Good morning, StockTwits Nation. I hope everyone likes the new music. It's very Studio 54. Our friend Michael Park is back, and of course Shai is here. We have a packed show. Guys, how's it going? First of all, Michael, I noticed you were bopping around. Is that your style of music?
That music? I mean, you said Studio 54, so that starts to date me.
Exactly. I like 54. No, exactly. And then I want that for my mixtape.
That's good workout music, too.
Oh, but yeah, the gym. Just like, you know, cranking some weights, listening to... Shai, how do you like the new song?
No, I love it. I'm craving a scorpion on a rooftop somewhere right now.
Yeah, it's kind of like Mika's vibe. It's good. I love it.
Yeah, 100% scorpion. It's good.
Well, welcome back to the show, everybody. Michael's here. We got a lot to cover. If you guys watched a couple weeks ago, maybe was it last week? I had a fascinating conversation with him. The CEO of IonQ is stopping by to share some news. But first, Shai, you made an announcement yesterday. Tell us about what's going on in your world.
Yeah, I mean, honestly, before I even talk about it, how about we just roll the tape, Keith. Today's a big one. I'm officially joining the Futurum Group to launch Futurum Equities alongside the man, the myth, the MBA, Daniel Newman. Look, retail investors are sharper than ever, but the research world is still built like it's 2005, and we're here to change that.
Yeah, that's exactly what we've been doing for the past seven years since we started Futurum. We've been advising the world's largest tech companies and their CEOs. We've been helping them navigate disruption, innovation, digital transformation, and now AI. What we're going to do here is we are going to extend that same level of analysis to the retail investing world, giving them clarity, discipline, and conviction, the insights needed to build a world-class portfolio. That's exactly why we're launching a full lineup of research products and subscriptions. But this isn't just content. This is capital intelligence. And in a market moving faster than models can keep up, AIR is how you breathe. Access, insights, research. That's the future. That's Futurum Equities. The days of retail just getting the leftovers are done. We're setting the table now.
That's exactly right. Welcome to Futurum Equities. Welcome to you. Welcome to everybody. Time to get to work.
All right, Shai. Back in the lab. Congratulations.
Back in the lab. I mean, yeah, this is exciting. This has been in the works for a couple weeks now. A lot of backend work and dialogue between Dano and myself on what our vision was going to be for Futurum Equities. And I think the biggest reason why this is getting formed right now, in my opinion, is everyone treats institutions as like the geniuses of the world and retail is just dumb money. That's why you always hear smart money, dumb money. No, they're not smarter. First of all, just look at Robinhood, Palantir, they've missed the boat on that. That's a perfect example of why institutions are catching up. But either way, a lot of people think these institutional investors win because they're smarter. No, I really do believe it's because they're closer to the actual data. They have access that retail doesn't have, where they're in the rooms with the decision makers like the CEOs, CTOs, CIOs, the people who are actually making the moves on where they're going to have these investments in AI, what the infrastructure, where the next bottleneck is going to become, and then they front-run it. And then retail just sees it in the earnings print, but by then it's usually too late. So I think there's been this kind of massive institutional barrier for retail that I'm just like, why is that still the case? That's why I mentioned it's like everyone's treating investing like it was 15 years ago when Robinhood broke those walls down through disrupting the brokerages.
What's the other component of that calculus for retail investing?
It's access. And I think that Daniel, through his connections in the tech world, and he has an institutional research firm already, Futurum Group, he's willing to take that wall down to give retail that access that they've been craving. That institutions are like, I don't care about the little guy, I want to make my buck. This is what's changing that now. They're getting access to actual capital intelligence that institutions have been having for decades, as Michael can allude to, he's been around the block for a while, where they have the information and retail's just guessing. Those days are no longer. So like when we mentioned AIR, I love the rhetoric, AIR, like we're providing AIR for retail. It's not just a word, it's access, insights, research. That's what we're providing, and I'm super excited to start this chapter of Futurum.
I mean, this is awesome because this just means more content from you, which we all know and love. And Michael, I'm curious what you think about all this because you came up in the institutional side of things and things have changed very rapidly. And Shai and Daniel, they're getting the meetings that people who are catering to retail weren't getting in the past. What do you make of this trend?
Look, one of the pleasures of watching technology over the last 30 years is how it democratizes every market. And you're right, my world, you know, I was at Goldman Sachs for over 20 years, the ultimate institutional firm on the planet. And yeah, my whole life's work was around institutions, but that was the mainframe era of investing, and today is the PC and smartphone era of democratization, and that's coming to financial services, and that's what Shai is talking about. You know, Robinhood absolutely broke down the barrier in the context of the trading part, because I remember, you know, I joined Goldman in '82 till 2003, '04, and our whole world revolved around 5, 10, 15 cents a share. And now all of a sudden it's all, you know, it's extraordinary, you know, for today's generation to even think of that. That's how, you know, I bought my first Porsche because, you know, you do $100,000, you know, it was... no, but I'm just saying, you know, that's that. And that was, by the way, three days before the crash of '87. I, you know, I told my Porsche salesman, it was on a Friday night and the crash had happened on Monday. I took my delivery of my car and I said, you know, this is probably going to be the last car you're going to sell in Manhattan for another year. And savage. And six months later, six months later, the Porsche dealership was gone. There were no Porsche dealerships in Manhattan for the next two or three years. It was that bad, that crash. Anyway, sorry. Sorry for that.
I always say that to my Porsche salesman, too. Always. It'll be the last one you ever sell, buddy.
No, it's look, it's markets sell markets and reality. And what Shai is talking about is exactly the democratization of investing, and research is one of the big next access points. We've already seen this in the meme stock, what's been disparaged as the meme stock. It's just that individuals are now very well-informed investors. And whether they're trading short-term, mid-term, long-term, they have the same tools, access, everything that institutions did. And now in a world where IPOs are scarce, the liquidity in the markets is very low, they have as much clout to move things as anybody. So anyway, so I'm very excited. Shai, congratulations. You're on to an exciting new journey.
Yeah. And Shai, I have a question for you. I want to get in the headlines, but I'm curious. You know, you have your pillars of themes you're into. Is this endeavor sticking with those pillars and kind of going deeper? Are there new pockets of the economy that you and Daniel want to identify and explore?
I mean, it's all fourth industrial revolution. I think everything I've been doing the past couple years is just me sticking my finger out in the air and seeing where the wind's blowing a bit. Now I'm essentially going to be talking to God, and God's going to tell me the wind's going to go southeast, essentially. Like I'm going to be in the room now hearing it from the Lisas, the Jensens, etc., or what they're actually saying, and then I can make my educated assumption on what they're telling me rather than just reading channel checks, piecing all these pieces together to come up with the final puzzle. I don't really need to do that as much. Now I'm going to actually hear from them, and it's going to be less assumptions and more intelligence, and that's what's going to be exciting, where maybe I'm going to be bearish on software now. Who knows? Like I think I'm bullish on software, but maybe I'm going to get some information of like, whoa, the AI tsunami is hitting the software even harder than I expected. Like it's giving me more layers of information to make these assumptions I've been making. And that's going to make me look smarter, which is great, but that's why institutions just look smart. Like these analyst notes, it's because it's verbatim what they're hearing from someone else. They're not making really that great of assumptions. It's just an added layer to what someone else is saying. That's what I'm excited about, where I'm not going to stick my neck out as much because I'm going to have the facts now in front of me.
Very cool. I mean, exciting. There's been hints of your bearishness towards software lately. Shai, we talked about the move to consumption model. Let's start with MongoDB because this is a spicy headline with them because they're in your doghouse and they just put up some numbers. I think it was the best they've reported in like six years or something. So, can you quickly remind us what MongoDB does and give us a quick summary of what happened on their earnings call?
Yeah, so if you guys follow MongoDB, you guys know that they are, or hear me talking about it, which is constantly, they're positioned perfectly into the agentic AI world we're about to enter because they capitalize on the unstructured database essentially, where it's images, content, stuff like that is going to be rapidly changing. And Dev, their CEO, he's great, but he's so technical. He's a very technical CEO. He's not really a showman. And he's too technical to a fault where it's hard for the analysts on the calls, for example, to get a straight answer from him. And they've traditionally been the most annoying company that consistently beats on the top, beats the earnings, but guides down significantly and sends in bad guidance to make their earnings beats look drastic. So I was like, I'm sick of that. I think this environment, you can't do that no longer. You'll be stuck in this long range. But for three years, Atlas is their AI consumption metric to track going forward. Like that's what they do. That's what the whole, if they're capitalizing on AI agents or agentic AI, that's where the eyeball should go straight towards. Three years, their revenue has slowed consistently. I was like, okay, I know that Amazon has an incredible competing product with MongoDB's offering. Every engineer I talked to that uses it doesn't like it. So I was like, maybe I'm just wrong on this. And I'm seeing the numbers decelerate. Oh God, this was the first, yesterday was the first earnings print in like three to four years. It reaccelerated. That was huge for me because AI agents is just getting started right now. I'm underwriting that it's going to be one of the immediate beneficiaries of that thematic. If they're not showing any reacceleration, it's not a benefit for them. We got that yesterday. And if you look at their stock price, it's gone destroyed. And when a stock finally shows reacceleration on an early sector growth theme with a ton of runway, a rerating occurs. I do believe that is going to be going through that rerating because Atlas grew 26%. I think that's really a lead in this environment where tariff wars caused paralysis on enterprises, and that paralysis caused a slowdown in business spending. In that environment, MongoDB just posted 26% year-over-year growth, and customer additions were the highest level in six years. Free cash flow doubled, operating margins hit 16%. And even the CEO on the conference call, Dev, he was the most bullish I've heard in quite some time. And he didn't say this verbatim, but fill in the blanks: slowdown is over. He believes that the Atlas growth has bottomed at 26%, and that's not cautious optimism in my head space. It's a line in the sand. I do think that's the payoff of that multi-year strategy of doubling down on AI architecture, embedding the vector search directly into their core platform, acquiring Voyage, which was, I think, a great acquisition to own that retrieval layer, and going upstream rather than just focusing developers towards that mid-level enterprises who are constantly trying to modernize, like their own self-serving flywheel. I do think this is a full-stack play on AI, and if real-time data, smart retrieval, it's all in the one system. And I do think this is going to go through a rerating. And I added yesterday at 2:15. It's on my DCA plan list now. I want a 4% position, but it's a great, great earnings to buck the narrative.
You adding is big. You were about to block them. You were like, you were done. You were like, I don't know, I might cut. Michael, well, I think Shai calls out something interesting, which is we've talked about, comes up a lot now, is these software earnings looking specifically at their AI products and the growth there. So I want to, if you have anything to layer on about MongoDB, but also wanted to ask you about DocuSign, which I feel like is the punching bag of SaaS, but they report today too, and the big thing everyone's looking for is their AI growth as well. So I'll give the floor to you.
No, just starting with, I agree with Shai on his assessment here. And I especially like founders and CEOs that are very bottom-up, tech, nerdy, geek-focused, which is the point I think you're making, that they're less concerned about hyping the stock from a quarter-to-quarter point of view and just look at it from a bottom-up perspective. And that to me is actually some of the most attractive stuff that I look for in these kinds of things. And so yeah, there was a tonal shift here that is interesting. And obviously there is the macro headwind of AI and vector databases and unstructured data and AI and training and inference and all the stuff that we talk about. But the core thing, as you said, Shai, is just quarter-to-quarter execution, get the funnel in, start seeing the traction. And so they seem to have surprised, and you know, we've got to watch the trading action relative to float and liquidity and all of those factors, technical and financial, and then the fundamentals. So definitely worth paying more attention to, and I like how you're reassessing it.
Also, they announced an $800 million share purchase plan. That's always money where their mouth is, essentially, of like they think it's undervalued and the narrative is changing. I was like, wow.
Yeah, that's always a plus sign. So on DocuSign, look, I mean, this one has been kind of in the doghouse a lot of issues for a very long time, you know, especially, you know, with everyone has been so focused on the shiny new AI toy and what does this thing do in that environment. So it's been a show-me kind of a thing. There's a lot of questions of how they productize the AI components as AI and, you know, where they are even in terms of thinking from a context of, you know, first principles and AI-native kind of work. So it still feels to me a work in progress. I don't know, Shai, if you have different views.
I think it's been a dead stock for four years, and they're in the uncomfortable middle of being too big to get acquired and too little of an add value for that price tag of, I don't know, it has to be like 25 bill probably. Like, who's going to pay 25 bill for a company whose product is a commodity? Like, it's a low barrier of entry to create. There's no network effects at all through DocuSign. They are a dead, just zombie company. Maybe there's some new thematic in the future that I think can capitalize on with the data they have in their ecosystem. But DocuSign and Zoom, there's just no future for those companies. They're too big, financial engineering. They're going to constantly do buybacks. They're profitable, but like they don't grow top line enough, and it's just going to be opportunity cost, opportunity cost, opportunity cost if you own this name.
I do, this DocuSign conversation reminds me of something you've been writing about, Michael, which is this concept of AI hallucinations, because I do think the legal application of AI is interesting, but it's very not perfect and it's very risky. And AI hallucination, essentially, we've all had that experience where like sometimes they just make stuff up. Like I've been researching something and they're like, oh, 18% of this is this, and I'm like, wait, is that true? And it's like, oh, no, my bad. I'm like, wait, what? So tell us, you wrote about this, tell us about AI hallucinations and how that limits some of the potential of softwares like this that are trying to sort of replace that human layer, but it's really not perfect.
Yeah, hallucinations in AI is a very critical issue, and I wrote a post yesterday where I basically said, look, for now, think about AI hallucinations as a forever problem. In other words, as we all just get used to using AI, whether it's in our businesses or consumer, just take it as a matter of fact that 20, 25% of the results you get are not accurate. So you have to check it. You have, regardless of what the companies, OpenAI, Anthropic, you know, the biggest LLMs or the smallest enterprise SaaS companies, whatever they tell you they're doing to reduce hallucinations, you cannot believe it. Why? And for here, you have to go back down into the secular way AI technology is fundamentally different than 70 years of computing. 70 years of computing is just, you know, it's ones and zeros and calculations precise. You can count it out to the 19,000th decimal place. AI does not work that way. It's probabilities on questions. You know, someone very smart three, four years ago talked about AI as stochastic parrots. You know, these are statistical repetition machines, and what we're doing is putting ungodly amounts of NVIDIA GPUs to work, and then they run extraordinary amounts of matrix math, run these intelligence tokens and so on, but it's all probabilities, so you cannot trust that the actual answer is what you think you will get. So even when you buy OpenAI's product for $200 a month or Google's product for $250 a month, and tomorrow we'll be paying subscriptions of $2,000 a month, don't expect 100% accuracy. Why? Because of the nature of which these systems work. They are fundamentally incapable of reducing hallucinations because the researchers are stupid or, you know, the technology can't be harnessed and so on. But for now, it's just the nature of the beast. So what does that mean? If you're using it as a consumer or as a business, make sure you have discipline to double, triple, quadruple check your results and then run it. So when you talk about it in the legal field, you know, Harvey, all of these great startups in verticals in AI, they've learned that lesson and they're executing on it, that when they work with their top clients, whether it's the law firms or the customers of the law firms, they make sure that they have both humans in the loop and then they have technology around with traditional hybrid approaches of computing to triple, quadruple check the results that the citations in the court cases are actually real. And that is a friction point that you have to accept in the process. Even though, you know, as a technology company, you'd rather not, you'd rather just take the result and run with it, but that's not the way AI works, and it's probably not going to be the way that's going to work for at least the next three, probably five plus years, just the nature of the beast. I'll just say one more thing. You know, we know about the godfathers of AI, you know, the Hintons and the Bengios and so on. Yann LeCun is one of them. He's the head AI researcher at Facebook, and he literally a few months ago, you should read his interviews and podcasts, he literally said LLMs are a dead end. He's at this point, you know, because of the hallucination problem, he says, and he goes through technical reasons why of this that I kind of tried to summarize at a high level, but he said there are approaches to fix hallucinations, but they are in new directions that are not part of what the hundreds of billions of dollars are being spent for AI today. There will be the start, next startups will focus on this. So it's a problem that needs to be solved. The solutions are still in the labs and research houses. It'll be here for a while. So hallucination is an issue. So as we're talking about, Shai, for AI in enterprises and so on, this is a critical issue. This is what the big companies are absolutely realizing, that this is not a magic tool. Many CEOs have been convinced otherwise that this is just something that we can replace a PhD with or a scientist over there. Uh-uh. We're going to need more PhDs, more scientists for a lot longer to make this work. Just by the way, just like, you know, Uber, Travis thought, you know, we were going to get rid of drivers in the cars, you know, 10 years ago, and today we need more drivers supervising or teleoperated drivers who are watching the Waymos and the other equivalent driving cars. Humans need to be in the loop because they will hallucinate. There will be long-tail data recognition problems that will cause problems, and that's an issue. So hallucinations are here, a forever problem for now.
Mm-hmm. That's fascinating. I mean, Shai, you probably relate. I think everyone has that moment when you're doing research about something you actually know about. That's when you see the limitations. It's like you're like, wait a minute. Like, Shai, you're looking up a company that you know like the back of your hand. It spits something you know, but you don't have that insight, that level of insight when it's an area you don't know. So I think it's a good, just as users, but also when you look at these businesses, how are they going to get around this? Because, I mean, trust is so important with these software solutions.
Yeah. I mean, for me specifically, I can't use AI for facts at all. I use it for Grammarly, essentially. It's just a Grammarly thing for me. Like, I hate grammar. I just like to put my content in writing and it's like, make this grammar make sense. That's literally it. But the guardrails of AI, like, it's going to be here for a while, like what Michael was referencing. Like, it's not the intelligence is going to get replaced across a lot of industries, but it's not going to be replacing headcounts as drastically and as efficiently and as quickly as people anticipate. Yes, I think inflated teams of 10 are going to turn to six or seven, but that six or seven to three to four, that distance will take a long time. I think from 10 to six very quickly, but six to four, that's going to take a much longer time. And that's what I think Michael's referencing. And I do believe, though, like what a comment that Michael mentioned, I don't know if it was last week or the prior episode, where Google was founded like nine years after the dot-com era, essentially. I do think that that next layer of like all the solutions are now going towards LLMs or expenses are going towards LLMs, that's not really solving the terminal solution. There's going to be, once they make that pivot away from the initial phase to the second phase of whatever the next derivative of AI will be, there's going to be a trillion-dollar company that gets created during that process. That's what's exciting, and that's why it's so important to do your research now. Learn the MADCIC and see how it progresses because everyone thinks, oh, the Nvidia is a Palantir, the move already happened for the winners, disruptors. No, that's a first phase. There's multiple phases. There's going to be another Nvidia, another Palantir sometime next decade.
Yeah, I mean, and go ahead, go ahead, Michael.
No, just because you mentioned Grammarly, because that was literally the name I was thinking when we were talking DocuSign, because here's a very relatively boring company, been around for 20 years. They just signed up to raise another billion dollars on a performance set of metrics. You should go back and look at it. They're taking traditional grammar checks and now starting to add AI components to it and growing their market very slowly, just like DocuSign, and now all of a sudden they have a spur opportunity, and that's why General Catalyst is going to put in a huge chunk of money on performance. They're not even getting direct equity. Just as a kind of a rabbit hole, if people are interested. Just talking about how technology takes a long time to perfect, to apply even to a field that is supposed to be in the wheelhouse of AI, which is grammar checks and summarization and so on. And then the last thing, Katie, to your point, you know, we're talking about fact-checking and accuracy from an AI, and that, you know, again, hallucinations will be for a while, but what you were saying where it comes up with new things, new ideas, that is the level three, level four on OpenAI's map. It's in my post that I put up on the one I did a couple of days ago. And so right now we're going from chatbots, which is level one, reasoning, level two, agents, level three, and that's basically occupying all of us for the next two or three years. The ones after that are innovation, and that's where AI is theoretically supposed to come up with new things, and that is even tougher to do with today's tech LLM technologies for a whole bunch of reasons. Yann LeCun talks about this because today's AIs are basically hoovering everything that humans know and are put online and is running stochastic matrix math on that. There's not enough stuff there for them to come up with innovative new ideas. The only place where we're seeing some of the glimpses of new innovation potentially is things like AlphaGo, you know, DeepMind's things on the healthcare side, because they've orchestrated that specifically for the new innovation stuff. So that's to come, and that's another place where I mention all this because especially CEOs and CIOs, when they're investing billions of dollars of their IT budgets, sometimes they get ahead of themselves on their expectations that the AI is going to do all these amazing things, that they will need less human people to do their work, and they'll be able to rely on these technologies to do what humans do today. And I think it's going to take a lot longer for that to happen.
I mean, the health, the healthcare implications of this is fascinating. I'm in this like meetup group in New York that's decentralized science meetups, and they talk about this all the time, like, and it ties into like university funding and like when this stuff, when you could run tests and simulations using AI in like a fraction of the time, it gets really exciting. There's the opening bell. We're going to bring in our next guest, but real quick, I want to just touch on Google for a moment because they are battling it out in court. Actually, hello, Niccolo. Welcome. Welcome. How are you doing today?
Hi, guys. Can you hear me okay?
