OpenAI & Anthropic Execs Leading the Pentagon AI Showdown, Amazon Explores AI Chabot Ads

18:59Stephanie Palazzolo:They already have so much access as it is.

19:02Steve Koenig:Yeah, and they might want to say, hey, we have a unique audience that maybe not all of them are using Amazon. I don't know if that's true, but maybe that's true. and we want to keep that to ourselves or we don't want Amazon to know how our audience is shopping on ChatGPT. And that would be fair, but we'll see. Great.

19:18Stephanie Palazzolo:Well, Catherine, I want to thank you for coming on. That is Catherine Perloff, our Amazon reporter here at The Information. MongoDB reported its quarterly results last night. Shares were trading more than 25 % lower this morning. The company said its chief revenue officer was leaving the company to understand what is going on. I want to bring on Steve Koenig, U.S. Head of Software and Services Research at Macquarie. Steve, welcome to the show. It's great to have you here. Thank you, Akash. Thanks for having me. What did you make of the results last night?

19:50Tom Schmidt:Well, the results were okay. You had Atlas, their cloud database, decelerated by one point, but still growing at 29%. Apart from the departure of the CRO, the other big negative was in the minds of investors was the guide, the initial guide for fiscal 27. The revenue guide skewed down from consensus. You know, the low end, I think, fell something like 40 million below consensus. So that was taken as a negative. And then, you know, the departure of the CRO with no replacement, you know, that was also a concern, you know, for investors too. So that, you know, hence the big down move in the stock today, we see some big, pretty big outside moves in this space especially lately do you think that this is a mongo db story today the stock shop or a sas apocalypse story today well mongo db is trading trading down a lot more than everyone else you know uh but but yeah we're seeing you know i think i think software is um you know software's certainly multiples have gotten compressed uh significantly in the last couple of months, but today's price action looks to be more kind of geopolitically driven across the board than some kind of like major resurgence of AI fears.

21:17Tom Schmidt:The AI fears are out there and they're going to stay out there for a while.

21:20Stephanie Palazzolo:And more what I meant was, let's say maybe two months ago when the SaaSpocalypse was not so much in the conversation, let's say you have the CRO leaving, I mean, do you think the stock drops 27 % or do you think that that is very much amplified today by the SaaS apocalypse fears?

21:39Tom Schmidt:Yeah. Well, I think it's both. I think that's fair. But I think you've got a big part of the price action here is it's unsystematic risk. It's negative alpha associated with the events going on at Mongo, the guys and the CRO leaving. So I think it's both, but mostly it's company-specific stuff going on here.

22:04Stephanie Palazzolo:And do we have any rationale or reason as to why the CRO might be leaving? I mean, I was looking at the results, and I mean, the company did seem to be accelerating a bit this year. Then the guide is kind of what's dipping it lower. It's a bit of a boomerang. What do we know about why they might be leaving? Yeah.

22:21Tom Schmidt:Yeah. You know, we don't have the inside scoop, you know. I'm not sure, you know, and I'm not sure, you know, that the info is out there in the public markets. If I had to guess, you know, I'd say, you know, the new CEO, CJ Desai, he's an operator, you know, and he gets out in the field, he talks to customers, and he sets the tone on, you know, what happens in the sales organization, no matter who the head of sales is. And so, you know, my sense is there's probably some degree of, if not outright friction, just lack of alignment between those sales leaders. The head of field ops also left. So lack of alignment between the CEO and the sales leaders and whether they got pushed out or whether they decided to leave, it's unclear.

23:08Tom Schmidt:But I'd probably most likely chalk it up to kind of alignment with the CEO. Yeah.

23:15Stephanie Palazzolo:Now you talked about CJ being out in the field. We actually, we had him on our show a couple months ago and he said he was spending a lot of time talking to customers. And one of the points that he made on the call last night was in those conversations with customers, something that has come up is that these agentic workloads or basically the agentic products that everyone is waiting for enterprises to adopt. He said, it's really not happening in the ways that investors might hope. And I was struck by the comments because it was some of the most candid comments that we've heard from CEOs talking about this.

23:50Stephanie Palazzolo:And I wonder what you made of that. I mean, look, I'm not going to say that the stock drop was because of those comments, but MongoDB, this is an infrastructure company that I would have thought would have at least had a little bit of a tailwind from people adopting AI.

24:05Tom Schmidt:Yeah. Yeah. Well, Kash, I think you hit the nail on the head there. You're spot on. And I was struck by the same comment as well. And I think the surprising thing to me about the comment isn't the content of the comment. It's the forthrightness of that commentary, which echoes a lot of our checks in the field. Certainly indicate that over the course of the last 12 months, adoption of a Gentic has been slower than what people initially thought maybe a year ago. So, you know, that's fairly clear. And there's a whole host of factors. But I think Mongo, you know, the CEO coming out with that was maybe surprising.

24:50Tom Schmidt:But keep in mind, like, there's not a whole lot of traction yet from what we can see with their AI products in the enterprise. They're just starting to get some traction marketing it to the AI native privates, you know, which Mia Culpa, they made a Mia Culpa, said they should have done that. a year ago or two years ago. So they want to probably some motivation to explain why there's not more traction with enterprises. But I think his comments about the backdrop are largely true.

25:23Stephanie Palazzolo:Great. And last question for you, Steve. I always try to get a little bit clearer on what the actual products are that these companies are. Can you try to explain to us, as if you were talking to the smartest person you know outside of tech, What is the actual database software that MongoDB actually sells?

25:40Tom Schmidt:Yeah, Mongo sells a really differentiated database product. So, you know, this is data management and it helps companies persist data. And Mongo's database, their key product, Atlas, runs in the cloud. And it's designed to power operational workloads. So what that means is when companies have transactional applications, you know, whether it's e-commerce or supply chain applications where people are reading and writing data to manage business operations as opposed to like analytic or data warehouse applications. That's what Mongo is really used for. And then over the last couple of years, they've supplemented that product with acquisitions and internal development to help it be useful in really agentic use cases where you're going going to also put AI against those operational data stores.

26:33Tom Schmidt:So that's really what they're all about.

26:35Stephanie Palazzolo:And do you think that Mongo is going to be a winner in the AI era here? Or are you thinking that today's stock drop is actually reflective of what's to come?

26:44Tom Schmidt:Yeah. Well, I think the AI strategy, it's still nascent. There's more products, you know, products they need to develop. They want to become part of the reference architecture for people, for enterprises. that are implementing agentic AI and for the AI natives that are developing, you know, frameworks and LLMs, right? And that's, it's still really early days for Mongo to be relevant there. And I'm going to plead agnostic on that. I think the verdict is still out there. And, you know, I think investors are going to need a lot more proof that there's actual traction here in terms of monetization, you know, either with the AI natives and or with enterprises.

27:25Tom Schmidt:Great.

27:26Stephanie Palazzolo:Well, Steve, I want to thank you for coming on. That is Steve Koenig from Macquarie here on TI TV. The big AI funding rounds are not showing any signs of stopping. Basis, an AI agent's accounting company, recently raised$100 million led by Excel at a$1.15 billion valuation. I want to bring on Matt Harp, CEO and co-founder of the company, to talk to us about what he is building. Matt, welcome to the show. It's great to have you here. Thank you, Akash. It's great to be here.

27:55Matt Harpe:Tell us about what you're building at Basis. Yeah, so we build an AI platform for accounting teams. So we have long-running agents that are able to actually do real work for accounting teams that allows them to automate a substantial portion of the manual work that's involved in doing this accounting work. We currently serve about 30 % of the top 25 firms in the country and 20 % of the top 150. and we're essentially allowing those firms to take on a bunch more accounting work on the back of these AI agents.

28:27Stephanie Palazzolo:And how did you get into accounting here? What do you have accountants in the family? Were you an accountant in a past life? How did you get so passionate about this space?

28:35Matt Harpe:Great question. Well, I personally have no background in accounting. Mitch, who I started the company with, has a background in accounting and was working at an accounting technology company. My route to accounting is slightly circuitous, but maybe potentially interesting. I was focused on machine learning projects in healthcare, was working with a bunch of healthcare organizations. And I had a kind of crazy realization in that process of working with healthcare organizations, which is if you go to any hospital in the US and just ask a very simple question, which is how much does it cost us as the hospital to provide a procedure to a patient?

29:06Matt Harpe:Let's say, Akash, you're going to go get an x-ray or an MRI or something like this at a hospital system. There's really no hospital in America that can actually answer that question for themselves. And this is just a ridiculous proposition because if you're a hospital, you should probably have at least two goals. One is to increase the quality of care for people and the other is to reduce the cost of care for people. And if you can actually measure the cost to serve in the first place, you really don't know whether you're getting better or worse at that second objective over any given period of time.

29:34Matt Harpe:You don't really know how the decisions you're making as an organization are affecting one of the core outcomes of your organization. And so our view is that that is one of the core jobs of accounting. that fundamentally accounting is about structuring and systematizing all the economic activity that happens in and around an organization so that people can understand it there are certain compliance things that accountants have to do like doing the appropriate structuring and systematizing for preparing a tax return or you know other filings for preparing for an audit where there's all this other accounting work that simply doesn't get done um right could do we probably should do and so this kind of realization got me quite interested in the importance of accounting in the functioning of our market economy and led us to start thinking about what we could build.

30:12Matt Harpe:And then as we started thinking about, okay, how can we leverage the emerging abilities of these large models to sort of change the way that we work? It became very clear there was a significant opportunity to do this in accounting. And then we started building and we're fortunate to find a lot of excitement among the initial set of firms that we started working with.

30:28Stephanie Palazzolo:Okay. So that's a great story. Now it leads me to my next question, which is that, how did you get all of these investors participating in year round? Because I'm just going to read some of the names here and it's not a small list i mean excel led the round but then i mean you have coastal ventures on board then you've got lloyd blankfiend nat freeman daniel ghost aaron levy jack altman kyle vote adam d 'angelo i mean how do you get all these people on board man well we're yeah we appreciate the question and we're very fortunate to have

30:56Matt Harpe:a great group of folks uh around the table i think one of the things that uh we think is is really important is as you're trying to build a company there's so many important questions that uh come up. Oftentimes, we like to try to think about them from first principles, but there's so much scar tissue that other people have that they have experiences that they have that are useful to us as we think through these decisions. So having folks around the table that we can loop in and who have some context on the company when we come across these various items is super helpful. Right.

31:24Stephanie Palazzolo:But how did you get such an all-star roster? What did you learn about fundraising? It's a big valuation. It's a lot of money. I mean, what surprised you about the funding process? What did you learn through it? I mean, there are people watching the show that are going to be doing fundraising themselves too. How do you get these executives on board?

31:45Matt Harpe:I think I've come to realize over time that the most important thing about fundraising is just to build a great business, to have a clear vision of where you want to go, to state that vision very clearly, and then to exudicate against that vision at each sort of stage of the company, we put together a memo where we just write out, this is what we're going to do as a company. And then when we get to the next stage, we look back and say, did we accomplish those things? And what are we going to do going forward? And I think being extremely clear about what you're trying to do and then delivering against those results is a clear recipe for success.

32:18Matt Harpe:I think if you do those things, a lot of the other aspects in terms from a fundraising perspective, take care of themselves. People want to be involved in the company because they think there's something interesting going on. And so you're able to bring a great group together. So really it's as simple as that. And we've just been fortunate that a lot of folks who have great experiences in various areas have participated in the company and been great advocates for us.

32:41Stephanie Palazzolo:So you're building agents and outside of talent. What's the biggest cost for you right now? I imagine compute is pretty significant. Absolutely. For us, they're clear too.

32:52Matt Harpe:Number one is talent. Number two. Yeah.

32:55Stephanie Palazzolo:So how do you manage compute costs? I mean, I know you're early and, you know, margins are still something that you probably may want to worry about second after you develop the product. But, I mean, what's the advice that all these advisors are giving you about how to manage those compute costs, which, you know, they're kind of slow to come down?

33:14Matt Harpe:Yeah, I think there are two things to think about. The first is how do you manage compute costs today and what's the impact of that on margins going forward? I actually think there are two totally different questions. Our view, at least internally, is that the cost per unit of intelligence has dropped precipitously over the last couple of years, and we expect that to continue. I think if you believe that trying to calculate the margin profile today is useful as a point of reference, but it doesn't actually tell you that much about the margin profile going forward. The other factor, of course, is that as the abilities of these models continue to improve, the types of things that you can do continue to expand, that you can provide more value to your customers, you can take on more complex workflows.

33:56Matt Harpe:And so you're constantly doing more things as the margin profile changes. So I think it's very difficult to forecast those things. We're quite confident that if you just drive significant value, if you're able to essentially end-to-end complete a lot of these workflows, that the margin profile makes a lot of sense. And so then today, the question in terms of managing compute costs is making sure that you have good visibility into what the agents are doing, how much inference they're using for various tasks, and to make sure that you're sort of operating in the best possible way. And so we just have a system that essentially monitors that with a fair amount of granularity across all the various things that the AI is able to do.

34:32Matt Harpe:We keep an eye on that and make sure that it's in check with sort of the expectations that we have. And if it's not, then we take a look. And there are usually ways to optimize it if it's not. So that's our approach to it. I think that the question of monitoring today and thinking about long-term margins are actually largely independent, at least in terms of how we look at it.

34:46Stephanie Palazzolo:And very quickly, are you using OpenClaw at all? You're in the agent's world. Are you using it?

34:53Matt Harpe:We certainly don't use it for any of the work that we do in terms of building agents for customers. I think there are lessons.

35:02Stephanie Palazzolo:But personally, I mean, could you see a roadmap at all, you know, in which your personal use of it maybe could start to come into play with the program?

35:12Matt Harpe:That's a great question. I think there are lessons from OpenClaw about giving agents autonomy using as much intelligence as possible that probably are pervasive elsewhere and that you can learn from. We actually have a bit of a different approach, which is that we have a team internally called Atlas whose only job is to make everyone at the company 10x more efficient. And that includes me and includes everyone else at the company. A large part of what they do is they figure out what's the internal infrastructure we need to build as a company so that we can essentially let agents loose to complete all sorts of various workflows in the most secure way possible.

35:45Matt Harpe:And of course, OpenClaw, I think, has all sorts of security challenges. It has all sorts of challenges in terms of collaborating across companies. And so a lot of the internal infrastructure that we build essentially allows us to put similar, very capable, long-running agents to go work on various tasks, but in a very safe and contained way. And so I think you might call that a version of OpenClaw in some sense. It's giving it wide-ranging access to various things and sort of empowering it to take on large tasks, but doing it with core company infrastructure. And so that's part of what we've been building for almost a year at this point internally, which gets you, I think, a lot of the benefits that people have started to realize with OpenClaw.

36:20Stephanie Palazzolo:Great. Well, Matt, I want to thank you for coming on. That is Matt Harp, the CEO and co-founder of Basis here on TITV. Crypto venture capital company Dragonfly Capital last week raised$650 million in its latest venture capital fund. It is not every day that we hear nowadays crypto venture firms raising new money, especially amidst this latest crypto winter. Bitcoin is down nearly 40 % over the past six months. I want to bring on Tom Schmidt, a partner at Dragonfly Capital, to talk a little bit about what he is investing in in this moment. Tom, welcome back to the show. It's great to have you here.

36:56Stephanie Palazzolo:Thanks for having me. Okay, can you talk about how difficult it was to raise this fund? Because I mean, let's just stack the odds here, okay, for a second. And I think you know where this is going. But we've got crypto bear market, we've got VC fundraising bear market, we've got, I don't know, crypto VC fund bear market. Yes.

37:20Catherine Perloff:We like to play on hard mode. How did you do it? I think this is a little bit of a misconception or misnomer in that, yes, obviously VC fundraising has been going down. Crypto fundraising has been going down. But in practice, it's really just consolidating into the best performing funds that are winning both across crypto venture and more traditional venture. To your point, LPs are very pro-cyclical. And when you're looking at the public market, when you're looking at something like Bitcoin, Centemic tends to kind of track that. But when we look at crypto, I think really like the aperture of what we invest into with our Fund 3 and increasingly our Fund 4 is broad.

38:02Catherine Perloff:I think people are maybe over-indexing on this kind of NFT creator economy story that was in vogue a couple of years ago. But in practice, when you look at the up and coming new crypto companies, it's stuff like Polymarket, which is running on stable coins, is running on Rails, or it's stuff like Rain, which is a new company we just backed that allows you to spend stable coins with a debit card instead of going to have to have to have to off ramp through traditional banking infrastructure. And so if you look at overall investments, I mean, YC just put out a bunch of RFSs for stablecoin based startups.

38:36Catherine Perloff:It's more about, hey, crypto being this sort of underlying financial plumbing, this new sort of substrate that you're all doing all these more interesting businesses on top of versus like crypto just existing for the sake of existing.

38:48Stephanie Palazzolo:I want to get to the to the startups in a second here. But with the fund itself and the fundraising, was it mostly returning investors? Did you get new investors?

38:58Catherine Perloff:pretty much every existing LP re-upped and we got a couple new LPs as well. We're pretty institutional these days. And I think people who have a long-term time horizon, if any, see this as a great opportunity to be deploying in the space. I think crypto has died, you know, four or five times at this point. I think I've lost count. And it's kind of found a way to reinvent itself after every cycle. And it always looks a little bit different. And I think that's also why LPs like working with a fund manager who is spending day in, day out, looking at what's actually happening on the ground, finding new companies and trying to spot those new trends, which is, I think, what we've had a good track record of doing here.

39:33Stephanie Palazzolo:Now, you talked about prediction markets and stable coins as two areas that you're looking at with this new fund. Are there any intersections with AI? And I don't mean to bring everything back to AI because that's not what we try to do here. But I have been hearing that these agentic protocols that are coming up, I mean, you know, one possible way to navigate the payments issue here is with blockchain. And look, I don't know if it's the same group of people that were working in crypto that have now come to AI that are saying this or if there's actually potential here. But have you seen startups that are working on this?

40:09Stephanie Palazzolo:Is there a real intersection?

40:11Catherine Perloff:Yeah, I think in many ways they're very complementary technologies. You know, I think if you look back at the intersection, we haven't really done any investments. They've been things that have been focused on, you know, decentralized compute, decentralized inference, decentralized training. I think for a variety of reasons, those don't really make sense. You have this sort of inherent cost to decentralized compute that just adds to the already enormous cost of training a new model or, you know, running inference on an existing model. But kind of the interesting things that we're seeing now, or to your point, more around agentic payments where, hey, if you want, you know, I think Stripe recently put out a post saying that they, you know, imagine a future payment system that's going to require a billion transactions per second.

40:55Catherine Perloff:When you have all these agents making micropayments, you know, buying a fraction of an API or doing cross-border payments, it makes sense they would maybe want something that is global and programmable and 24-7. You're seeing things like Hyperliquid and LiDAR, which are two perp-dexes, really leading the market. I mean, even this weekend with the news in Iran, people are trading commodities pre-open on these decentralized futures platforms. And so you can get a sense of what the market sentiment is going to be before the futures open on Sunday. And I think, you know, interestingly, there's also this overlap of personnel between crypto and AI.

41:28Catherine Perloff:People forget, you know, Sam Altman was also a co-founder of WorldCoin, which is a, you know, decentralized humanity verification startup. And recently, I think there was a leak or a post, I think maybe the information picked it up actually, that they're looking at doing some collaboration with OpenAI to create a human-only social network. And so it seems like there is going to be things that's kind of overlap between the two. I think people are also just excited about distributed compute. So for us, we're an investor in Exo, which is a company that allows you to basically shard inference across multiple different Apple Silicon devices on your network.

42:02Catherine Perloff:So if you want to run, you know, an open source model that's kind of beefy, you don't have to go and buy like a really, you know, beefy, you know, GPU or like a, you know, H200 or something. You can just run this on, you know, not to meme too much, but a Mac mini or a couple Mac studios that are running on your network. And, you know, that allows you to really take advantage of edge compute. So I think the two will continue to collide, you know, naturally in a world where there's insane abundance and insane just flooding of content and things coming out of AI. Crypto is sort of this natural scarcity function, this natural sort of gating function.

42:35Catherine Perloff:And we're looking for ways that there's going to continue to collide in front of war.

42:39Stephanie Palazzolo:On the talent piece, how much harder is it to get the smartest people to come work at these crypto startups now, given that now there's this other shiny thing that wasn't really in vogue three, four years ago?

42:52Catherine Perloff:The talent wars are very real. And we do see, again, talent going back and forth between the two. I would say sort of mainstream crypto these days is actually competing more with traditional finance and financial services and pulling people out of those. So when we see people, you know, leaving JP Morgan or American Express or Visa and going into crypto, that's really more where the industry is going. It's kind of building this new sort of financial substrate. And so, I mean, you're right that, you know, hey, talent is scarce. And, you know, I think people see those multimillion dollar comp packages coming out of open AI.

43:25Stephanie Palazzolo:but um well even for payment i mean i'm just thinking about payments here you know somebody is working at jp morgan let's say or a big bank they have an option to go work at a crypto company or they could go work in this thing agentic commerce agentic payments you know open ai's building i i don't know i just feel i i feel like um it's it's hard yeah i i think the two are are

43:49Catherine Perloff:you know really again you know one of the same if you look at what's leading the road uh or leading the way in agentic payments. It's people implementing X402, which is this new standard for doing payments over the internet, Stripe's adopting it, Coinbase is adopting it. And so, if you think about, hey, where is the future of commerce going to go? It's probably going to want something programmable, something definite, something global, something you can custody. And already, I think we're seeing a lot of companies coming out of accelerators, coming out of seed, focused on this part of the market.

44:19Catherine Perloff:So, who knows? Maybe the future is also you get one dude in a garage vibe coding, some sort of thing. I think that's maybe more likely than not.

44:31Stephanie Palazzolo:Right. Great. Well, Tom, congrats on the fun. I want to thank you for coming on. That is Tom Schmidt, a partner at Dragonfly Capital here on TI-TV. That does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. I want to thank you all for tuning in. We really do appreciate your viewership. Make sure to subscribe to the information on YouTube and follow us on x instagram and tiktok check us out wherever you get your podcasts i'm already excited for our next show tomorrow have a great rest of your tuesday bye-bye for now