Sam Silvershein:
Welcome to Driving Alpha, the podcast where outperforming investors share their insights and paths to success. This show is brought to you by Alpha Partners. We’re a growth stage fund that co-invests alongside top VCs by sharing economics with early stage investors in their most exciting companies.
Today, I’m thrilled to welcome Kyle Harrison, General Partner at Contrary. Kyle’s venture career spans heavyweight firms like TCV, Coatue, and Index Ventures, and he’s now focused on building out Contrary’s later-stage practice.
Kyle is one of the sharpest minds writing about the mechanics of our industry today through his newsletter, Investing 101, and his leadership at Contrary Research, which includes his recent 300-page deep dive that resulted in a book called The Anduril Thesis. In this episode, we’re unpacking exactly how Kyle generates alpha.
We’ll dive into Contrary’s unique people-centric sourcing model that identifies and invests in top-tier talent even before they start companies. We’ll also get his take on portfolio construction, navigating VC hype cycles, and what separates the best firms from the rest.
So Kyle, welcome to Driving Alpha.
Kyle Harrison:
Thanks for having me.
Sam Silvershein:
I’d like to start by talking about your career, which has taken you through some of the biggest names in venture. Yet you’ve said Contrary is the renegade of choice for building a later-stage practice. When you’re competing against massive funds, how does building a people-centric platform that invests in top talent before they even start companies fundamentally change how you source and how you generate alpha?
Kyle Harrison:
There’s a bunch to unpack there. I joke that I’ve had a very Goldilocks experience over the course of my career – I’ve tried a little bit of everything. TCV was very much a private equity mindset, at least when I was there; it was diamonds-in-the-rough hunting. Coatue is much more hedge fund – big idea arbitrage was the leading thesis there. Index is a little more venture classics.
There are a lot of different styles you can compete with. In terms of Contrary’s approach, and what distinguished it from everything else I’d done, the most significant thing is that the people-centric model of venture gets meme-ified – everybody talks about going earlier and earlier, sourcing in high school and whatever. But what it really means is building a relationship over time, and the ability to build a relationship requires having something relevant to offer people, even years before they start a company.
I used to joke that at Index and Coatue, we’d meet really sharp people very early and play the hurry-up-and-wait game: as soon as they put their hand up in the air, we were ready to give them a term sheet. But before that, there’s not much a firm can do beyond, say, inviting them to a dinner once a quarter to stay close. There’s no real product offering.
Contrary built the firm around the ethos of creating a flywheel – all the different things we can do for someone over the course of their career. Even if they’re early in undergrad, grad school, or a PhD program, we expose them to a network, get them in front of interesting companies, and source opportunities for them to invest, including leading scout checks. That gives them a learning experience, and it lets us get to know them and understand their judgment. Then, as they move into the talent world, we help them find interesting roles and connect with other people.
A lot of the companies we’ve invested in – Ramp, Anduril, Basepower – when you look closely, five to ten percent of the first fifty or sixty employees came out of the Contrary network, in large part because we helped facilitate those relationships. So in many cases, when we invest in a founder, we’ve known them for five, six-plus years, and we’ve helped them at every major phase of their career. In some cases they don’t even run a fundraising process – they just come to us because we already have an established relationship, and it works great.
In other instances, as the world gets more competitive, exposure to that broader talent network becomes a competitive edge for us, because we’re helping companies plug into that network and hire out of it. When you think about the DNA that makes up the Ramps and Andurils and Basepowers of the world, it’s that pool of people we want to help companies plug into.
Sam Silvershein:
You guys have such a breadth of network, and it really shows in the research you’re able to publish. You’ve written before that you don’t know what to think until you write what you say. How does deep research and public writing help convert your thinking into deal flow, especially when you’re competing against other funds for the same opportunities?
Kyle Harrison:
I shouldn’t take credit for that – it’s a Flannery O’Connor quote, but it is very much how I think about the world. I need to be able to research, write, and articulate something, and then I understand how much I know and how much I don’t.
There are two sides of the coin for me on writing: my personal blog and essays, and the Contrary Research brand. They were born out of very different needs, but both are advantageous in a venture career, and in being chronically online, as my wife would say.
The Contrary Research writing came out of needing to do more marketing as a firm. The first four or five years were very much one-to-one relationship building – a lot of finding people and building those relationships. But that’s not scalable, so the goal became getting exposure to a broader set of people. Eric Tarziski, the founder of the firm, and I sat down and said Contrary needed to do more marketing. We were pretty averse to typical VC content about how smart we are, so we asked what we could do that would actually be additive to the ecosystem.
We were already doing a lot of prepared-mind thinking about companies and markets, articulating our perspective on those spaces and what matters. So we put that out into the world, and it let the world vibe back – one of the benefits of writing online is that people reach out. They’ll say, “We wrote a report on Stripe and were thinking about how their business works this way, and your memo was a helpful articulation of that strategy.” Or someone considering joining a company will say our work helped them evaluate one company versus another. Or someone deep in a market will say we articulated it better than anybody else. That kind of validation, letting our ideas speak for themselves, has been an important and continuing part of why we do Contrary Research – providing a public research arm as a service to people.
The personal writing side is different – much more like what the comedian Mike Birbiglia calls his “secret public journal.” I think of it similarly: it’s primarily for me. I don’t really care how many people subscribe or read it. It’s more a consistency exercise. When I started, Rex Woodbury, an investor I worked with at Index, pushed me to focus on one goal: publish something once a week, every week, no matter what – no matter how good or bad it is, no matter how tired or busy I am. I’ve kept that goal since the beginning of 2022, so I’m coming up on four years of consistent weekly writing.
What that really forces me to do is see the world in articulable bites. When I look at a market or a problem or an incentive – I’ll hear about another VC’s misbehavior, or think about a new fundraise and what it means for the world – I don’t just think “interesting” and move on. I ask myself how I’d write about it, and that forces me to chunk it into A leads to B leads to C, which helps me better understand what I’m thinking about.
Sam Silvershein:
You started writing in ’22. How has your writing or research process changed with the AI tools we have now?
Kyle Harrison:
I love Claude – I use it constantly, and I’ve started using Claude Code to build out a personal website and a kind of personal wiki, a la the Andrej Karpathy LLM wiki. I had a debate recently with Jack Raines, an investor at Slow Ventures, about this. Writing does at least two distinct jobs: it’s thinking, and it’s saying.
It’s the same debate as when people say it’s only a matter of time until we’re all wearing AR lenses that automatically translate languages into our brains, so what’s the point of learning a new language? There’s a lot of research showing that learning a new language, or complex math, isn’t just about being able to do the math or speak the language – it’s literally rewiring your brain to think in a specific way, forcing you to wrestle with complex ideas. I think the most disadvantageous thing people can do is let AI replace that: say “I’m not going to think about it, just give me this,” take the answer, say “yep, that sounds good,” and move on without any thinking. That’s the most problematic thing.
But when it comes to something like building a bookshelf and needing to do some geometry to figure out angles, I’m not worried in that moment about what my brain will look like afterward – I just need the thing done. If something can do the math for me, great. Language is the same: if I’m just trying to communicate and the state of my brain before and after isn’t the primary consideration, then use the tool. AI is an incredible tool for shortcutting unnecessary complexity. The danger is that this becomes intoxicating – it’s very satisfying to hand a complicated problem to AI and receive a simple answer. You have to develop the ability to sit with discomfort and complexity, and writing forces you to do that the old-fashioned way.
That said, we’re constantly doing research, synthesizing a ton of different sources, identifying where to go deep, and pulling ideas out of a lot of conversations. There’s so much AI can help with, and it’s great. But I do worry about people who aren’t giving any thought to the shape of their brain today, let alone what it’ll look like a few years from now. Otherwise your brain ends up like the people in WALL-E, sitting in chairs, zooming around. That’s the fear.
Sam Silvershein:
That’s a really good point, and I think everything compounds – you’re constantly referencing things you’ve already written, and without that context, new information either wouldn’t stick or you wouldn’t know where it fits in the rest of the puzzle you’ve built through writing.
I want to talk about something you’ve written about that feels very relevant today – it’s kind of a David-and-Goliath dynamic in this industry. I’m curious about your take on these massive funds with seemingly endless capital, and how that’s changed how you’ve constructed your portfolio at Contrary.
Kyle Harrison:
I don’t know that I’ve said it quite this way before, but I actually just taught my six-year-olds’ Sunday school class at church, and we went through the Bible story of David and Goliath – so maybe that’s why this connected as you were asking. I actually think David and Goliath is a poorly used analogy. People use it to mean big versus small, but the actual story is a moral debate – David represents standing up for a belief system in the face of a threat.
The problem with applying that to what I call capital agglomerators as Goliath and cottage keepers as David is that I don’t think there’s a moral debate there. I don’t think being a capital agglomerator makes you a bad person, or means you want to build crappy companies, or that you’re filled with hubris. It’s a strategic difference – like choosing a sniper rifle versus a shotgun, not a moral one. But a lot of people have already decided where they fall on what they see as a moral debate between size and specialty, and they end up hating whichever side does the thing they think is wrong. I always stop short of passing that kind of judgment. Larger firms can have bad incentives that create bad systems, but so can smaller firms.
So I think it’s really a strategic problem. The framework I find more useful for the environment we’re in is what I call the unholy trinity of venture capital. It’s not that large firms are inherently bad – it’s that there’s an ecosystem feeding off itself that’s becoming problematic because it’s being slammed against a form of company creation that doesn’t scale the way this engine requires.
You have LPs representing massive amounts of capital – hundreds of billions of dollars, the CalPERS and OMERS of the world – who need to park two or three hundred million dollars at a time to drive yield. They’re not necessarily looking for venture-scale returns; they need something like a six or seven percent replacement rate, sometimes three or four percent, on very large pools of capital, because they have specific obligations to meet each year. Their bar is much lower, so they need to park capital. Then you have capital agglomerators happy to oblige, taking in larger and larger pools of capital and deploying it. And then you have capital absorbers – the standard way many companies now frame how they need to be built.
When you see companies raising $300 million seed rounds, or billion-dollar seed rounds or Series As, that’s capital absorption becoming the default mode of company building: demonstrate success, tell incredible stories, go after massive TAMs, raise giant war chests, hire a huge number of people, and pursue world domination in your chosen field. Elon Musk is sort of the patron saint of the capital absorber – people love pointing to his TAM ambitions, the twenty-three trillion dollars of intergalactic B2B SaaS, and saying “we need our version of that story.” What’s ironic is that SpaceX as a business didn’t raise ungodly amounts of capital. Last time I added it up, I think it was around thirteen billion – chump change compared to what we’re talking about today.
Sam Silvershein:
A lot of that is secondaries, too.
Kyle Harrison:
Totally. So Elon Musk as a North Star ambition-setter is the language people use, but the strategy is really defined more by the Ubers and WeWorks of the world – raising tons and tons of capital. Those were the early generation of these companies, and they’ve given way to the OpenAIs and Anthropics of the world, and now every company down the stack wants to be that and thinks capital absorption is their tool.
So you’ve got this unholy trinity: capital agglomerators ready to shove in a ton of capital, yield farmers letting them raise larger and larger funds, and it just gets bigger and bigger. That’s not problematic at all if companies like OpenAI and Anthropic are, first, sustainable businesses in the long run, and second, a repeatable approach to building a business. People will say there’s only a handful of companies that matter because of the power law – fine, but there are hundreds of companies getting funded on the assumption that they’ll be one of those companies, which is why they’re raising so much capital and are going to burn a significant amount of money.
My hesitation – and this is as close as I get to passing judgment – is that when the default mentality for building a business is shaped by capital absorption as a first principle, it’s like the old saying: to someone with a hammer, everything looks like a nail; to someone with a billion-dollar war chest, everything looks like a spend-money problem. The reality of most company building is that it’s not a spend-money problem. There are a handful of cases – first-mover advantage, network effects, getting a flywheel spinning, high capital intensity – where spending more money is genuinely advantageous. But that’s not the majority of business-building cases. Most problems don’t get solved by throwing more capital at them.
We have a dozen examples like Quibi to point to: just because you spent $250 million and you’re one of the greatest people to come out of Hollywood doesn’t mean you can turn on product-market fit. Actually, I think Quibi could have crushed it, and could crush it today – I found a feature on the Disney+ app where you swipe through clips of shows and movies to figure out what you want to watch, and I thought, that’s actually a really good idea, ahead of its time. Quibi could have made it, but it treated the problem as a money-spending problem when that wasn’t what it needed. That’s the biggest problem with how things work right now.
Sam Silvershein:
I’m curious what patterns you’d expect to see emerge from the 2021 hype cycle, where everything was spent on growth marketing, and once the tide went out, people realized a dollar in on marketing wasn’t a dollar out in revenue – it was really just funding your lifestyle through VC. It’ll be interesting to watch as people start using the word “bubble” again.
I want to drill into the high-capital-intensity side and move to your book, The Anduril Thesis, which you said took three years to write. You mapped out a hundred years of military history – it was a fascinating read, actually got me through a twenty-hour trip to and from San Francisco last week. You cover decades of bureaucracy and cost-plus contracting that hurt innovation and drove out talent. I’m curious about the key takeaways from that history, and from underwriting founders like Palmer Luckey, for listeners who are looking at industries and companies trying to rewrite the rules of entrenched legacy industries.
Kyle Harrison:
The book was an evolving, moving-goalpost kind of project – it shouldn’t really have taken three years. It grew out of a lot of research we were already doing, investing in companies in the space, and publishing long-form pieces online on companies like OpenAI, Stripe, and Databricks. When we went deep on Anduril, we realized it had a pretty unique setup. Most companies point to some heritage – OpenAI has a rich history of ML development over many years – but very few are built almost as the rebellious stepchild of an industry, deliberately counter-positioned against it. Anduril looked at fifty, sixty, seventy-plus years of military-industrial history and said, this is broken, so we have to do this differently.
That’s why they focus on high-volume, low-cost assets – because cost overruns forced the industry to concentrate capital into a small number of exquisite, expensive assets. That’s why they spend on their own R&D – because the defense primes built their entire business on overcharging the government to reinvent the wheel every time they’re asked to build something. Understanding that history helps you appreciate what’s required.
The framework I keep coming back to is the Lindy principle. People look at an industry that’s operated the same way for a hundred years, like life insurance, and think it’s ripe for disruption. But there’s an evolutionary logic to it: if something hasn’t changed in a hundred years, there’s probably a reason – if it sucked, it would have changed, and plenty of things have changed in the last hundred years because they sucked. Usually the reason isn’t that the thing is great; it’s that it’s complicated, or it’s held hostage by the money and interests involved.
In defense, it’s a combination of both. There’s money in politics and jobs tied to congressional districts that keep incentives misaligned. But there’s also the evolving nature of warfare and conflict – what the world looked like after World War II, in the ’90s, and today are three completely distinct worlds requiring distinctly different approaches to conflict. It’s hard to move a big, lumbering machine along such a rapidly evolving landscape.
So when we underwrite companies in these spaces, we look for that same pattern. We recently invested in a company building liquid propulsion systems for missiles – solid rocket motors are a critical, structurally constrained bottleneck in the supply chain, and this company has a unique approach to solving it, built on understanding why that bottleneck exists and counter-positioning against it. That’s where the real opportunity is. People who go in saying “this is broken because people are stupid and lazy, and we’ll just do it better” usually fail.
Sam Silvershein:
Backing up – you were very early backers of Anduril. What did your LPs think when you started investing in defense tech?
Kyle Harrison:
For better or worse, Contrary started very much on entrepreneurs’ capital. Eric, the founder, was twenty-three when he started the firm, and built it as a solo GP for four years before bringing me on. In the early days, our LP base was basically entrepreneurs – people who’d built businesses themselves. Not having a heavily institutionalized LP base let us be much more belief-driven than check-box driven. It wasn’t about filling somebody’s investment-committee checkbox; it was founders wanting to invest in a firm that invests in people building great things.
It’s the same reason we were the first investor in Hallow, the large Christian prayer app – we invested in 2017, and at the time investing in a religiously focused consumer app was considered a weird thing to do. Today Hallow beats out OpenAI on consumer app downloads at certain times of year – it’s a great business.
Defense was similar. A lot of the early investors in Anduril got angry messages, or damaged relationships, over that investment. There’s a vibe shift that different categories go through at different times – if Hallow were started today, I think it would be completely non-controversial. One benefit of not having a traditional LP base is that we didn’t have LPs telling us not to invest in specific denominations, or in weapons, or whatever. Anduril also didn’t start out explicitly building weapons; it evolved into that over time.
I actually think it’s good for capital allocators to wrestle with the implications of what they’re investing in. You’re seeing a similar divide play out in AI more broadly – there’s one group of people who see an unbroken chain of globalization and believe technology competition is just meritocratic across the world, and another group who recognize that China and the CCP operate in a way that’s implicitly confrontational with the US – our companies aren’t allowed in their markets, but their companies are allowed in ours. That’s an adversarial relationship even without open conflict. When you see something like China blocking Meta’s acquisition of Manus, or Meta itself being effectively locked out of the Chinese market, that’s the kind of thing that creates real problems. As capital allocators, you have to decide whether you’re going to invest in companies with implicit ties to the CCP. I think it’s good for people to have that debate within themselves about where they’re willing to allocate capital.
Sam Silvershein:
Eventually, as valuations keep climbing and these companies scale, we’re going to have to figure out how to access the Chinese market if we’re going to justify multi-trillion-dollar valuations for AI businesses. It’ll be interesting to watch how that evolves.
I want to go back to Anduril and Palmer for a moment. You’ve talked before about how defense tech was deeply unpopular when Palmer started Anduril – he was ghosted by engineers, investors, and people he thought were friends. Fast-forward about ten years, and one might argue we’re in a defense bubble. As someone who’s been in this for a while, what’s your take on the shift in VC psychology away from high-margin software businesses toward more capital-intensive businesses, often in single-buyer markets?
Kyle Harrison:
Palmer jokes that 2022 and 2023 were his “I told you so” tour. After the Russian invasion of Ukraine, there was this acknowledgment of something we talk about in the book: there was a bestselling book written in the early 1900s arguing that large-scale conflict was basically inconceivable – that Europe was foolish to keep building up arms in such an interconnected world. Five years later, World War I started. It’s a perfect example of a recurring mentality: we convince ourselves that humanity has graduated past the willingness to use violence to achieve its ends, and we keep getting proven wrong. That happened again in 2022 and 2023 – Ukraine, Israel and Palestine, conflicts that shook a lot of people’s worldviews.
On the defense tech bubble specifically, I think companies built on a momentum-first mentality – loud, massive buildouts, raising as much capital as possible, sprinting as fast as possible – are, to some extent, doomed to fail, because they’re often chasing a pretend goalpost. I think about a company like Tesla: in the early 2000s there was a green-tech bubble, when John Doerr famously pivoted a huge amount of Kleiner Perkins capital toward sustainability and clean energy. Most of those bets didn’t pan out, but Tesla was able to prevail through that bubble. Amazon is a similar story – a quintessential dot-com company. Bezos was a Wall Street guy who looked at the internet growing fast and thought, what can I ship easily? Books. That’s a textbook example of a company that could have gotten washed out in the dot-com crash but survived it.
Some of that is survivorship bias – I’m sure there were plenty of companies built on good first principles that just weren’t able to weather the storm, bubble or not. But a good chunk of it comes down to the fact that companies built around a specific worldview, with their product and approach structured around that worldview, weather storms more effectively. There’s a great line from Palmer about the ChatGPT moment – people asked him whether ChatGPT was a huge deal for Anduril’s business, and he said their products don’t use ChatGPT. In the early days of the ChatGPT fervor, it benefited Anduril’s business not because they were using GPT-3.5 in products like Lattice, but because people who’d never paid attention to the cutting edge of AI were suddenly using ChatGPT in their everyday lives, and that woke them up to how powerful automation and AI could be – which in turn made them receptive to Anduril’s products and approach, even though the two things didn’t actually have much to do with each other at the time. Palmer’s attitude is: happy to take advantage of that tailwind while it lasts, but he also knows it will come and go.
The defense bubble will ebb and flow the same way. You’re already seeing conflicts become heavily politicized – Iran, Venezuela – and administration changes will shift attitudes toward defense buildouts. But I think Palmer, like Elon with Tesla, has this mentality that the way you weather the storm is by having a deliberately held worldview that’s true independent of which way the wind is blowing. If the wind’s at your back, take advantage of it. But if your entire ship only works with a tailwind, and you’re screwed the moment it becomes a headwind, you haven’t built a durable business.
That goes back to my core problem with capital absorption as a first principle: it’s a dangerous way to build a business, because most people don’t have a deliberate worldview of what their product should be in the long run. They have a sense of the game they should be playing – the fundraising game, the markups game, the big press-release-partnership game – but no worldview they’re actually building toward. That typically ends badly. It happened in 2021 with crypto companies – there wasn’t really a worldview behind most of them, just a moment in time.
Sam Silvershein:
They sold a nice story for the moment – Web3, crypto, all of it.
I want to end on one thing in that vein. You’ve written about the “hero generation” of VCs – those building instead of extracting, trying to fund things worth building in some of the biggest markets. What areas of the market are you most excited about today?
Kyle Harrison:
The core hype cycle we’re avoiding is what I’d call the “Neo Lab for X” rat race – anybody who leaves a major AI research lab to start their own bespoke research lab. That doesn’t make sense for us to play in, partly because we’re not deploying billions of dollars a year, and partly because we want to make eight investments a year, not forty or fifty. So we generally avoid that knife fight, with the occasional exception.
Our work has really bifurcated into two directions. One is a strong belief that a lot of existing software can be disrupted – not because AI can write the code more quickly, but because there are advantaged ways to build software around an ancillary component that’s very difficult to replicate. We have a company called Carta Health, a digital health company providing in-home therapy for things like cardiac and pulmonary rehab. It’s AI end-to-end using Claude in a lot of ways, but most people wouldn’t describe it as an AI-native company – it’s not that they have AI agents replacing therapists. The company is tripling toward almost a hundred million dollars in revenue, growing profitably, because building a network of therapists and deep relationships with insurers and distribution partners is genuinely difficult. Once you’ve built that, software becomes an advantaged distribution mechanism for your product and your network. We also have a company doing privacy-first productivity – rebuilding messaging, email, and docs from scratch with a Signal-like protocol around privacy, selling into highly regulated industries like defense, healthcare, and financial services. The benefit is that it becomes a foundational layer that unlocks sovereign AI in the long run. Those companies aren’t mind-bendingly complicated on paper – they sound like software companies – but they’re able to build something very specific and hard to copy.
The second bucket is that we genuinely believe the physical world hasn’t had its ChatGPT moment yet, despite what people say about robotics and humanoids. The physical world is still massively complicated and difficult to traverse, so companies building very specific, capable engines around physical-world processes are advantaged in our view. One of our companies, American Housing Corporation, builds in-factory housing units – not cost-optimized, cheap modular housing, but very high-quality, well-designed building projects that are heavily automated to dramatically reduce the cost of construction. That’s a difficult, messy environment to get into, and they’re tackling it head-on. We’ve essentially split at the labs level into software and hardware layers, both asking the same question: how do you get this thing deployed into a messy, complicated world?
Sam Silvershein:
And how do you build moats around that? I imagine a lot of those founders are coming through your existing network – people you’ve known since undergrad who’ve scaled within the Contrary ecosystem and spun out.
Kyle Harrison:
Yeah, exactly.
Sam Silvershein:
That’s awesome. Kyle, I really appreciate your time today. Thank you so much for joining the Driving Alpha podcast.
Kyle Harrison:
Thanks for having me. This was great.
