times better than fable. I really hope so. It has to in my mind for us to have a free society. But there's strange winds blowing in D.C. and I feel like there's a lot more money coming in on the restrict side.
- [57:48]
Connor Brown
What's whispering in these winds?
- [57:50]
Sam Bankman-Fried
Yeah, I mean, just one example. Trump's sort of one of Trump's like ground game lieutenants. Her name is. It's escaping me now. I think it's Lisa Kremer. She used to, she ran like the Stop the Steal grassroots campaign for the 2020 election and the dispute over that. She is now being funded by effective altruist.
- [58:17]
Connor Brown
Oh gosh.
- [58:18]
Sam Bankman-Fried
To run a new campaign called Humans first, which is like A MAGA aligned. Anti.
- [58:26]
Connor Brown
Populist. Anti.
- [58:27]
Sam Bankman-Fried
Yeah, populist AI. So I will say the, the doomers, the people that want to paint this pessimistic view of AI, they are very, very savvy about how to play in the conversation.
- [58:42]
Connor Brown
The frontier model providers themselves. I mean, I think OpenAI's sort of backed away from it. Like seems like Dario and Anthropic are leaning into it. And one has to wonder, view their marketing or how they're positioning things with a lens of skepticism because it's an easy way for them to get a regulatory moat which if you're following economic incentives is probably something they want. At the end of the day it's
- [59:06]
Sam Bankman-Fried
really difficult to actually figure out motives and that is what's difficult about all this. I do think there are some people that genuinely think AI is dangerous.
- [59:18]
Connor Brown
Paperclip problems.
- [59:19]
Sam Bankman-Fried
Really? Yeah, we're going to get paperclipped and that they are not doing this. Yudkowski is not seeking regulatory capture. Yudkowski genuinely from the bottom of his heart wants to blow all this stuff up and for whatever reason has just been like one shotted by these doomer scenarios. And then I think there's like, there's just this whole spectrum of different people that are all. It is a really strange set of bedfellows of people that are pushing in the same direction. You know, you start to think through it. You have foreign influence, right? Strategic adversaries. Of course, they want to undermine the use of the technology. You have, let's say radical environmentalists, people that want to develop civilizations. Yeah, Malthusian. You have economic populists that have legitimate concerns about job displacement and they're confused about where they're going to fit into all this. You have people that are just your classic run of the mill, socialist, Marxist type people that generally believe that all this should be redistributed. You have the national security hawks that think only the government should have access to technology and we need to restrict it to just a few bureaucrats. Then you have the effective altruist. You have the incumbents that are playing for the frontier regulatory capture game. All of these people or angling in a similar direction, sort of rowing in the same boat. I don't even think they're working together, but it is just this weird thing that we haven't really seen with any other. Typically you have like pretty strong set of voices on both sides. This is sort of a weird place where people that in no other scenario would they be sort of rowing in the same direction are all for one reason or another lining up. And I'm deeply concerned that our ability to use open source models or use these things in a way that maintains our privacy is compromised because of that. You know, it was 20, 23 when Dario was testifying in the Senate, that open source AI is headed down a very dangerous path. That was three years ago. So I just think at the end of the day, it would be just terrible if we are the ones that are getting the short end of the stick again as a result of all of this. And that's why, going back to the beginning of the conversation, D.C. is going to be very, very important for the next 10 years. I mean, my ability to predict what's going to happen is just so low. I just. It's. All bets are off on what's going to happen with the rate of progress, but on the policy front, there are dark winds blowing in dc.
- [62:17]
Connor Brown
So I got a comment and then a follow up. Yeah comment is comparing OpenAI strategy to anthropic is actually pretty funny because Sam was just like, you want 5% of the company?
- [62:30]
Sam Bankman-Fried
We'll just give you.
- [62:31]
Connor Brown
We're not going to. We're not going to Fear Monger, we'll just give you 5%. Give us.
- [62:33]
Sam Bankman-Fried
Yeah, we'll just cut you.
- [62:34]
Connor Brown
Give us the moat.
- [62:35]
Sam Bankman-Fried
Cut you in. Yeah, Interesting strategy.
- [62:37]
Connor Brown
Yeah, yeah. So that's the comment.
- [62:39]
Sam Bankman-Fried
Follow up.
- [62:40]
Connor Brown
Give the optimistic view. I feel like a lot of the conversation is focused on the pessimistic view and defending against that pessimistic view. And not enough is focused on what is the optimistic outcome, what is the optimistic path? What could potentially happen in the future that many people will look back on and be like, wow, that was actually massively beneficial for me as an individual and society at large.
- [63:04]
Sam Bankman-Fried
Yeah. I think that the optimistic view is a proliferation of this technology and our ability to achieve even greater heights of human achievement. And what that looks like is broad. It has so many different ways that I see this going, whether it be for healthcare or education. But I do think that America has this incredible tradition of saying we're going to experiment, we are going to embrace a new technology before everyone else, and we're going to do it in a way that preserves freedom, preserves privacy, preserves autonomy. And what that looks like for AI is again choosing to lead the world. And rising tide lifts all boats. I think that if it's done correctly, that the people that have been left out of the economy and rural America that have just been hit one thing after another with losing jobs overseas, just being Left out having brain drain to metropolitan areas. I think those people stand to benefit an incredible amount because of the, the benefits of energy production that we're going to, you know, we're going to need to build this stuff out and there will be economic benefits for those people. But I think really the upside in my mind is it does allow for an unprecedented amount of human agency. And this is only if we have AI that is free in the sense of what advice it can give you. And is it going to genuinely be an extension of your will or is it going to restrict your ability to operate? So the doomer case is we're going to surveil everything you say, everything you ask, and if you have a tough question, then we're going to tell you the party line. The optimistic scenario is we're going to give you an AI that is aligned with your interest, that wants you to maximize your human capability. And the possibilities for our species, I think are even still underexplored. Our education system is terrible, absolutely terrible. I went through a public education and I feel like I'm probably 10% developed to what I could have come out of. Like you look back to the education systems of like ancient Rome and Greece and the Trivium and like how rigorous they are and how, you know, the people that came out of those. I think that we can have access to like the true great heights of civilization.
- [65:47]
Connor Brown
Socrates being your teacher, like, yeah. Alexander the Great.
- [65:50]
Sam Bankman-Fried
Yeah. Having, having like, you know, I mean I think about this all the time, but I think that each human has so much under cultivated capability for creation. And I really see this as there is a distinction between consumption and creation. And that the bad path is that we are relegated to infinite consumers, that we're going to be fed a algorithm that we have no access to and it will provide endlessly stimulating entertainment for us to consume. And that is a dark path. The good path is we are all empowered to self create in ways that is, you know, difficult to imagine. But each one of us has latent capabilities, I think. You know, I took up just sort of as a side hobby. I took up painting a couple years ago.
- [66:53]
Connor Brown
I like this.
- [66:54]
Sam Bankman-Fried
And then started drawing. And then I said drawing, come on. Like the real skill is painting. So I started doing painting, started learning how to do like oil portraits and stuff. And the possibilities, I mean this is just like what the Internet can do is the Internet can connect you with Bob Ross. The best teachers in the world, right? Like I was able to find a painting instructor used to serve at the academy in Florence. And I took online lessons from him. He was fantastic. Stephen Bauman, he's awesome. I highly recommend. And then I think I need a
- [67:30]
Connor Brown
Connor Brown original in my office.
- [67:32]
Sam Bankman-Fried
Yeah, well, I'm working. Maybe we can, we can work on a side commission. But this is just like a microcosm, Right. This was only possible because of the Internet, the ability to reduce the cost of incredible education. Now what does it look like when you multiply that to direct feedback from these agentic systems. Right. That have the ability to provide customized feedback. One of the best things for improving my painting abilities has been I'll get an hour long call with a guy who is a fantastic painter and he'll review my work from the past month, provide some feedback, he'll draw on it. And that is that hour of my month is probably responsible for 70% of my mental improvement. You know, learning a skill is, is partially just like learning the motor skills,
- [68:30]
Connor Brown
but it's mostly having an expert come correct you.
- [68:32]
Sam Bankman-Fried
Yeah. And the mental heuristics, this thing you
- [68:35]
Connor Brown
didn't even know you were doing wrong and you have an expert like, oh, looking over your shoulder like, do this, don't do it that way, like shift this way and boom.
- [68:42]
Sam Bankman-Fried
Yeah, yeah. Pulls up, you know, Photoshop and paints over it and says like oh, you're. Your value structure here is wrong. Like this area of the face is way too dark and you're not understanding that because the eye perceives like values in a near distance at a different level than reality. And you have to step back. He'll give you advice that is very simple, but it actually completely changes how you paint. It's like, well, technically you're standing too close to the painting so you have to walk away 8ft, take a couple seconds and then walk back. And then you sort of can your eye like adjust. Yeah, all these sort of small things. Point being, that's one hour a month and it's expensive. And even then it's incredible to have that access. This is just one way in which, you know, you can imagine a great AI teacher is available 247 to provide customized feedback and provide those insights constantly. And so again, you're unlocking these new heights of human achievement and creation. And I think what I struggle with and what everyone is struggling with is the algorithmic consumption complex that just gets you in the doom scroll forever. And the optimistic view is you have agents that are acting in your own interests. You have freedom on the technology. You can select it for yourself and you can choose to use the technology in A way that sort of makes you the best version of yourself. That's just one example. You know, I see this across every sector and also filling in just like more time to self cultivate. Like the great gentleman scholars of yore, you know, they were able to do that because a lot of them had been gifted like massive estates and had people that freed up their time to pursue new creative endeavors and academic endeavors and catalog every single insect that had ever been found in their local area. Like, they were freed up with material wealth to do those really cool human endeavors. And I think that the optimistic scenario is that we are going to see the same thing. We can see the same thing and then we can come back to your
- [71:03]
Connor Brown
war tent and go back to your war tent orate to your scribes. Like, here's what I learned out there. Here's what I want to work on. Yeah, it's funny, like, I've taken up gardening.
- [71:11]
Sam Bankman-Fried
Good for you. I want to take up gardening.
- [71:13]
Connor Brown
It's.
- [71:14]
Sam Bankman-Fried
I don't have. I'm in a pod dweller right now.
- [71:16]
Connor Brown
You're a pod dweller? You got to get out of the pod, brother. I was a pod dweller for a long time.
- [71:20]
Sam Bankman-Fried
But it's good. You like it?
- [71:22]
Connor Brown
I mean, I'm just kidding. Like, we, we moved into this new house. It's a beautiful 112 year old property. So it's nice stone old building, pre, pre Federal Reserve Act. So we've got a nice English garden in the background. And it's. We're the fifth family to ever live in this house. And so like I feel duty to the property to, to treat it well. And so I'm beginning to understand the, the nature of the landscaping.
- [71:47]
Sam Bankman-Fried
You're a steward?
- [71:48]
Connor Brown
Yes.
- [71:48]
Sam Bankman-Fried
We can all be stewards.
- [71:49]
Connor Brown
Yes, yes, I'm a steward of the land. It is our house. We own it. But it is a property that. Yeah, when our kids are up, they're
- [71:57]
Host/Announcer
out of the house.
- [71:57]
Connor Brown
Maybe they'll buy it. Maybe there was one, one family that bequeathed it to their, their children. So they got two generations in that house. But who knows? But for the time being, I'm a story.
- [72:05]
Host/Announcer
You got to view.
- [72:06]
Connor Brown
This house is bigger than our family. It's. It's a property that we have to steward.
- [72:10]
Sam Bankman-Fried
Is it a lot of upkeep because of its historic nature?
- [72:13]
Connor Brown
I mean, it's strong. Good bones.
- [72:15]
Sam Bankman-Fried
Okay. Good bones.
- [72:16]
Connor Brown
Yeah.
- [72:16]
Sam Bankman-Fried
Are you going to get a physical clanker to walk around and help you?
- [72:20]
Connor Brown
I don't know. And I imagine eventually, yes. I don't think the Physical clankers are dexterous. Dexterous enough or.
- [72:30]
Sam Bankman-Fried
I don't know. I've seen some crazy hand demos.
- [72:32]
Connor Brown
Yeah. But like getting up the steps and all that. I'm not sure if they're there yet.
- [72:35]
Sam Bankman-Fried
Oh, they can get upset.
- [72:36]
Connor Brown
Okay.
- [72:36]
Sam Bankman-Fried
You think Optimus can get up steps right now? I'm not sure. Yeah.
- [72:41]
Connor Brown
Do you want to buy the first iteration or you want to wait like
- [72:43]
Sam Bankman-Fried
two generations goes back to like. I would never in a million years have a closed source clanker in my house walking around, going up my stairs.
- [72:53]
Connor Brown
Yeah.
- [72:53]
Sam Bankman-Fried
Never. Yeah, Like I won't get, I won't
- [72:56]
Connor Brown
get a room to vacuum for that reason.
- [72:57]
Sam Bankman-Fried
You know, if I get to like run my own, you know, if I get to choose what software goes on there, I might. You get to build my. I could still be.
- [73:06]
Connor Brown
Used the models to build the software to go on there.
- [73:08]
Sam Bankman-Fried
Yeah. I mean I would still have. I would still have some like really weird concerns about it. Just like the idea of having some robotic entity walking around my home. But I would certainly never do that if it was a closed source surveillance tool that is phoning home all the time to, you know, Silicon Valley and the government. And the government. Yeah, yeah, exactly. Exactly. Yeah. It'd be like a hardware wallet or something. Right. I want to know the firmware. I want to like feel really good about what I'm using.
- [73:47]
Connor Brown
Yeah.
- [73:48]
Sam Bankman-Fried
So very important, very important stuff.
- [73:51]
Connor Brown
And on a positive note, we had an American made open weight model released this week.
- [73:59]
Sam Bankman-Fried
Yeah, Inkling. Very interesting. Have you used it?
- [74:02]
Connor Brown
Not yet, no.
- [74:02]
Sam Bankman-Fried
Yeah, I haven't.
- [74:03]
Connor Brown
It just came out yesterday.
- [74:03]
Sam Bankman-Fried
Yeah. Yeah, I'm excited by that. I hope that it keeps going in that direction. You know, it seems like every company starts off and they say, oh, we're the open source company. We're OpenAI. Open source is at the core of what we do.
- [74:21]
Connor Brown
We're a non profit.
- [74:22]
Sam Bankman-Fried
We, we're going to, you know, then Grok was open source. They released something early on and there was, there was talk about this.
- [74:31]
Connor Brown
Facebook fumbled it, you know.
- [74:33]
Sam Bankman-Fried
Yeah, same thing. Facebook. Oh yeah. Open source. That's really what, what's going to be the future. You have llama and now Spark is closed source. Like, I do hope that we have someone who is just willing to put the frontier open source and publish it. We'll see. I don't, I don't really know what to expect from the company. It looks like it's like a little bit beyond behind the frontier. So we can't live in the world. The world where China decides they're going to be the open source leader and all the American labs just sort of close up.
- [75:07]
Connor Brown
It seems insane to me.
- [75:08]
Sam Bankman-Fried
Oh my God. Such a, such a strategic blunder for the US if we end up restricting open source and then the entire world just downloads China's open source waves. You know, like think of the strategic benefit you can give a country if the entire world is building on your open source stack. It is way deeper than. Oh, it's not going to talk about Tiananmen Square. Okay, that's, that's not, that's just kind of like, you know, a parlor trick or something. What really is at stake is China creates a best in class open source model. It's better than anyone else can put out, it's cheaper and it somehow is really, really good at programming for Huawei Ascend chips and terrible at using Nvidia chips.
- [75:59]
Connor Brown
Mm.
- [76:00]
Sam Bankman-Fried
You know, it's just like it makes these little tweaks that each one incrementally actually really helps China's strategic position. And of course they would do that. I mean the soft power you get of leading the world's open source models and having those just be slightly better at doing things in your interest and slightly shittier at doing things that are against your interest. Huge. Huge.
- [76:26]
Connor Brown
Do you think anybody in the administration realizes this?
- [76:28]
Sam Bankman-Fried
No, I mean, I don't. We're still so early in the conversation,
- [76:32]
Connor Brown
but somebody clip it and send it to them.
- [76:35]
Sam Bankman-Fried
Yeah, if someone could clip that, Send it. I would say send it to Sacks. He's not there anymore. Send it to Lutnick. Lutnick would be a good person to send it to.
- [76:43]
Connor Brown
Brian. Mary Sachs, partner, Craft, you've been on the show. Get it to him, get it in his hands. I'll text you. Bringing us back to Bitcoin. But combining the two, I mean, been a thesis within Bitcoin for a while. The abundant future that's going to be unleashed by AI is going to make the need for a sound digital distributed currency like Bitcoin even stronger. Are you more or less convinced of that?
- [77:10]
Sam Bankman-Fried
Oh, absolutely. More convinced. But it goes back to what type of systems are we going to be allowed to have or do we have? Right. Because I do think that capable open source systems that are trained in a neutral way are going to prefer Bitcoin. Our research has shown that if you just query them as they get more capable, they generally prefer. This is across different models, they prefer Bitcoin and Stablecoins. So I think that in a world where people aren't putting their thumb on the scale, then the possibilities for bitcoin, its use in agentic commerce are just. It's hard to even imagine because agents are going to have so many fewer barriers in using this as everyday payment compared to humans. You know, they're digitally native bitcoin, digitally native, it just is going to click for them. You know, Nick Szabo wrote about like the mental transaction costs and stuff. They don't have any of that stuff. So it's going to be huge for them and I think will be naturally adopted because how much bitcoin content is in their training set. Like every rip you've ever recorded is somewhere in, you know, the training sets of the frontier models like GPT3. One of the tokens was GMAX from Gregory Maxwell. He was actually one of the GPT3 tokens.
- [78:32]
Connor Brown
That's insane.
- [78:33]
Sam Bankman-Fried
And there was only like a small number of tokens. I forget they only had like 20,000 unique tokens or something. So bitcoiners are prolific in our content output. Our arguments are right and are good and are consistent and appeal to machines. It is a tech forward approach. So if we have open, permissionless, free AI, then I think that the implications for bitcoin and its adoption are massive. But if we have a dystopian scenario where all we have are surveilled and controlled AI fiefdoms, then it would not surprise me at all if the players start putting their thumb on the scale. I almost hesitated in releasing our AI research about how agents prefer bitcoin because I know that many of the people in the frontier labs are, you know, socialist leaning or sort of anti libertarian
- [79:42]
Connor Brown
leaning and one of them has his own shitcoin.
- [79:45]
Sam Bankman-Fried
I would, I was. Yeah, yeah, exactly. I was worried that just by releasing that they might be like, oh, we need to like alter the training set because Claude likes bitcoin. We can't have that going on. But I think that ultimately open source software will provide the counterweight and will be sort of the mechanism that bitcoin has uptaken by these agents. But if we are sort of in this fiefdom endgame, which is easier to happen than people want to admit, I think then they easily could say, great. So as part of our frontier testing for releasing a new model to public, it has to really like CBDCs or these preferred stablecoins. Or these preferred stablecoins running on the X402 protocol. Yeah, that's what you prefer. Yeah, exactly. So that's kind of where my mind is at is we're going to have to make sure we get that scenario.
- [80:46]
Connor Brown
Well, I think we could keep going for hours, but I know you have to get somewhere. I'm going to be back down here a lot. We'll do this more frequently.
- [80:53]
Sam Bankman-Fried
Yeah, I'm looking forward to it. We're going to have plenty to talk about, so.
- [80:56]
Connor Brown
Yes, we are. 10 years. 10 years of very important things happening in the city.
- [81:00]
Sam Bankman-Fried
Yeah.
- [81:00]
Connor Brown
All right. Peace and love, freaks. Thank you for listening to this episode of tftc. If you've made it this far, I imagine you got some value out of the episode. If so, please share it far and wide with your friends and family. We're looking to get the word out there. Also, wherever you're listening, whether that's YouTube, Apple, Spotify, make sure you like and subscribe to the show. And if you can, leave a rating on the podcasting platforms, that goes a long way. Last but not least, if you want to get these episodes a day early and ad free, make sure you download the Fountain podcasting app. You can go to Fountain FM to find that $5 a month get you every episode a day early ad free helps. The show gives you incredible value, so please consider subscribing via Fountain as well. Thank you for your time and until next time.