80,000 Cold Emails, One Surprising Result | AI Venture's Bubble Or Boom Moment | The Moat Myth

I think part of the problem in cold email is a tragedy of the commons. Most of it is terrible. Even if you're a credible person and you have a good message, you're competing with all of the perception that it's junk because a lot of it's junk. What I always think is, why is this person reaching out to me cold? I wouldn't tell most people to go cold as the first option, but if your choices are not reach out to the person at all or write a really good cold email, cold it.

It's the week of July 6th. Welcome to the Learning Corner, a Precursor Ventures podcast. Welcome to the Learning Corner by Precursor Ventures. I'm Mia Farnam, and I'm joined by our GP and managing partner, Charles Hudson, where we will be talking through the articles and Goodreads that caught our team's attention over the past week. Our first piece this week comes from our friend Samir Khaji, who draws on his 27 years in private markets to argue that we are living in two truths about AI venture all at once. On one hand, the froth is unmistakable. Companies are repricing every 5 to 6 months—Precursor Venture Ventures, everyone knows went from roughly $2.5 billion to $60 billion acquisition by SpaceX in about 18 months—and hundred million dollar plus seed rounds are pricing in the option value of being early rather than the business itself. The access market has grown noticeably undisciplined too, with these multi-tiered SPVs piling fees on top unauthorized secondary transfers. And on the other hand, every supercycle has produced bigger winners than the last, and this one really is no exception. SpaceX's roughly $1 point $7 trillion IPO, we've got Anthropic's raise close to $1 trillion itself, and OpenAI at $852 billion could together just exceed the total exit value of the last decade in venture. Samir's conclusion is that the percentage of real winners will be far smaller than the percentage of companies people believe are winners, which means the split between winners and losers this cycle will be more extreme than anything venture has produced before. His practical takeaway for allocators is really blunt. Portfolios without at least one top 5 to 10% asset of this cycle will significantly underperform. Perform, an average exposure to the AI theme will be worth close to nothing. So Charles wanted to pause here first. Curious if you feel like the broader sentiment beyond Samir's article is also headed in a similar direction to what he was saying.

I feel like there's been this interesting vibe shift in venture where for the last 2 years it's been like, just invest in AI. It's the only thing that matters. It's where all the alpha is going to be. There's going to be a lot of winners. To a market where that's not the case, where everyone's like, oh, okay, well now that like the dust has settled a little bit, maybe it's only a really narrow set of companies that are going to make the majority of the money and the rest of the stuff is going to be less valuable. In my mind, that's a pretty big narrative shift from people, which is that maybe 3 or 4 years into the cycle, we have a better sense of the companies that are most likely to be valuable. And does that mean that everything else ends up being remnant? And I, I don't know if that's the answer. It goes back to this whole, something we've talked about. If returns are really concentrated in a handful of companies, if you're not in those companies, there's almost no way to catch up on a comparative basis. The other question I have for some people, like how repeatable is some of this stuff. If you do catch lightning in a bottle, will you catch it again? And I think if I were a fund manager who had one of those companies and was in market, I would absolutely say yes, we can. And let me tell you why. We— the process we have will help us find the next version of one of these. I do think if this narrative shift continues, there will be a moment where things change and it becomes hard for AI companies to raise. People express real skepticism about revenue terminal, and that just hasn't, in my opinion, really been happening in the sector. So I think this could be a big shift and In some ways it sort of reminds me of the turn that happened in 2021 where everyone was like, well, I guess it's okay to pay 100x revenue for a really amazing software company, but not just for any company. I think when that shifted, it really did change the way people operate and behave.

And I have to believe that some of this vibe shift has to do with the fact that we're seeing some of the most consequential outcomes hit the public market. And then you have to look at your portfolio and say, are my companies going to hit the public market in that way? Because those could quite possibly be real. And once it's real, it gives the opportunity for a lot of things to be shown as fake. And I think that had not been available to people over the past couple of years because everything could truly be quote unquote fake for so long. And so I do believe in Samir's statement that you must be in the top 5 to 10% in some way, shape, or form because the outcomes are going to be so large. On the flip side, is there some point about access that still rings through that if it's maybe, I don't know, in the top 20%, but it's a hot company that still gets rewarded from LPs. And I Maybe these are two different arguments happening, like, yeah, return versus like what's helpful for fundraising.

I think we might be going back to a world where absolute and relative return are different. So I think if you're like, hey, my goal is to produce a top decile fund, that's a relative comparison. And I agree with Samir's argument, which is that you probably do need to be in at least one of the big names because ironically, some of the people who are going to be competing with you will be in multiple of the big names. So I don't know how you get there if you're not in at least one $10-15 billion outcome. It's going to be really, really hard to get there. On the flip side, if your goal is to produce great absolute returns by historical venture standards, you can probably do that without being in one of these really big companies. And you could produce a 5, 6, 7x fund, which by historical standards would be really, really great. And maybe in this moment feels like less impressive.

The last piece that I wanna talk about is a quote that Samir said, which was, after a period of disillusionment, we will see another wave of slope of enlightenment, and that period will likely see far more expansive growth than we have ever seen from technology. How much do you agree with that statement?

Somebody called me a Luddite a few weeks ago. I was like, I didn't know that expressing even a little bit of skepticism about whether the first generation of AI companies will be the ones that matter. And when I look around, I'm just like, okay, what's really happening? We've had this boomlet of this new technology. It's clearly powerful. It's also clearly not being fully utilized by most companies. It just isn't, partially because they don't know how to use it, partially because to really get the gains, you'd have to rip out processes and start over. And it does remind me a lot of when people discovered digital technology. They're like, oh, let's just digitize faxes. So now we have like the ability to send a digital fax. Like, why don't you just get rid of the fax? Well, you have all the stuff that's tied to the fax machine, so we're gonna just digitize faxes. And I feel like we're in the digitize faxes era, which is why I think everyone's so excited about these AI-native companies, because there's no rework. You're not like, oh, we have this accounts payable process, or oh, we have this other process that stinks, we have to retrofit AI. And you're like, no, we could just decide today that we're just going to skip the landline era and go right to mobile. And so I think it's going to take another generation of people really getting comfortable with the technology, really understanding what it can do, and really being willing to like break the way that they work to imagine new ways in order for this to happen. I think that's typically what happens when you get a breakthrough technology is it feels incremental for a long time until people are willing to just do something different. And we're meeting companies who are just like, hey, we're going to build mobile first. Everyone's like, no, you got to have a website. Like, well, no, you don't. Like, we're building for an audience of people who don't care about desktop web. Desktop web is a secondary experience for them. And that produced a ton of really interesting breakthroughs. And I could say the same thing about almost every phase where there were companies who just said, we're just going to skip the old stuff. We're not— I don't think most people are there yet on AI.

Well, our second piece is from Stanford's Ilya Srebuliev, who ran one of the largest field experiments ever conducted on cold outreach, sending over 80,000 fictitious pitch emails. To more than 28,000 real investors, despite VCs predicting the emails would go nowhere. So on true cold outbound, close to 10% of investors responded with interest, and the top-performing pitches hit response rates of 13 to 17%, meaning roughly 1 in 6 VCs replied to a total stranger. The piece traces this back to asymmetry and how VCs think about regret. Nobody remembers the deal they lost money on, but everybody remembers the unicorn that they passed on. And the practical findings are very specific here. Keep the email short if you're doing cold outbound. Lead with the problem rather than the solution. Establish credibility right away and match the pitch to an investor's actual thesis rather than mass blasting a generic template. All things that I think we've probably heard in different forms. He also flags a real operational risk that roughly a quarter of the emails in his experiment never even landed because spam filters caught them. So infrastructure and deliverability matter as much as the pitch itself. The data suggests Again, that about 10% of all VC deals start with a cold email, holding steady across fund sizes and track records. Does that number surprise you? Because I know early in his piece, he mentions this experiment in a large conference room with a bunch of VCs, and they all nodded when he said that cold outbound is just not a great way to get in front of investors. But then he has this 10% number. I have a lot of thoughts about this.

I think most VCs, my hunch is they tell people cold email doesn't work because they don't want more of it. And so I don't know if you saw this on your LinkedIn feed. I saw tons of people who had the CEO of Cursor Ventures had reached out to them on a LinkedIn DM and they had like never replied or left them on read. And I think part of the problem in cold email is like a tragedy of the commons. Most of it is terrible. And so if you're going through the cold channel, even if you're a credible person and you have a good message, you're competing with all of the perception that it's junk because a lot of it's junk. And I think AI has accelerated the production of bad cold email, what I always think is, why is this person reaching out to me cold? So I think students have an unfair asymmetric advantage in cold email because I don't expect a student to have a network in venture that would land them in my inbox through a warm intro. If you've been working in Silicon Valley for 20 years and you reach out cold, then I was like, well, this is strange. There's probably somebody who knows you, who knows me, who could help bump you up in the queue. I think people who are reaching out to me from geographies or places in the country, again, where I'm like, well, I don't have a network there, so I understand why this person's going direct. And I read them. I replied to a cold email late last night. A student wrote me a funny— it had a funny subject line and I opened it and I thought the thing was interesting and I was like, nice job. Like we should chat purely because I was like, I liked the style of the email, but the ones that are hard for me are the ones that are 5 paragraphs and I get to the bottom and it's like, oh, I'm a placement agent for this company and the company is raising a $25 million mezzanine round. And I'm like, why did I even get this email? Why did I even get this email? So we try hard to tell people like, this is what's in spec for us. This is like what's not in spec for us. We talked about this internally as a team. I wouldn't tell most people to go cold as the first option, but I think if the op— if your choices are not reach out to the person at all or write a really good cold email. Cold and, you know, in our CRM, increasingly I see companies that reached out cold and we were working our way through them in the queue and then somebody else flagged it for us. And then you have this like multi-touch attribution. Well, was it really cold or was it warm? You're like, well, does it really matter? I don't know if it really matters. But this whole attribution problem, I think, I think this is why VCs would think it's lower because if you had, if someone reached out to you cold and someone else nudged them and made you pay attention to it, what's the real source?

Also, I think VCs would all love to claim that their deal flow is proprietary and relationship-based. And so there's something a little bit like if you said that a large percentage of your deal flow that you looked at came in cold, I think there's something that they're like, oh, is that like an interesting enough story to tell people to show them that like I have like a very ring-fenced deal approach, but a VC, I think, would say, like, don't stop sending me stuff if it's good. Like, if it's cold, I don't want you to do that. But like, I wouldn't go around saying that I'm answering all of my cold inboxes, emails, when in fact there is some good stuff that comes out.

I also just think a lot of people associate the cold channel with desperate, and I'm like, sometimes it is, but sometimes it's just disconnected. I know some people who are just like, I'm going to make a zero investment in cold. I'm just going to assume that like it's all worthless, which is a historical VC perspective. And I think the hardest thing to do is to do some cold. I think you'd be like, look, I'm gonna look at everything. I'm gonna give it all 2 seconds and decide if it's interesting. Then you're like, all right, well at least I looked at everything. I saw it all. Or you're like, I'm gonna look at none of it and I'm just gonna accept that like if there was a gem in there, I'm gonna miss it. I think it's hard if you're like, I'm gonna look at some cold and that seems to me hard.

And I would say the flip side is true as well of VCs reaching out cold to founders. There is an art to that as well of what's a good note to send out that's effective. And like what isn't. I had a whole conversation with my friend yesterday who works at a different fund being like, how do you approach cold outbound to founders and what resonates, what doesn't? And then also just being realistic that if there's founders with specific signals, there's 20 other investors who've probably sent them the same note. So like the same dynamic is happening. It just happens on both sides.

I find it funny that leading with a message is like, I have money for you, for someone who has no problem raising money does not solve a problem for that person. Yeah. That person's probably thinking, where do I find my first founding engineer? They have other needs, but capital's not one of 'em. And I'm surprised by the number of of people who reach out to people who are clearly gonna have no problem raising money, where the first offer is money. And I'm like, okay, it's gotta be something else. It's gotta be something else. If you want, if you wanna bubble up and get that person's attention, you have to offer them something that solves a real problem they're facing or, or a real need they have.

Well, our last piece is a LinkedIn post from Tomasz Tungas tackling the question every founder dreads in pitches, which is, what is your moat? His framework splits moats into two types: leading moats that exist at founding, so like a proprietary dataset or a novel architecture, and then lagging moats that only get earned through years of execution, like brand and scale or embedded workflows. Infrastructure companies tend to need a leading moat just to get off the ground. Snowflake's decision to separate storage from compute bought it the runway to later build brand and switching costs, while application layer companies like Salesforce, Slack, and Figma, they won without one beating better funded or better built companies purely on commercial execution, timing, and speed. Dimash maps this onto Hamilton Helmer's 7 Powers framework, noting that scale economies, brand, and switching costs are lagging by considerable construction, while counterpositioning and cornered resources can be leading if a founder actually has them at founding. His honest conclusion for most application layer startups is that the moat isn't there yet. And the correct answer to the moat question is simply, we are building one. We obviously write a lot of checks into application layer companies that by Tomasz's own frameworks have no real moat at founding. When a founder tells you honestly that they are still building the moat, what actually convinces you that they're capable of building it versus just aspiring to build one?

About 2 years ago, I started going down a similar journey, which was like, if we're going to invest in these application AI companies, there's only a couple of answers. One is like, there's no moats and it's all commodity software and it will all be a race to the bottom. I was like, we shouldn't invest if you believe that. Second, there are moats. We just haven't figured them out yet. And the only way to discover them is through discovery. So you make the investment and you go on the journey with the founders. Or the third one is the moats already exist. We just don't know how to describe or we can't see them yet, but there's no discovery. They exist. And if you can articulate speculate what they are. There are some categories, like for example, if you wanna build a foundational model, capital and GPU access is a moat. You either have enough money to like pay for inference and pay for GPUs and do all the things you have to do, or you don't. You can't, as far as I know, bootstrap one of those. So that, that's a case where I think like, I don't even think of it as a moat. There's like a minimum requirement of capital to get in and stay in those businesses. And I think that's why you've seen this sort of bimodal distribution. You have a lot of really interesting open source models. And you have a very small number of closed-source expensive models and kind of very little in between by just by construct. And I think there are, you know, people laughed at me when I said this before, but I still think it's true. There are some commodity markets where simply being the most well-known, most well-referenced vendor is part of the— it's not a moat. And I think if you say like a moat is like some weird defensible thing that people can't go by, I tend to think like, well, what's your strategy for creating a winning company? And like, what do you think your advantage is? And if you're just like, hey, we're going to become the company that's synonymous with this category and in a world where all the products feel the same, we think the buyer will choose the best branded product. I think that's totally a reasonable strategy. I'd argue also a lot of vertical AI companies, if you really look at what they're doing, they're doing the classic like bundling strategy. Oh, you have a bunch of commodity parts. How do you make money? I'm like, well, because you're not allowed to swap out this piece for that piece for this piece for that piece. You have to buy the whole bundle. And so if you have to buy the whole bundle, the only people who can compete with you are people who can offer the entire bundle. And I've seen some of our vertical AI companies, particularly in like voice-enabled vertical AI, start off with products that today would be considered commodity, but now they've built 4 or 5 other things that are attached to that product. So the only way to rip it out is to replace it with the same stack. That will change eventually. So I think more like when I talk to founders, I'm just like, how are you going to win? And sometimes moat is the issue, but it's really like, how are you going to win? And if they don't have a strategy for how they're going to win, it could be we're going to be really good at marketing. We're going to be really focused on developers. We're going to be the cheapest. We're going to be the fastest. We have this channel. There's got to be some thoughtfulness around how you win. And sometimes I meet companies that have no plan. Think we are better than everybody else. I'm like, well, I I hope you are. Yeah. And I have some of those in our portfolio too, where I'm just like, I still don't know in the long term how the customer's gonna decide between 2 or 3 different competitors.

And the moat's going to continue to evolve, especially in this market. And I think anyone who tells you otherwise, I think is trying to probably push their own book of business. Of course. But I've found that like, if you're pitching authentically and you can tell them the moat will come down the road, but here's how I'm gonna plow through for the next round with the capital I'm raising. I find that to be a lot more thoughtful than creating an arbitrary answer about like how your dataset is proprietary when maybe it's not. And you're doing something that's similar to 3 other competitors, just tell me what your plan is with the capital and how you're planning to run and execute on those milestones. And we can figure out the moat as it's happening because you can also have a plan on how to build a moat and it's not working. And then we have to pivot and do something different, which happens all the time for successful companies in the end. Well, we are all out of time today. Thank you, Charles, for joining in. Thank you to our subscribers for joining as well. Please like and share with friends and we will talk with you all next week. Thank you. Bye!