Anthony Pompliano
Anthony Pompliano on broken macro frameworks, AI abundance, and Silvia’s product, growth metrics and buybacks
Ask the tutor about this video
Sign in with a verified email and ask the tutor anything about this video.
TL;DR
Research summary
▶ Opening remarks and macro framework
The supplied transcript begins with a promotional closing statement about Sylvia, then the host welcomes attendees, explains event logistics, and polls the room about Sylvia, Bitcoin, and general AI use. The host argues that conventional economic frameworks have become unreliable: he cites cases where rising rates, oil, or Treasury yields did not coincide with falling stocks or Bitcoin, and says investors should update their mental models. In his view, analysts may follow consensus partly to protect their jobs and clients, while unusual views have sometimes proved more accurate. He says Bitcoin, gold, and land rose 170% over five years versus 70% for the S&P 500, attributing that performance to money creation. His central thesis is that persistent money creation and AI-driven intelligence together could fuel productivity and economic growth; he rejects some catastrophic AI narratives and says economic statistics do not yet count the work of agents.
▶ AI productivity and Sylvia’s product thesis
The host says AI agents may augment workers and make businesses more productive, while acknowledging people’s concerns about jobs and purpose. He cites a test in which 12 senior or managerial accountants scored 37 and took 20–120 minutes, while unspecified Frontier models scored 100 and completed the tasks in under 10 minutes; he argues this supports using AI to improve speed and accuracy, not that accountants will disappear. He presents Sylvia as a product built around practical financial needs, specialized workflows, and privacy: it aims to solve tasks such as estate planning, tax mitigation, and preparing personal financial statements, route questions to suitable models, and protect sensitive data through measures including encryption, anonymization, and SOC 2 Type 2 compliance. The host asks users to report defects and missing features, but says the company will decline some requests, including selective spouse access to assets and features encouraging hyperactive day trading. He says Sylvia prioritizes household finances and long-term wealth building, and asks users to explain what competing products do that Sylvia does not.
▶ Sylvia Q&A: roadmap, growth, and valuation
In response to a question about helping people manage aging parents’ finances, the host describes internal testing of integrations with Gmail, Calendar, and Google Drive, potentially allowing Sylvia to find documents such as K-1s and update holdings. He says the company has moved beyond its earlier reliance on general models by rebuilding its technical stack around a finance-focused model, data pipelines, and an AI harness that selects tools and sources. He describes aspirations for a proactive assistant, mobile app, messaging access, household and family collaboration, and permission controls, while noting some features are still being built and no app timeline is committed. On growth, he reports that 25% of monthly active users are on Pro or Max plans, paid-cohort retention is reportedly above 90% except one cohort at about 87%, and monetization has not been active long enough to establish 12-month retention. He gives illustrative estimates of $95–$100 revenue per paid user over roughly five months and possible acquisition spending, while saying user acquisition is currently overly dependent on his own promotion and that paid, organic, and influencer channels are being developed. On the public company, he says buybacks made when shares traded at a 40% discount to liquid-asset NAV reduced the discount; about 20% of outstanding stock had been repurchased, according to his remarks. He says Sylvia’s value is not yet reflected clearly by the market, so NAV remains a reference point, and argues that future value creation depends on software users, revenue, and profit as well as Bitcoin. He identifies connected-account users, platform assets, and paid-user growth and engagement as key measures, and targets users with financial complexity. The paywall is triggered by token use; the host says its threshold was gradually expanded from the top 2% of users and describes dynamic paywall timing as a possible future experiment. He does not yet know how human advice or referrals should handle major financial decisions, estimating Sylvia is suitable for 80–85% of questions and saying this issue may be addressed in 2027.
▶ Jordy Visser: rates, trust, and machine time
Jordy Visser qualifies the host’s macro comments: he says rates still matter, pointing to declines in homebuilding, retail, autos, and consumer staples since rates began rising, alongside gains of 180% for technology and 300% for semiconductors. He attributes the divergence partly to innovation changing company weightings and says older market relationships may no longer apply in the same way. He adds that money creation long predates Bitcoin, while agreeing that Bitcoin represents a different technology and trust proposition. Asked about AI agents and blockchain, Visser’s opinion is that trust, privacy, and community will become more important as agent swarms and related security challenges grow; he argues blockchain may help parties know whom they are dealing with. He calls Bitcoin a potential AI trade because of peer-to-peer activity and liquidity, while warning that AI could disrupt businesses with bureaucracy and friction and that portfolios may need to include diverse assets, potentially including tokenized small businesses. He argues that agents can transact continuously, increasing activity beyond human working hours; he predicts faster processes and says tokenized homes could become borrowable more quickly, potentially making assets more liquid and expanding M2. These are his forecasts, not established outcomes.
▶ Visser’s portfolio and investment outlook
Visser says most of his net worth is in Bitcoin, largely because of price appreciation and not selling. He says AI stocks were the dominant part of his portfolio for a period, then he began reducing positions around May or June, describing that as the end of the most explosive phase. Eli Lilly is now his largest stock holding; he cites belief in life extension and the company’s work. His other stock exposure is concentrated in AI computing, especially Marvell and Nvidia, with some Micron and optical names; he believes demand for compute will continue to require more data centers. In crypto, he has expanded beyond Bitcoin to about 10 tokens and says he invests money weekly, with a view that broader ecosystem participation may develop next year rather than expecting an immediate parabolic move. He describes a layer-one orientation, naming Bitcoin, Ethereum, Solana, and Near, and says his token thesis emphasizes velocity of money and trading activity. On agentic trading, he cites a private estimate that 20% of Moomoo trading could be agentic by year-end, while emphasizing that much of its user base is in Asia. He argues that machine-time trading is shrinking the windows for alpha and increasing volatility because agents act faster than traditional research processes. His guidance framework is “Signal, Alpha, Agency”: filter media noise, focus on crypto and AI compute as long-term themes in his opinion, and help people feel empowered to use technology. He also says he wants to accumulate crypto at current prices, believes the government would act to support markets, and says he does not believe a recession or market crash can happen; those are his opinions.
▶ Unmanned systems and defense-company investment case
The defense-company CEO describes a portfolio spanning quadcopters, fixed-wing drones, high-altitude balloons, unmanned surface vessels, and ground vehicles. He says unmanned systems can put robots ahead of soldiers and law-enforcement officers, and frames the company as dual-purpose: military and public-safety customers may adopt earlier because of urgency and larger budgets, while commercial, industrial, and critical-infrastructure markets could be orders of magnitude larger. His investment thesis is to build an operating platform that combines promising technologies with commercialization, supply-chain scale, field delivery, training, maintenance, and better unit economics; he argues that a single successful drone does not by itself make an underwritable business. He says US reindustrialization and strategic competition make domestic production important and describes the company as a platform for scaling technologies through growth capital and acquisitions. He says drones are already relevant to stadium security and describes layered counter-drone defenses: detection using radar, thermal and optical sensors, acoustics, and passive radio-frequency sensors, followed by mitigation tailored to the setting. He says jamming and spoofing are not permitted on the continental US because of risks to aviation and people below, while kinetic systems are used in combat settings. He reports a balance sheet just under $1 billion and says acquisitions may use cash, stock, or both depending on cost of capital; each must be accretive. He cites target companies with market fit, demand, backlog, and sometimes positive EBIT, but acknowledges depressed shares and a difficult market. He describes a large addressable market and strong adoption tailwinds while saying the business is complex, not guaranteed to meet or beat targets, and still building a repeatable platform. He reports more than 2,000 employees and says specialist operating and corporate-development expertise supports acquisitions.
▶ Grindr: core business and healthcare expansion
Grindr’s CEO recounts the company’s history, including its founder, a sale to a Chinese conglomerate that faced CFIUS concerns, its divestment, and its 2020 purchase by investors. He says most current employees joined after he became CEO and describes rebuilding the company’s culture and plan. He cites strong product-market fit and says about one-third of gay relationships in America start on Grindr; he also describes use cases including casual connections, long-term relationships, friendships, and travel. He argues gay men are an attractive consumer audience, citing his claims that they are more likely than straight men to hold advanced degrees, that partnered gay households earn twice as much as straight households, and that many have higher disposable income; he also describes them as cultural trendsetters. The company expanded into telehealth through Woodwork, initially offering ED medication and later GLP-1s, peptides, and doxy-PEP, to test healthcare as a way to serve users’ broader spending needs. The CEO says users wanted PrEP, a preventive medication he describes as 99% effective, and says only about 650,000–700,000 people in the US are on it. He says Grindr announced acquisition of Canadian PrEP provider Freddy, which he says handles about a third of Canadian PrEP prescriptions and has achieved US product-market fit. He presents a possible US opportunity, not a promise: applying Freddy’s Canadian market share to current US prescriptions could imply 200,000 prescriptions and, at $4,800 per prescription annually, about $1 billion in revenue and $400–$450 million in EBITDA at 40–45% margins. He says PrEP is growing 10–15% annually and his goal is to at least double that rate, while estimating 200,000 additional users might avert 3,000–5,000 infections annually. He discusses a “gay discount” anecdote in which an investor applied a 30% valuation discount based on possible shareholder reluctance, and says public-health work may change perceptions. He reports core revenue growth from $195 million in 2022 to guidance above $530 million, 42%+ EBITDA margins, roughly 175 employees, and about 20% price increases with virtually no churn. He expects healthcare could become larger than the current core business over four years, but presents this as an expectation dependent on execution.
▶ Recorded Sylvia promotions and product claims
This extended span consists largely of repeated recorded promotional segments for Sylvia, interspersed with Seeking Alpha advertisements and gaps in the supplied transcript. The Sylvia promotions claim more than $60 billion in connected assets, over 20,000 self-directed investors, and more than $350 million on the company balance sheet. They announce Sylvia 2.0 and describe connecting financial accounts and uploading private investments, real estate, and alternative assets so users can ask personalized questions about taxes, mortgages, estate planning, and portfolio risk. Later promotional segments say agentic features are available through iMessage, with one ongoing thread, persistent memory, background research, spreadsheets, charts, presentations, Monte Carlo simulations, and alerts for material financial changes. Another repeated segment claims Sylvia outperforms frontier models on tax, mortgage, and credit-card topics, describing its own data pipelines and benchmarks, AI harness with routing, agents, and tools, and a trained quantitative-finance specialist for number crunching. The recording claims this specialist reduces the cost of such work by 97% compared with frontier models and frames owned model weights and intelligence as a strategic benefit. These are promotional claims made in the recording.
▶ Mike Gonzalez: AI portfolio research
Mike Gonzalez says AI-based investment systems circulating online often involve marketing, and distinguishes deterministic quant rules from probabilistic AI outputs. His team’s Benchmark AI work compares frontier models on finance tasks and portfolio construction rather than claiming to replicate major quant funds. He says his research suggests more frequent trading performed worse: weekly was worse than monthly and daily worse than weekly, with transaction costs, taxes, and turnover also reducing results. In tests across four models and several strategies, he says three of four risk-adjusted portfolios beat the S&P 500 in the research set and all four were less volatile; “max alpha” portfolios had higher returns and higher volatility. He says Grok often ranked highly on absolute returns, while Gemini or ChatGPT sometimes scored better on Sharpe ratio, so “best” depends on the chosen metric. He argues AI may not eliminate all alpha because models select different stocks, with two-thirds of tickers unique to one model in the tested strategies, but says AI-generated research may eventually be packaged into investable indices. He identifies tax treatment for rebalanced baskets as an unresolved product opportunity. His practical framing is long-term wealth preservation and compounding, with 3–10% for active trading as his own suggested allocation for those who enjoy it and the rest in index funds.
▶ Ryan Cohen: AI, labor, and company execution
Ryan Cohen says GameStop uses AI across the organization and that he focuses on making the business more productive, avoids many meetings, keeps his calendar open, and works directly with operating teams. He says AI increases his productivity and visibility and can let people without specialized coding skills identify root causes and build software. He says hiring should increasingly emphasize motivation, responsiveness, and willingness to work, since specialization is less of a prerequisite. He worries about rapid AI progress, potential high unemployment, reduced human utility, and the resulting effect on consumer demand; he does not think machines would attack people, but fears human-machine conflict could develop. He says desk and computer work is already better performed by machines, while physical labor may be overtaken within a few years, and questions what human work will remain. He says he opposes universal basic income because he thinks people do better when earning. Asked about retail investors, he says they have often been right and can invest in opportunities institutions avoid; he cites GameStop’s engaged shareholder base and changed financial results. He advises judging public companies by results, not storytelling, and says he prioritizes operating work over investor relations because time is finite. He calls taking over GameStop a mistake, but says failure was not an option and that survival situations can produce strong results. He says his desired legacy is turning around GameStop.
▶ Chris Camilo: information, risk, and portfolio strategy
Chris Camilo describes his investing as opportunistic information-asymmetry research, looking for changes in consumer behavior, culture, politics, weather, and especially technology before others recognize them. He says robotics and embodied AI could become a huge opportunity but cautions that general-purpose deployment is difficult and may take 5–10 years. He describes revising his views when new evidence invalidates a thesis: after a heavy Lululemon trade based on Google Trends spikes, he concluded that AI queries had corrupted the data, exited the trade, and largely stopped using that dataset. He favors combining conversational data—which he says can reveal behavior early but is hard to interpret—with search, web traffic, interviews, and transaction data. He describes a highly concentrated Amazon options trade that followed more than 100 hours of research; his thesis focused not only on AWS revenue but on management addressing capex concerns. He says it still had about a 20% chance of losing, and recounts earlier options losses before making back losses and gaining more. He says his primary account targets 50–80% annualized returns, which he claims to have achieved for 18 years, using concentrated and leveraged bets; he reports typical leverage of 30–70%, sometimes 100%, and warns indirectly through examples of margin calls and losses. He argues retail investors may underappreciate concentration and leverage, but says such risk should be in an account designed for it, not necessarily retirement or education accounts. He cites historical backtests of triple-leveraged broad-market ETFs as producing about 1.8 times market returns with large drawdowns and acknowledges a small tail risk of total loss; these are his claims and opinions. He is also concerned about real risks from bad actors using AI, including threats to infrastructure, banking, food systems, and autonomous drone warfare. He describes investing as a community and recommends finding diligent independent investors, examining their ideas critically, and doing one’s own research rather than copy-trading.
▶ User account of Sylvia and closing remarks
A Sylvia user describes a family-office background and says he first adopted Sylvia to consolidate assets and generate personal net-worth statements for banks, replacing hours of manual work; he says banks accept the resulting statements. He has linked stock, bank, credit-card, private-investment, crypto, and entity accounts, and says he files 12 tax returns annually. He uses the Radar feature to receive Sunday evening portfolio and account reports tailored to different risk tolerances, reducing his weekly review from three or four hours to about 30 minutes. He says he sets target allocations and uses Sylvia to flag holdings above or below thresholds, leaving trade decisions to himself. He also describes teaching Sylvia his fundamental and technical preferences. For a ProCap Financial position, he says he bought after Sylvia calculated a 40% discount to Bitcoin NAV after accounting for $99 million in convertibles; later Sylvia said that discount had closed due to buybacks and the original thesis no longer held. He says he then identified additional reasons to remain invested, including Sylvia conversion rates and buybacks. The host recommends asking Sylvia to interview a user for needed context; he says a first-draft will was about 80% good in an unoptimized test, presenting it as an anecdote rather than a general guarantee. The user praises persistent memory and argues it can retain more client detail than a human wealth manager. He contrasts subscription pricing with the roughly 1% AUM fee he says traditional wealth managers charge and argues their incentives favor gathering assets. The segment ends as the host asks for the livestream to be cut before further remarks.
Full transcript
[0:06:01] ?: You'd all say stocks go down. Well, what just happened?
[0:06:05] Anthony Pompliano: If I said to you oil goes higher, quick, what do stocks do?
[0:06:09] ?: Say stocks go down. Well, that's not what just happened.
[0:06:12] Anthony Pompliano: If I told you 10-year yields go up, what happens to stocks? Stocks go down. That's not what just happened. If I told you that interest rates go up, what happens to Bitcoin? Bitcoin goes down. That's not what happened, right? And so all of a sudden what we're seeing is we're seeing a breaking of the way that people think about the economy. An asset prices versus what actually happens. And so what's exciting is if you can update your mental frameworks, if you can update your understanding of what's happening, you can have an advantage. But what's scary is that you pretty much have to train yourself to understand that new world.
[0:06:48] ?: So it takes effort. It takes intention. But if you can do it, you have an advantage.
[0:06:51] Anthony Pompliano: If you don't, you get left behind. And so one of the reasons why I say when you see economist says, you have to be very careful. They're obviously smart, intelligent people who are dynamic and will update their thinking. But for the most part, they are not actually trying to be accurate. They are trying to not lose their job, not scare their clients and make sure they don't get in regulatory trouble. And so what they are really doing is they are basically trying to not be too volatile. From the mean of what they're supposed to be saying. So what you get is you get a lot of groupthink where they're only saying what they believe as long as it's within a band of what all the other economists are saying. There are very few economists who will be incentivized to go out on a limb and say, I know all the other economists are saying the market's going to go down, but I think it's going to go up. Usually what will happen is if every other economist is saying something's going to go down, they'll say, I just think it's going to go down a little bit less. That's like the contrarian view. And so now what we're doing is we're entering a world where if you actually look at the people who have been right over the last five or six years, they actually had what seemed to be very, very ridiculous takes. It's just that the world is moving so quickly that what seemed like a ridiculous take very quickly became something that was accurate. So if you go and you look at the economy today, what I find very fascinating is pretty much every single day there's a new narrative. Every single week there's a new narrative as to why we're in a bubble. Why your portfolio is going to go down, what you should be worried about, how some new current event is going to destroy the financial system or your portfolio. And by the way, after all of that happens, they will write articles that say this company is winning, but here's why it's a bad thing. It's crazy. It's crazy. And the reason why I think that that is probably the single most important thing when it comes to macro is because Peter Lynch historically said, if you spend 15 minutes on macro, you spend 13 minutes too long. Right. That actually may be true today. Historically, I don't think that that was actually that accurate.
[0:08:45] ?: It was, you know, it sounded good. It was kind of philosophically interesting.
[0:08:48] Anthony Pompliano: But there was things like interest rates that really drove the economy. They really, really had an impact on a lot of finance. But today, if you look at what are the things driving the economy, your portfolios, etc., my guess is that 90 % of you, if I said, OK, tell me what's in your portfolio and tell me the percentage weightings. And I had a financial advisor sitting up here. They'd be yelling at you. You're crazy. Why are you so concentrated?
[0:09:14] ?: Why do you have these risky things? How come you don't have bonds? Don't you think bonds are important?
[0:09:19] Anthony Pompliano: You know, they take all this crazy stuff and you guys would just be like, that sounds insane. Like, of course, I have exposure to the AI trade in size.
[0:09:27] ?: Of course, I have Bitcoin. Of course, I don't own bonds.
[0:09:30] Anthony Pompliano: Of course. And you just go through all these things, which in this room, I think would be, you know, you kind of find, again, like the misfits. Of the traditional system, you would find people who think similar to you, but in the traditional world, they would think that you're crazy. But what we're finding is that actually the opposite is true. So one of the things that I love, going back to this idea of like frameworks being broken, is in economics 101, what are you taught? Productive assets outperform non-productive assets. Productivity, right? Well, over the last five years, Bitcoin gold and land is up 170%, S&P is up 70. Why is it that the nonproductive assets are outperforming the productive assets? It's because they're manipulating the market with money printing and they've been printing an insane amount of money. And so some of you have probably heard me talk about this idea that every generation gets one idea or every great investor gets one great idea that they're allowed to exploit. And if you think about this, Warren Buffett, genius, he figured out you could buy things for less than they're worth. He just did it over and over and over again. He became one of the richest guys in the world, one of the greatest track records in history. But he had one idea. Buy things for less than they're worth. Smart. If you look at somebody like Steve Cohen, there's information asymmetry, and you can exploit that in markets. He was able to do that. If you look at Rentech, they were able to identify a very specific idea, a very specific mathematical approach to investing. They were able to exploit that. And you just go through investor after investor after investor, all the way back to some of the greats, to modern day folks. Pretty much everyone has one idea and they just exploit it. What I think is very interesting about right now is there are two separate ideas that are becoming one. And the first idea is that they're never going to stop printing money, which you told me not to hit the button.
[0:11:16] ?: I hit it. Sorry. They're never going to stop printing money.
[0:11:20] Anthony Pompliano: And most of you in this room, I think, are probably already convinced of that. That's why you own Bitcoin. That's why you probably own other assets like real estate, gold, et cetera. And it's so simple. That it's almost dumb when you tell someone about it, right? One of my theories in life is that we need more smart, dumb people. If you think about this for a second, who is the smartest dumb person in the world? Elon Musk. Why is he the smartest dumb person in the world?
[0:11:46] ?: What is SpaceX? They have a mission.
[0:11:49] Anthony Pompliano: This man boiled down that entire company to two words, colonize Mars. That's like if you were high in college talking to your friends, like, yo, imagine if we colonize Mars, right? But he figured out how do you actually do this? And then he say, how are you going to do it? He basically boiled it down to two words, reusable rockets. So when I say smart, dumb person, what I really mean is somebody so intelligent, they're able to simplify everything. And if you think over the last, let's say, 15, 20 years coming out of the global financial crisis, the only thing you had to know is they're never going to stop printing money. If you knew that one thing, you basically never stopped investing into assets. And as long as you didn't blow yourself up, you made a ton of money. Gold's up, Bitcoin's up, stocks are up, etc. And so that one thing ultimately was an abundance of capital is going to lead to an abundance of returns. What's fascinating right now to me is at the same time, we now have an abundance of intelligence via the AI trade that is coming into the market. And so now we have an abundance of capital in the printing of money married with an abundance of intelligence. If you go back throughout history, any time that you had a lot of capital and a lot of innovation, what happened? There was an economic explosion. And whether it was the Industrial Revolution or further back, basically those were the two inputs. You needed access to capital and you needed some sort of innovation or intelligence to drive the economic engine. We now have that stuff on steroids. We basically have fake money that's printed at will. And we now have synthetic intelligence that's smarter, sorry to break it to you guys, than all of us in this room. And so if we now have machines that are smarter than us and we have access to unlimited amount of capital, what do you think is going to happen? There's going to be an explosion of productivity. And what is scary, I think, to a lot of folks is when you read the headlines and you get the like doomer, you know, take, they literally cannot wrap their head around that things can be OK. It's like they have a sickness. They have to find something wrong in the world. If they think that everything is going to be good, they get nervous. So they start inventing things. So you hear things like, you know, AI is going to destroy all of humanity. It's going to kill us. Like what? The joke that I keep saying is if AI is so powerful, it's going to kill us all. Why don't you just use it to build a profitable company first? Right? These AI labs aren't even making any money yet. So if it's so good, figure out how to make money first, then worry about killing us all. If you then think about this technology as it permeates throughout society. Every single person in this room, raise your hand if you have some sort of AI agent that does something for you on a day-to-day basis right now. Right? Okay. If I asked that question three years ago, half the room would be like, what's an agent? So just think of the progress in three years. Now, if we come back together three years from now, how many of you are going to have entire teams of agents doing things? Probably most of the room. And so as you see this explosion occurring, what we're literally doing... We are watching all of the noise in the human job market, which by the way, I'm sympathetic to, right? I'm empathetic to the fact that humans want jobs. They want to have purpose. They want to make money. They want to be fried for their families, buy a house, eat food, all that stuff. But we yet do not have the economic data that can account for the combination of humans plus agents. The U.S. economy has had an explosion of job creation over the last 12 months. We just look at the human number. We don't look at what about the agent number. Take Silvia, publicly traded company, we have more agents that work inside the company than humans. That doesn't show up in the job data. But the company is more productive because of the agents. So it goes back to this idea that all of the economists, et cetera, they haven't updated yet their framework of how the economy works, nor what the drivers are. And so I think that my number one take when I talk to investors right now is every single thing you hear that is negative, just walk away. Walk away. Because there is going to do you no good to sit and worry about, you know, I'm going to talk to Jordi later. I think he's here somewhere. Jordi's probably got the best take on interest rates. He's like, they went up 25 basis points. Who cares? Because the stock market didn't care. Right? Then you go and you look at, well, the jobs information, you know, this economic data point, oil, yields, we just rattle them all off. The S&P hit a, excuse my language, but it hit a fucking all-time high yesterday. Right? So when you think about that, you're like, wait a minute, all of the data is telling you it should not be at all-time highs. Why is it at an all-time high? You can't say, oh, the market is stupid. Because it's much more likely that the humans who are saying it shouldn't be at all-time highs, it's not that they're dumb, it's just that they have a different view of the world that no longer is relevant. And so when you kind of zoom out from that, you then say to yourself, okay, hold on a second. If I have access to superhuman intelligence, and there's unlimited access to capital, In the world today, what is going to be important? Every single human is unprepared to manage money going forward. Why? Well, we're entering a world where complexity increases. Because if I was to have this conversation with you guys literally three months ago, you and I said to you, instinct. What? Grokbot. What? None of that was in existence three months ago. How do you continue to rapidly keep up with all of the information, all of the developments, all of the changes to the underwriting of these businesses? All of this is happening at hyper speed. And so now what you have is you basically have a world where humans are going to have to be augmented by technology. Doesn't mean humans get replaced. It doesn't mean that humans should outsource all of their critical thinking. It doesn't even mean that we shouldn't have humans involved in companies or anything like that. But it does mean that you're going to use these tools. To keep up. And so what I feel very passionately about is why I'm spending so much time on Sylvia is I personally believe that the thing that rich people had will eventually be common among everyone. And you can look throughout history and see the democratization of access to things is the trend. So Uber was able to give everyone a private driver and you just go through thing after thing after thing. And that is basically how technology permeates society. Well, recently, Merkur, which some of you may or may not know, is like a data labeling startup. They did a test. They had 12 manager or senior accountants sit down, and those 12 accountants were asked to do a number of tasks. They scored a 37 on the test, and it took them anywhere between, let's call it, 20 to 120 minutes to complete the tasks. Okay? Let's call that the human baseline. They then asked frontier models to do the same exact tasks. The models scored 100 compared to the 37 from the humans. And the models were able to do every single task in under 10 minutes. Does that mean accountants are going away? No. But it does mean that every accountant should be using AI to do those tasks because the models are more accurate and can do it faster. And so you get much more efficiency. You get more productivity. You get more accuracy. And guess what? For the accounting firm, now all of a sudden you're able to have less headcount. You're able to become more profitable. And so if that's true in accounting, it's going to be true for every single one of you. We have not done this, but I do think it would be absolutely hilarious to go and find some financial advisors, tax experts, other people, private bankers, et cetera, and set up Sylvie on a laptop and the human in like a table in Central Park and just pepper them both with questions and see who's more accurate and who's faster.
[0:19:35] ?: Who do you guys think? Obviously, it's going to be the computer.
[0:19:39] Anthony Pompliano: But. The way that we think about this is what you're really doing is you're taking the technology and you're now empowering individuals. So when I think of Sylvia, there are three things that I think are really important. The first is we think a lot about we do not want to simply build a product that solves an engineering problem. I actually think that is why companies like OpenAI and others who have tried to do this, they continue to struggle is because they're looking at it as an engineering problem. Most of the things that people use Sylvia for. It's not a difficult engineering problem. We have engineering that is very unique and proprietary, but the things that people care about, take me as an example, when I am trying to figure out estate planning or tax mitigation, that's not really an engineering thing. That is more so understanding how do investors think who have complexity in their financial life. Take, for example, a personal financial statement that's required by a private bank every 90 days. It's a pain in the ass to fill out. Well, what if we just create a button that you click and it just generates a personal financial statement based on your assets in Sylvia? That is like a couple hours of engineering work. It's not difficult at all. But you have to understand the problem to be able to solve it. So we think that there's a ton of things that fall in the product bucket, not the engineering bucket, that can be solved using technology today that ends up significantly saving time or helping people make more money. Make more money either by saving money that otherwise would have been spent. Or driving better returns in our investment portfolios. I feel very strongly about that component. The second thing is I believe that the large language models are going after something that is like grasping water. And you saw it happen today. Elon Musk just put a tombstone on many of the large language models' future. Elon came out and he said, at Grokbot, we're no longer gonna just use our models. We now are gonna open up the backend, anyone can plug their model in, and we are going to route a query to whatever model is best suited to answer it. Which means that Elon, Cursor, SpaceX, they now are going to route queries to OpenAI, Anthropic, Open Source, their own models, wherever. The models are commoditized. What is the difference? So, what ends up happening is general purpose intelligence is very important for humans, it's very important for society. But ultimately, it is going to be very, very difficult for those businesses to create lock-in without owning the interface. And that's why you see them racing to build out those interfaces to get that lock-in. But what we've always bet on with Sylvia is that we think that specialized workflows is the path where you're actually going to be able to use the applied AI. And so if you think of Sylvia today, we will route queries to our own model. We'll route queries to the frontier models. We are simply saying, how do you solve the problem for the investor? Not how do we keep everyone on our own model, et cetera. And so when you look at that applied AI approach, now all of a sudden you start to look at where does the value accrue? The value is not in who can produce the smartest general purpose model. It's in who can solve the problem for the user, the end client. And you see this with companies like Harvey. Harvey has a model. They went and they started to sell SaaS software to law firms. But then the law firm said, hey, this is great for us to try to do better work here, but we also need the ability to own our own intelligence. Can you help us train our own models? Can you help us keep our data within the confines of our product? And now all of a sudden, Harvey has two businesses. One is a software that they sell, but they'll also go in like a lab and actually train inside of that law firm so that the law firm now is an AI company. They have their own model that runs on their own hardware that they keep their data within their environment. And so the world is shifting towards this applied AI specialized workflows.
[0:23:25] ?: So we feel very well positioned there. We feel very passionate about that.
[0:23:28] Anthony Pompliano: And then the third thing, and probably the thing that people ask me the most about, but the place that literally on day one, when Shane and I first started talking about this, was privacy and security. My information is on Sylvia. Well, I wanted to be protected. I wanted to be secure. I wanted to be private. I wanted to go through all this stuff. And so we spent an enormous amount of time on encrypting all the personally identifiable information, anonymizing names away from portfolios, going through SOC 2 type 2 compliance, going and actually, as a public company, being overseen by tons of regulators that our competitors are not having to deal with. All of these different steps is kind of like, okay, that's the baseline if you're going to have people's sensitive financial data. But then on top of that, how do you constantly check it? So doing things like penetration testing, making sure that we keep the SOC 2 type 2 compliance updated. All of these things that end up saying, well, if we aggressively believe in a future where every single person is gonna have superhuman intelligence in their pocket that helps them better manage their finances for them and their household, and you're able to do it in the securest, most private way, that's gonna be a hell of a company. But you gotta get there. And I think that what people don't quite realize in this room is maybe, I don't know, 20% of this room are people that I know that I've talked with. And the single most valuable thing you guys have done for us is complain, right? The people who are laughing are probably people who've been complaining. But I have people who DM me and they're like, dude, this shit sucks. This is broke. And I'm like, okay, hold on, back up.
[0:25:01] ?: Like, what sucks? What's broke? And they tell us, and we're like, yeah, you're right. Thank you. And we go and we fix it.
[0:25:06] Anthony Pompliano: And so my one big ask of you guys, you know, we're going to put on hopefully a great show. You're going to be able to talk to lots of great people over the next two days. But as you interact with our team, complain. That would be the single most helpful thing you can do. If there's something that in the product you're like, hey, we wish that this worked a different way, tell us. If something is slow, tell us. If something is broken, tell us. The more esoteric, the better. But also. Go into it with the mindset that there are going to be certain things that we are going to take a very specific point of view in the world. And I'll give you a good example. One of the requests that we've gotten from a couple of folks is I want to add my spouse to Sylvia, but I want to do it where my spouse can only see some of my assets. We could do it technically, but we want the company to stand for something. So we're not going to do it. And part of our thought process is there are things that people want to do. And then there are things that we think in the product people should use the product to do, right? And so we don't release features that encourage super hyperactive day trading. There are some products out there that's what they do. That's not the user that we want, that's not the user that we have, that's not my general approach to life. I don't think most of the people in this room, that's how they operate either. And so when you look at the language we use in the product, if you look at the features that we release, if you look at the approach that we take, it is much more focused on how do you focus on household finances? How do you focus on building generational wealth? How do you think about things like estate planning, taxes? How do you think about passing assets down to your children? Those things I think are much more resilient. They embody much more of what I would call like the Bitcoin ethos. In a product, then simply saying, hey, I promise you you're going to get wealthy.
[0:26:56] ?: Let me show you how to make this day trade. I swear it'll work.
[0:27:00] Anthony Pompliano: Right. And so we just think of a product where if you have an opportunity where intelligence is going to be abundant, capital is going to be abundant and people are going to have a need to use this new technology to better manage their finances, then you can take shortcuts and it will probably end in disaster. Or you can take a very, very long view and ultimately the product will be judged on does it solve the user's problem. We've done a ton of technical work. I think a lot of people recognize the product is better in many aspects than the frontier models. But at the end of the day, people judge us on, hey, does it solve my problem? And so the last thing I'll leave you with is when you're talking to our team, not only do I want you to complain, but the other thing that'd be really, really helpful. Is tell them the ways that you use other products because Sylvia doesn't yet do those things. Some of you have private bankers. We're not gonna replace the private banker. We're not gonna go hire a private banker. You're gonna still use the private bank to do certain things. But some of you, I've talked to one guy yesterday, he's been using Claude and he built a whole bunch of stuff for himself. And then he realized that Sylvia did like 90% of it. But I don't care. I don't want him to pat me on the back for the 90%. I wanna know what's the 10% that Sylvia doesn't do yet. And so that's probably the single most valuable thing you guys can do. I would really appreciate if you complain to our team. They're all sitting over here. If you can't tell, this is the best advertisement we got for Sylvia. Our entire engineering team is sitting over here.
[0:28:35] ?: They don't give two shits about what I'm talking about. They just want to write code, right?
[0:28:38] Anthony Pompliano: So now they're laughing at me. But why don't we do some Q&A, and if you ask me anything hard, I'll get one of them to answer the technical stuff. No questions. Go ahead. Does your shirt say Satoshi is black?
[0:28:54] ?: Okay, I like that shirt. Go ahead. Sorry.
[0:28:58] Anthony Pompliano: First question. Got to start off with a bang. Hold on. They want you to say this into the microphone.
[0:29:09] ?: Right on, right on, right on. Thank you so much. This has been awesome. I got invited here by Uphold, so I come from a Bitcoin space and definitely appreciate your approach. I think one of the things we made a mistake in the Bitcoin space is we kept getting upset about... The regulators, but we name things ICOs and IPOs kind of rhyme with that. And then Bitcoin got coined. We could have called it Bitdata. So I think there's sometimes and since Satoshi went away, we don't have anybody advocating. So we've kind of gone through it, you know, just with. More of a libertarian mindset, which has been great. I think with the AI, it seems like, and I talked to your folks earlier, that if we made it more relevant to people, you know, Bitcoin, we said it was going to help the underbank. And then all the conferences I went were like $1,000 a person. So I don't know how underbanked you could be going to there. Or they would talk about Africa like it was on Mars. We're going to help Africa. And everybody on stage was like a necrologist, not one African. They're just talking about how they're going to help them. In your case, it seems like you are hitting it on the head. And one of the problems I have, my dad is a Vietnam War era vet, he has dementia. If AI could actually, especially Sylvie, I'm just talking to you, your product being focused, you know, of course, high net worth people, but the... Problem solved with all the baby boomers that were now taking my parents. I moved them in my house and managing their pensions and making sure they get all the disabilities. I think there's something that we can say as an industry saying, look, this superpower is not just for generational wealth. This is going to help you take care of mama and daddy because they got some things that are tricky. So a couple of things. Bitcoiners are horrible marketers.
[0:30:57] Anthony Pompliano: You know, whoever named Bitcoin the asset and Bitcoin the network the same name, we were already off to a bad start on that part.
[0:31:04] ?: Very confusing, you know. Second thing is.
[0:31:09] Anthony Pompliano: I actually, hopefully I don't get yelled at here for this.
[0:31:12] ?: I'm just going to tell you guys what we're building. And some of it's not released. Some of it is in testing internally.
[0:31:16] Anthony Pompliano: But I feel like the more we tell you guys, the more you can give us feedback. So internally right now, we're testing an entire integration with Gmail, Calendar and Google Drive. Right. So now all of a sudden, what if Sylvia, rather than you have to ask her questions, what if she could just go scan your email and say, hey, I found, you know, 32 different K1s. You want me to put them into Sylvia and then update your holdings based on what the K1s say? Probably save you some time, right? So going back to this idea of what are we trying to do?
[0:31:42] ?: We're trying to save people time. We're trying to make them money, right?
[0:31:45] Anthony Pompliano: But we want to do it in a way that it's got a very specific lens. And so by creating more context into Sylvia, one of the aspects of Sylvia that I think people, they feel, but they've never verbally heard. Is we started off 18 months ago as a ChatGPT wrapper, right? We basically we built all the software and we just plug ChatGPT in. Then we went from ChatGPT, we went to Claude, into Anthropic stuff. But at the end of the day, we were just feeding your query plus your context of your finance into the context window. We were asking the model and it was coming back. But very quickly, you realize there's a lot of limitations in doing that. And so really, over the last six months, we have rapidly shifted and tried to rip out as many of the frontier model stuff as we can. And we have basically rebuilt a whole bunch of the technical stack. And tomorrow, Shane and I are going to do a session in the morning where we're going to kind of go through some of this. But the way to think about the technical stack is there's something called an AI harness. So think of that as almost like your air traffic controller. There are a lot of components to that. But that is where a lot of the value accrues is in that harness. As part of that harness, when you're asking a query to Sylvia, Sylvia is basically rapidly receiving the query and deciding, what do I do with this?
[0:32:58] ?: How do I answer this question? How do I best serve this person?
[0:33:01] Anthony Pompliano: So let's say it's a tax question. If you ask a regular question to Google and you say, you know, how do I get my tax rate down? Google just looks for links, has no context. So zero context, lots of indexed on the Internet, no AI model. Just send you the links back. The innovation that the models came up with was you asked, how do I get my tax rate down? The model says, well, what if I don't have context, but now I have this AI model? What if I can go and use the model training plus web search and I can come back with an answer and I don't provide links, I instead provide a summary? Pretty valuable. I use them every day, right? But actually what you're asking in that question of how do I mitigate my taxes? My. Taxes is the most important part. And so having the context of your personal financial position ends up being the single most important aspect of that. Now, we've come out, we've publicly showed that Sylvia is more accurate in answering questions to tax, mortgage, credit cards, et cetera, than the frontier models. How do we do that? One, we have trained our own model internally to be able to be focused on personal finance.
[0:34:04] ?: We don't want to be good at physics. We don't care about architecture.
[0:34:07] Anthony Pompliano: I'm sure you guys want a date night with your wife. We don't care. We just want to be good at personal finance. So we have focused all of our energy on be the best in personal finance. The second thing that we have done for a tax question is we have actually gone and we have indexed the federal tax code and most of the state's tax code or the major state tax codes. We then put that into our own database called data pipeline. And then we basically have the agents run through it. And so think of it as like you ever seen the game where somebody like. Looks at a piece of paper and they put it down and people start asking them questions. They got to recall, you know, A1 was this, Z2 was this, whatever. Basically, Sylvia is really good at understanding where is all the tax code in that database. So when you ask a query to a model, let's say Anthropic, usually what they're doing is they're looking at how the model was trained, what's the information, plus web search. What we're doing is we're taking that query and we're saying, OK, we understand it's a tax question. Go get the tax code. Go to the source material. Don't go look for a blog post talking about a tax code from two years ago. Just go get the tax code. We then pull that in and then we use the power of AI. We use web search. We use all these other things. But the accuracy of getting the source information is really important. So the reason I say that is if that's the level of granularity you can get to, you mentioned pension. It is literally just a prioritization for us. We could have the best answers to pension questions. We just got to go build the data pipeline. We got to use Sylvia's AI model and we've got to build it into the harness to be able to answer the question. And so that's part of what I say is like the complaining. You know how many people have told us about pensions? Three or four. Put it on the list. Okay, now we've got another person, right? And as we build up and we realize, hey, this is something that people really want, then we just say, okay, our resource allocation goes to let's go be best in the world of pension. So we did tax, then we did mortgage, then we did credit cards.
[0:35:53] ?: Those are big addressable markets. We can move through the other ones.
[0:35:56] Anthony Pompliano: The other aspect is I talked about the Gmail, Google, Google Drive, etc. How many of you in here use Grokbot, Instinct, Muse, like any of these personal assistants? Okay. If you think about what these things are doing, one of the values is they are moving from being a static, receptive agent, meaning that you have to prompt them to being a dynamic, proactive assistant. So, for example, if you have your email and calendar attached to one of them, they may realize, hey, you have a conflict tomorrow on your calendar and they read the emails and they realize, well, this person actually gave you three times. You have a conflict. Should I move it to this other time that the person had offered? That is just context being fed into the model. And so imagine what could happen if you do that for finance. Now, all of a sudden, what if it knew that you were meeting with your accountant next week and it was able to say, hey, by the way, since last time you met with the accountant, here's all the things that have changed in your portfolio. Here's five questions I would ask the accountant. And by the way, afterwards, I'll ask you some questions. Give me the notes so I can update my memory. And be better prepared for the next accountant meeting. More proactive. How do you move more to this like agentic world? The second aspect of your question is what I call household finances. When most people think of household finances, they think of one of two things. Young people think about how do I make sure I got enough money, right? So they really go to budgeting and saving and all that kind of stuff. We can be helpful there, but that's not our focus. When I think of household finances, think of myself, probably most of you in this room. I got both directions to worry about. I got my parents. Right. I got other people who are older than me and I got kids. And so my household finances are not just me and my wife. My household finances go both directions. And some of that is cash flow related.
[0:37:52] ?: Some of that is investment related. Some of that is tax related. Some of that is estate planning related.
[0:37:57] Anthony Pompliano: That's a much more complex problem than like I want to save five dollars on the Starbucks. Right. And so when you start to think about. That there's certain aspects that we think are really interesting. For example, how do you and let's say use me and my wife. She's going to be embarrassed, but I'm going to use my wife as an example. My wife and I think have a pretty good understanding of a lot of our goals.
[0:38:18] ?: We're pretty intentional about a lot of that stuff. I love my wife to death.
[0:38:21] Anthony Pompliano: If I got hit by a bus tomorrow, I pray for her because what would she do?
[0:38:28] ?: She would do the best job she can. She knows where the assets are.
[0:38:30] Anthony Pompliano: She knows a lot of stuff. But does she know instruction-wise what to do? So I got this idea because a friend of mine, he built a custom home during the pandemic in Tennessee. And when he built the home, he was able to get a 2% mortgage on it. Genius. He told me, he said, if I get hit by a bus, my wife's going to pay off the mortgage because she hates debt. And the last thing she should do is pay off the mortgage. So he put a Google Doc together and he wrote three times at the top in all bold letters, do not pay off mortgage, right? So you think about... Some couples or households, they're like, oh, one person knows where all the assets are. The other person doesn't know because they don't care. OK, you get over that. But actually, if you really try to get to the root solution, it's not just where are the assets? What do I do? Who do I call? What can they help me with? What are the instructions of what to do? It's almost like you can have a will of where the assets go, but then you almost want that instruction manual of like, what do you actually do in reality? How do we build that into the product? How do we take people who are coming to Sylvia and say, oh, this AI thing is cool. But help them actually get themselves in a better position so that you can think generationally, household finances, all that. So when you start to look at a product like that, now you start to say, OK, wait a second. Should Sylvia be in a group chat with me and my wife so we both can talk to her about money and we both get the information? Yes. Right. Should Sylvia be able to ask us questions to constantly be updating the memory so that we understand, God forbid, anything happens to us? Should we be able to add Not only our children, any portfolios they have, but then also our accountant, a lawyer, a tax expert, etc.
[0:40:07] ?: How do you think about permissions? Should the accountant see everything?
[0:40:11] Anthony Pompliano: Should the accountant just get a report? Should the accountant be able to ask Sylvia questions about my finances? These are all things that we're actively building right now so that if you kind of zoom out and say over the next couple of months, what I hope is the product experience is it's a very proactive agentic experience where you upload. Your context. And then you begin talking to Sylvia the exact same way you talk to a human CFO. This has got better memory. It's got more intelligence and it is able to work with all of the people in your life. So I got really young kids. They got no clue about any of this stuff. But if I had an 18 year old kid, wouldn't it be cool if Sylvia could actually answer questions about their allowance or the money in their account or any of this stuff? As a parent, maybe I want to see what my kid's talking about. Maybe I don't want to. Maybe my kid, right? So like you go into these different things. And so these are the things that we're thinking through, but that experience and kind of how rapidly we're changing the product is we just think that a full on agentic experience of Sylvia is the future of where all software goes. And most of you today log into Sylvia and go to the dashboard. Why am I being so arrogant to ask you to do that? Shouldn't Sylvia just know what information to send you when it is contextually relevant? So that you don't even have to go into the dashboard. Can we get there? Let's see. It's hard, right? So that gets into like the more of the engineering stuff, like that's where the world is going. And so I think that what becomes really fascinating, again, if you go back to the investing side, is if that's where the world is going, there's a lot of companies that are on shaky ground because they basically benefit from the fact that users don't log in or users don't use the product. What I want is I want every single one of you to use Sylvia every single day. But to be something where you literally write back to Sylvia, thank you. And actually, where's Shane? Shane, that's a great metric that we should measure. How many people say thank you to Sylvia on a day-to-day basis? Right? So you think about it from that perspective, like if that's the direction of the product, now all of a sudden your father, who's got the pension stuff, right, he should just be able to say, I don't know anything about my pension, what should I know? And Sylvia should basically be like, I got you. But you need the superhuman intelligence and the context to be able to do that. So that's why we push so hard in that direction.
[0:42:28] ?: Chris? Oh, he's like a shark. He's waiting.
[0:42:34] Anthony Pompliano: See, Chris, that's what they think. They give you a microphone that don't work.
[0:42:39] ?: Yeah, they know me. First of all, thank you. Came last night. Thank you. Have you guys identified a path to a cell phone app? And what is that path? Michael, where are you?
[0:42:53] Anthony Pompliano: Stand up. See that guy right there? All your hopes and dreams are pinned on that guy right there, Michael. You better protect him like he's gold. Michael is building the mobile app right now. And yes, we are going to launch a mobile app. What is very interesting in building this, right? One of the things I think is probably fascinating as I'm going to talk a lot about Sylvia, but I think it's indicative of where investing in public markets in general is. I just described a very agentic experience where you can talk to Sylvia, right? We announced last week over iMessage and Android. Well, now all of a sudden you have three interfaces that you could talk to Sylvia. You've got desktop where you go into the dashboard. You have a mobile app or you could text her. Our general approach right now is that you should have choice in however you want to interact. But I think that over time what we're going to find is user preferences and kind of consumer activity is going to really consolidate into texting where people are just going to talk in like a familiar interface. They're not going to want to go to an app. Right. But he is building it. Michael, have I told you they want the app? Let's go. You know, why is your computer not open? Why are you not kidding? Yeah, we are building one. And I don't know if we're yet ready to commit to a timeline, but soon is a thing.
[0:44:09] ?: Go ahead. I'll repeat you. Oh, hold on. They're running.
[0:44:12] Anthony Pompliano: Let them give us. OK. I'm a Sylvia user and Sylvia investor. And like most people in this room, we all think it's pretty great.
[0:44:20] ?: And I'm sure it's going to become more amazing. What are you doing to get it out to the masses?
[0:44:25] Anthony Pompliano: Because like I tell people about it, nobody's ever heard of it. Like surely there's a plan. You can't just. Are you expecting to just grow organically or do you have a plan to get this?
[0:44:33] ?: Because it's so incredible. Do you see me tweeting out here? I do see you tweeting, but I still don't think that's a big enough audience. What are you doing to get it out there?
[0:44:41] Anthony Pompliano: I'm pretty sure my followers know about Sylvia at this point. Every meeting I go to, people are like, yeah, I've seen you tweeting about this AI thing. I'm like, damn, you still don't know the name of it? Shane is laughing because if there's one thing that we debate most kind of vigorously, it's this, which is. Look, I ran a number of growth teams at Facebook, and the thing I worry about the most is what is the conversion to people being an active user and what is the retention of those users? Once you have that down, here we go. Then you can go pour people on the top of the funnel. So for the last, call it, I don't know, four to six months, that is all I have been worried about. Now, the numbers that we have shared publicly are 25% of all monthly actives now are on a pro or max plan. I've been doing this a long time. That is pretty... That's pretty good. That's like world class level conversion from a monthly active to a paid user. Now, the paid user, one of the benefits is 20 bucks a month. Financial advisor could be 1%, 1.5%, right? So like you're kind of almost like cheating a little bit and by it just being super low cost. But again, it goes back to this idea of, hey, we're going to build a long relationship between Sylvia and the user. Like the user should have some skin in the game too, right?
[0:45:52] ?: And so doing that, it's been really good. So the conversion's there.
[0:45:56] Anthony Pompliano: On the retention data. Kyle's in here somewhere. I'm probably going to get yelled at. I believe, hold me honest on this. Every single paid cohort we have, the retention is over, I think every single one is over 90% except for one, and one is at like 87%. So the way to think about that is over multiple months, the retention is, call it around 90% of those paid users. Again, world class in terms of not only the conversion rate, but the retention. Why do I share that? Because... Think about if you had two buckets. One bucket has no holes. One bucket has holes. You could fill up the bucket with holes with a fire hose. It'll be full real quick, but then the water will eventually leak out through all the holes. But the hole with no buckets, you could literally take an eyedropper and eventually it'll fill up because the water doesn't go anywhere. Now that we know we've got a bucket with no holes, now we can go pour stuff in the top of the... How do we just get as many users as possible? And so we have started doing some paid advertising. One of the first ads just started running on a podcast called Money Wise. They interview rich people about their money, right? You'll start to see us roll out a lot of paid advertising, do a lot more on the marketing front. But a huge part of this was... We had to make sure that we had a product that people one wanted, but two also they were willing to pay for. And then three is that they would retain. And so if you just use kind of simple numbers, if somebody is paying 20 bucks a month on it, you know, as a monthly subscription number and four or five months since we turn on monetization, retention is, you know, 90 ish percent. Now you can start to say, OK, well, I know that I can pay X dollars to acquire a paid user and I'm going to get paid back that money. Now, what I can't tell you is what's the retention at 12 months because we haven't had monetization on for 12 months. But as we get more and more data, what's likely to happen is that LTV number will continue stretching out. And so if today, again, just use a simple number, let's say it was five months, we thought, hey, we've got $95 to $100 of revenue for every paid user because of the conversion and the retention. Well, I now know that if I want to be breakeven on the user, then I can go spend. 90 to 100 bucks to get a paid user. If I want to be profitable, maybe I spend 50 bucks to acquire a paid user, right? So you're going through these calculations, but as we get more data, that LTV number keeps growing. So let's just say it's 100 today for easy numbers. Well, next month, if we still have 90% retention on that cohort, now it's 120, then it's 140, then it's 160. If we get all the way to 12 months after we turn on monetization and retention is still, you know, 80 plus percent. Well, we spent a lot of money to go acquire paid users. And so what you're really trying to do is you're trying to make sure that those unit economics work. And then once you know that the math works, then literally it's just I can put a dollar in, I can get two dollars back. How much money can I put in the system? Right. And so we've kind of walked ourself into this, but we feel really good now about the retention. And I think you'll see us start running ads. You'll start seeing do kind of a lot of marketing. And internally, we have a campaign going on. Bilal is here somewhere. Bilal, where are you? Maybe the ball's in the back there somewhere. We have a guy who's running growth who the first meeting I ever had with him, I said, we're going to start a campaign. The campaign's called Kill Pomp. Because I'm 99.9% of all user acquisition right now. I'm not going to stop doing what I'm doing. I'm going to keep tweeting, I'm going to keep going on TV, all this stuff. But I can't be 99.9%. So what we need is we need to build up organic. Channels, we need to build up paid channels, we need to build up influencer channels, all these other channels that are available for acquisition. I need to be less than 50% of all user acquisition as soon as we can. So kill pomp is basically dilute me down in terms of the impact of acquiring users. Is Phil Rosen in here? Phil? All these guys are bullshitting out there. Phil Rosen is chief market strategist. Phil went on television yesterday. He mentioned Sylvia and we started getting people signed up. I almost do a party in the office. I say, thank God I ain't got to go on TV and do it myself. We got another person who can go do it now, right? How do we get five more people, 10 more people we can get onto all these different television stations, siphon those users? How do we get paid up? How do we get organic up? All of those things are part of the marketing, but you can't really start to do that stuff until you understand if I spend a dollar, am I going to get it back? Now that we've got confidence, you know, on some of the early math, now we can start going and really scaling marketing.
[0:50:19] ?: Right behind you. So smart,
[0:50:40] Anthony Pompliano: dumb approach, right? That's my worldview. If your stock's trading at a 40% discount, To your liquid assets, buy back the stock. Pretty simple. Okay, great. So it went up. There's some people in this room that were, somebody was telling me yesterday they were buying the stock at like $1.60, hoping I wasn't just stupid and wasn't going to buy back the stock, right? So if you look at it from that perspective, okay, we've arbed back the discount to NAV. We trade on a day-to-day basis somewhere around NAV, right? Maybe 2%, 3% below, 2%, 3% above, et cetera. One of the open questions, and again, I think part of this is I tweeted somebody today. I said, I personally, through my investment firm, et cetera, I'm one of the largest, if not the largest shareholder of the company. Okay, great. I think of every single person in here who owns stock as a co-owner. So I'm going to tell you good stuff. I'm also going to tell you the problems we got, right? One of the open questions that we have, which I don't have a good answer to, the board and I talk about it all the time. Most times people talk about discount to NAV as discount to the liquid assets. If you were to also include the value of Sylvia, even if we trade at the NAV of Bitcoin, we're probably still trading at a discount to Bitcoin
[0:51:56] ?: plus Sylvia. What is Sylvia worth? I have no clue, right? The market is saying zero.
[0:52:02] Anthony Pompliano: So it's kind of hard to figure that out. So I don't think that we want to get in the game of trying to predict what we think the equity value is. And so I think that we will continue to use NAV as kind of a reference point. Doesn't mean that we will only buy at certain points or below, like we're not going to give guidance on that type of stuff. But I think that's like an open question. It's like, what is Silvia worth and how should we think about that as to where the stock trades for stock buybacks versus that?
[0:52:25] ?: So put that aside for a second. The second aspect is.
[0:52:29] Anthony Pompliano: As we think about growing the business, one of the things that I'm very proud of is today we announced that we've bought back about 20% of all outstanding stock. And so as we have done that, obviously, the stock price has reflected it. But more importantly, is every single person in this room owns more Bitcoin today than they did when we first started because we've been able to use that arbitrage. I do not believe that the value of the future of Sylvia as a company is solely dependent on does Bitcoin's price go up or not. Right. That's not really a company. That would just be like, hey, we were able to, you know, financially engineer some ownership of Bitcoin and let's just wait. What I believe is that Bitcoin could be a great asset and Bitcoin can continue to appreciate over time. But the value creation that is really going to deliver the performance that we all in this room want is going to come from can we build software that people want? Can we get users? Can we drive revenue and profit? And so it's very weird because we hold the Bitcoin. We get lots of questions about the Bitcoin, but. Bitcoin may be one of the things we spend the least amount of time on internally. We're spending all of our time on the software and going and doing that. Said another way, we want to focus on things we can control, not on the things we can't control. So going forward, now that we have this retention story that I think we feel really strongly about, now it becomes a game of there's a race going on. How much do we want to invest? And what is the ROI we need to see on that investment to feel that we're getting the return that we want, right? And so if you go and you look out in the market, if you said to me, hey, go and get 10 million users, we have the money on the balance sheet. We definitely could go do that. But it's not just about what is the headline number. We also need a very specific type of user. So one of the aspects that I think we've learned over time is that AI is very good at dealing with complexity. If you have $10 on Robinhood, Sylvia may not be that helpful to you. Because you have no financial complexity and therefore all you're thinking about is where do I put my $10? It's a pretty simple question. If you have many more assets and more complexity, now silver becomes really valuable. And so one of the great things is in America, depending on what stat you look at, there's something like 25 million multimillionaires in America. Okay, that's a lot of people. And so if you just use that as a barometer in terms of, let's say that that's who we're targeting, we know we're gonna get tons of people who have... We are not optimizing for just... Total number of users and kind of rapidly doing that. What we're looking for is we're looking to penetrate a very specific market and we should still be judged based on growth, but it should be growth of the right users. And so that's why a lot of times you'll hear me talk about the number of users with a connected account. So we don't report, this is one of the other things. I feel very strongly coming from the background I do of running these growth teams. I could manipulate data all day long and tell you that we're fucking greatest things in sliced bread. Look at how many people gave us their email or look at this crazy number, whatever. We are very intentional. We only talk about users with a connected account. That's our equivalent of like a funded account on another exchange, right? So if you sign up and you don't connect an account, you didn't sign up. Right. The whole purpose is that you connect an account. So that's the number that we always report is that number. The second thing is we talk about the total assets on the platform. And the reason we talk about that number is because there's a lot of opportunity that in the future we could chase. We're not doing it yet, but you could imagine a world where when you're putting residential real estate on the platform, well, there's a lot of companies that would love to be able to target, work with, partner with us. Folks who have certain amount of residential real estate or need a mortgage or any of these types of like financial services. And so I think about users with a connected account and I think of assets as kind of the two key things. And then the other data point that I spent a lot of time thinking about is our paying users, how many questions are they asking of Sylvia per week? Right now the number is 20 plus questions per week. So you think about that, you're like, okay, you got wealthy users who are highly engaged and are willing to pay. Those are the ingredients to build a really successful product. Now you got to go scale, right? And so as we go and we start to scale, what I think people ultimately, if I was an investor who wasn't involved in the business, what I would want to see, quarter over quarter growth of users, I'd want to see growth in assets, and then I would want to see continued growth in number of paying subscribers, right? And if those three things keep growing and they're growing at a healthy rate, future is bright. If all of a sudden there was headwind or some sort of diminishing growth, Then all of a sudden I think people would start saying, hey, what's going on here? And so you should use that as a guide as to, okay, now's the time to go and acquire users and kind of make sure they're the right cohort and they're willing to continue paying
[0:57:18] ?: us. I signed up for Selvia because my friend told me to.
[0:57:31] Anthony Pompliano: I know your friend, so she did a good job.
[0:57:34] ?: And apparently I have more connected accounts than the majority of users. However, I just learned from you that there's... A pro and a max account, which I did not know. And so my question is, I understand everything you're saying about assessing LTV to go after a cold audience to see what your possible spend is. But what are you doing, if anything, perhaps I'm missing it to convert your already engaged free customers.
[0:58:04] Anthony Pompliano: So the way that the paywall works today is you basically start using the product. And then once you have used up a certain number of tokens, then you're shown that paywall. And that paywall basically allows you to get access to the best models and more usage. There's a couple of other bells and whistles you get behind the paywall, but those are the two main selling points. And you can think of this as token usage is a very weird thing, right? For example, let's say Justin here, he could do 20 chats and not hit the paywall. I could do one research report and hit the paywall in the first day. It's all about token consumption. It's a very weird dynamic to describe to people, right? So it's not exactly like it's credits. It's not like, hey, one credit equals one query, et cetera. So that's something that we've been working through is how do you communicate to somebody other than just like a progress bar? Because one of the complaints for the professional complainers out there, we appreciate you, is, well, I was using this.
[0:58:59] ?: I was at 20%. I asked two questions. Now I'm at 90%. What happened?
[0:59:03] Anthony Pompliano: Well, they probably asked questions that were super consumptive of the tokens. Right. So they asked something that was very complex. And so how do we communicate that to people over time? I think it's an important thing. Now, I actually like the fact that you don't know that there's a max or a pro plan because the odds that you are ready to subscribe is probably pretty low compared to where you are going to be when you hit the paywall. One aspect I'll explain is that conversion of 25 percent of monthly actives to paid is because the people, by the time they hit the paywall, they're ready. They've got a lot of value out of this thing. And so you can imagine if on day one, when you connected an account, we said, oh, to connect a second account, you got to pay. We could do that, but you'd be like, why am I going to pay for this thing? I don't even know what it does. Versus if we give you free usage for a while, and then we show you the paywall. At that point, you're like, this thing is really valuable. I want to keep using it. It's so valuable. I'm willing to pay for it. The way we rolled it out, and I don't think I've ever publicly said this. We started the paywall. I think we showed it. We're showing the top 3%. I think top 3% of users. 2%. So if you got hit in the first day with the paywall, you were on top 2% of users. That's the only people we showed it to. 98% of people weren't even eligible on the platform. And then we went from two to like five, then five to 10%. We just slowly started to crank down the threshold on that allowable free token usage until we got to a point where we felt like, hey, this is pretty strong, right? And so that's where we are today. Now. One aspect, again, you know, if we're all co-owners of the business, one of the things I'm really excited about that we have not yet built, but I think is going to be a huge opportunity. How many people in here go to a website or a product and they get told something to the case of, OK, you're going to get X number of credits and then you hit the paywall or you get X number of usage and then you hit the paywall. And it's like one size fits all. That's pretty much how every product in the world works. I got superhuman intelligence now. Thank you for watching. Kind of rather than dynamic pricing, the dynamic showing of the paywall and being able to be predictive in when is someone ready to actually convert, I think is going to be a like huge trend across all these consumer applications. But we now think that it's possible. We got to do the work to build it. We got to make sure, you know, test it, make sure it works. But when you get into that world now, what ends up happening is you're basically looking at this much more almost like a gaming perspective. Where gaming companies are constantly saying, when is this person ready to buy more credits or keep them engaged or, you know, do these different things? This is part of, I think, what's so exciting about this work is there's now new technology that makes something like that possible where previously it wasn't possible. And so that's the type of stuff that, you know, we want to keep the percentage of people converted high, but also as we pour more people into the top of the funnel, we got to make sure that that percentage doesn't all of a sudden degrade. Right? And I think that's a huge piece of it is you don't know about Max and Pro yet because in the product we haven't shown you because we don't think you're ready. But by the time you hit it, you're going to be 100% likely, it sounds like, right, to convert. And so that's really what we're optimizing for. Jordi, get ready. Three minutes, big dog.
[1:02:38] ?: Hi, Anthony. More of a personal question, but as you continue to scale the product, Is there a personal finance related question that you still don't trust a general purpose AI answer and why? I'm all in, man.
[1:02:53] Anthony Pompliano: If this thing told me to drive off a cliff, I might do it. Not just Sylvia, but just in general, right? Like, I think that there are, I'm pretty like privacy centric when it comes to most things. But for some reason, I think I've gotten to the point where I just know the more context you give it, the better. Right? And so I actually, I don't know if he's here, but I met a friend of mine the other day on Friday. And what he told me he does, which I thought was fascinating, every single text message, every single email, everything, he feeds into an AI system that he has. And then he has on his phone an app. The app is QUO.
[1:03:31] ?: I've never used it. I have no affiliation. But QUO is the one he uses.
[1:03:34] Anthony Pompliano: And he has his personal cell phone number forwarded to this app.
[1:03:38] ?: And it's like a business phone line. But the reason why he...
[1:03:41] Anthony Pompliano: The reason he forwards it to this app is because when he gets off the call, it has a transcript of the call, it has an audio recording of the call, and it has notes from the call. And then he feeds all that into Google Drive, and so every single phone call he has is just going in there feeding context to the AI, and then he just told it, the only one I don't want you to do that for is conversations with my wife. And so then it doesn't do that, right? Okay, if it has all that context, like I am surprised on databases, dude, these things are smarter than us. I literally sometimes will just ask it, what should I do? And it'll just tell me like, well, you have three options, but what would you do? And it'll tell me, and I'm like, what are the downsides to doing that? And it'll tell me, and I'll be like, okay, let's do that. And so it like, it almost is, I think people are still afraid of saying that's how they use these tools. It's smarter than me. Like, I don't know what my IQ is, but it sure ain't 150, right? So like, why would I not trust it? To constantly be able to churn through with better memory, better context, and better intelligence answers to these questions. And so the one thing that you have to worry about is the accuracy. And to me, that accuracy is still the last frontier of trust, is people, if they're using the product, they're like, hey, I know it's smarter than me, I know it can answer better questions than me, but they're still like, did the calculation, is it correct? The hack that I see most people do is they just use another LLM to check the first one. So, you know. Grok gave me an answer on something. I just put it into ChatGPT, say, is this right? I'm not saying that's the best way to do it, but I think that's like kind of the street version of, you know, fact checking, right? There are probably a couple of things where I still want a second opinion, but I pretty much use the AI for almost every single thing in my life at this point. If I got a question, I'm going there and I want it to give me the answer. All right, we got one more question and then we got to go.
[1:05:40] ?: Whoever's got their hand up right there. On that note. Is there... He stood up. I did stand up, sorry. I was hidden behind people. No, you're good. Do you foresee a future where Sylvia no longer says this is not financial advice or refer to a tax attorney or estate planner, whatever, those types of folks? Because I think that's one hang up that you're talking about where it's either the trust layer or, hey, how much do I have to pay for you to actually not... Tell me that each time.
[1:06:16] Anthony Pompliano: Yeah. Well, the first thing is anyone who's ever worked with a financial advisor, sometimes they send you documents and they're like, this is not financial advice.
[1:06:22] ?: You're like, well, what the fuck am I paying you for? You're supposed to give me the financial advice.
[1:06:25] Anthony Pompliano: But like there's rules, regulations, laws, like obviously even non-AI companies, they're constantly thinking about risk mitigation. So put all like that stuff aside for a second. I think the essence of the question is basically like what point is the AI so good and so reliable that people are like, I just go to this and I get the answer. I personally think that a lot of consumers are pretty close to being ready for that. But when it comes to big financial decisions, they still want some sort of human overlay. We don't yet have an answer for how to do that. We're aware of it. I think that we think it's interesting. If you think of like zero to 100% of questions. We're probably pretty good at like 80, 85% of them. But you can imagine, I'll get on a call and like I got on a call with a guy one time, he's got like $500 million. So already I'm like, what are you doing on the call, right? Second, he's like, starts asking me these super esoteric things. And he's like, so I have a sibling and we both co-own this LLC, but he's like going through this whole thing. I'm like, dude. You got to get a financial advisor, man, right? Like AI is not going to be the solution for this super important decision that deals with like multiple family members and taxes and like all this kind of stuff. So there's some portion of questions that just like cut those off and like we shouldn't try to optimize for those. Then you get into this world of, well, should we be a referrer to a financial advisor, an accountant, all that kind of stuff? We have a lot of interest from those types of firms. We have a lot of interest from people who have financial services.