E429: Dr. V on AI, Market Bubbles & Finding the Next Anthropic

How I Invest with David Weisburd

In this episode of 'How I Invest with David Weisburd,' Dr. V (Vaibhav Agrawal), former Lightspeed partner and founder of ODDBIRD VC, challenges the traditional venture capital

Key takeaways

• The venture capital industry has shifted from a reputation-based model to one driven by media, distribution, and continuous liquidity management.
• AI is already disrupting labor in predictable ways—especially coding—due to low cost of failure and deterministic outcomes, but medicine remains resistant due to high stakes and complexity.

Main topics

• The evolution of venture capital from reputation to media-driven sourcing
• AI's impact on labor markets and coding productivity

Notable quotes

"The market always seems to be overvaluing technology. And 10 years later, you realize it wasn't." – Dr. V on venture capital cycles

"If you look at the impact that cursor and claw are having on coding... AI kind of eat into a whole bunch of that." – On AI's early disruption in engineering labor

Conclusion

Dr. V argues that the future of venture capital lies in proactive liquidity planning, leveraging AI for trustworthy

Transcript preview

Speaker 1 (0:00) Dr. V, you spent nearly a decade at Lightspeed before deciding to spin out. And before we started recording, you said that the market right now is undervalued in venture. Why is that? This Speaker 2 (0:10) is something I learned from Chris Chappie, who was one of the founders of Lightspeed. He said the market always seems to be overvaluing technology. And 10 years later, you realize it wasn't. I've kind of like learned to tune these things out. You look at the data, right? I mean, if you look at, let's say, 2008, 2009, when I was starting my first company, in the aftermath of the global financial crisis, the total market cap of private technology companies at that time was maybe close to $50 to $80 billion. And then the largest company was Facebook, $15 to $20 billion in market cap. And you fast forward today, we're talking about multi-trillion dollar companies. We're talking about a private market that's an aggregate of about $5 trillion. If somebody had said we would be sitting at a $5 trillion private market cap even three years or four years ago, it might have seemed ludicrous, right? And here we are. And mind you, we've only seen AI disrupt coding in the world in a realistic way. Most of the revenue of anthropic and open AIs are coming from coding. So whether it's medicine, engineering, physical automation. When I look at the degrees of... possibility, it just feels very, very large. And Speaker 1 (1:13) the very rough back of the envelope math on that is market cap of the private companies versus venture capital going after those private companies. Speaker 2 (1:22) For perspective, if back in 2008, 2009, the top funds in Silicon Valley were probably raising like $500 to $1 billion funds. And even that was like, wow, that's crazy. There was no soft bank. $100 billion Vision Fund. There are no mega hedge funds. And I look at even just Lightspeed, Andreessen, Sequoia, now increasingly interestingly Benchmark and many others, you're already in excess of 50 to 70 billion. The markets have completely reset in terms of the capital scale and expected value. And Speaker 1 (1:56) the big unknown is whether AI will actually disrupt labor. I Speaker 2 (2:01) don't think it's an unknown. If you look at the impact that cursor and claw are having on coding, which is inherently engineering labor from like writing sophisticated code and research level code to implementing basic systems, like systems implementation and stuff, very different levels of sophistication. You're seeing AI kind of eat into a whole bunch of that. For me, the question is less whether it can disrupt labor. The question is what kinds of labor get disrupted first. My point of view is that categories of labor where There is low cost to going wrong, i.e. there's a little bit of course correct available. You don't require a lot of trust. And then second, ones where there's a deterministic right answer and you can actually close the feedback loop, improve the model on a continual basis, which is why coding is so perfect, right? Because when a code is written, it's expected to perform in a certain way. It either works or it does not. Along those two axes, you're going to see things evolve. And if one end is coding, probably the other extreme end is medicine. Cost to going wrong is... Very high because of multisystemic nature of biology. It's like very hard to attribute a response or unintended response of something solely to the algorithm. Speaker 1 (3:13) You've said that the old venture capital playbook no longer works. Why is that? Speaker 2 (3:17) The venture playbook has changed. Ten years ago, venture was not a media business. It was a reputation business where investors would invest it over decades and they were known for investing in certain companies. and therefore became aspirational founders to raise capital from. Fast forward today, right? Sourcing, especially at the early stage, is very much a media business. You've seen the advent of like TBPNs getting acquired and recent horror talks about the new media. Speaker 1 (3:44) Harry Stubbings has deployed over a billion dollars reportedly. There Speaker 2 (3:48) you go. And so there was kind of like this moment where a lot of podcasters were raising funds to varying degrees of success. It is now widely established that distribution is important. for a venture firm to be able to see the best opportunities. And then all the way at the end, when you're exiting, right? 10 years ago, investors were not really thinking about exits that proactively. The conventional assumption was you're gonna have 10 investments, one or two of them are gonna hit it out of the park, eight are gonna like probably go sideways, which don't need any management. The other one or two that hit it out of the park are probably gonna go public. So it was kind of like a low effort, more laid back style of planning exits. Fast forward today. We're in a very difficult public environment. And so as an investor, you have to kind of build this muscle of continuously asking yourself whether you should be telling a company, buying more in a company, consequently have an ecosystem of relationships of buyers or sellers that you could trade with. That was not necessary many years ago. Companies are staying private much longer, but the average tenure of an investor is reducing. The average time an investor stays at a big firm is reducing. Speaker 1 (4:51) Why is that? Speaker 2 (4:52) I think it's just like maturity of the industry. It's become more competitive. I would say- Are Speaker 1 (4:57) they getting poached from other firms? Speaker 2 (4:58) Absolutely, all the time. I used to joke, there are three types of investors, right? There's ones that don't perform. You're going to hopefully find them over time and transition them out. Ones that are high performers, but then you're at the risk of losing because they want to go start their own thing or somebody poaches them. And then what you're left with is the stable middle, right? And that is the dilemma of a large firm. What's happening more structurally is that- several parts of what an investor used to do are getting institutionalized. They no longer sit with the investor. For example, marketing. All of these funds now have big marketing teams and investors are collaborating with them. That was not the case when Bill Gurley was writing his blogs or Fred Wilson started writing his points of view. Think about exits again. Lightspeed built a capital markets team to create that discipline at that scale. And so now if an investor needed to come into a Monday meeting and talk about their plans for exiting a position, That conversation is now happening. Different set of people on a much more regular cadence. Speaker 1 (5:53) And they started this capital markets team while you were at Lightspeed. They were just starting when I was at Lightspeed. What does that capital markets team do exactly? Speaker 2 (6:01) Number one, list of companies that a venture fund owns and segment them, discipline into what you want to buy or hold, and then what do you want to sell. That itself in a big firm where you have 30, 40, 50 investors. And by the way, you have thousands of companies even getting to that level of clarity takes a lot of work. Speaker 2 (6:23) The second is developing relationships in the ecosystem that actually allow for these exits to happen. For example, big funds are now using a variety of tools to actually exit companies. They're doing continuation funds. They're doing GP secondaries, LP secondaries, encouraging M &A. Firms like Lightspeed are also doing buyouts now. So they kind of like could be a seller and a buyer, right? Arguably. And so that ecosystem, as it continues to develop, requires a continuous... Speaker 2 (6:54) relationship building and awareness of how people's strategies are changing. Speaker 1 (6:57) Saying that you want to manage liquidity and you want to provide liquidity to LPs sounds extremely obvious and sounds like everybody should be for that. But the question then becomes, who is buying these assets? Speaker 2 (7:09) Everyone I talked to on the show is chasing the same thing, an Speaker 1 (7:12) edge. And more and more, the edge comes down to your information, not just having it, but being able to trust it when the stakes are highest. AI is doing more of the information gathering for you every day, and most tools are very good at sounding right. The summary reads clean, but can you trace it back to the filing, the transcript, the specific passage that drove the answer? Or are you just trusting the confidence of the output? For investors, that's not a minor concern. A missed filing, a missed weight of source, a context that got lost somewhere in the retrieval chain, those aren't edge cases, they're how decisions go wrong. AlphaSense