Nov 2, 2024 · 50m · american-optimist
Secrets Of Investing Early – Lessons From A Billion-Dollar Investor · Joe Lonsdale
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this 100th episode of American Optimist, host Joe Lonsdale interviews prolific Silicon Valley investor and entrepreneur Elad Gil about his career journey, early-stage investment strategies, startup scaling, and the economic and societal impacts of artificial intelligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Joe holds 31.6% of the talking time here. How this is scored →
Joe as informed peer 5.5 Guest teaching 3.3 Guest disagreement 1.2 Joe pushing back 1.9
speaking balance: gold is Joe, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 30:27 Contrarian claim that AI is dramatically under-hyped
Elad directly rejects the consensus media and investor framing of an AI bubble, arguing that current massive revenue gains are occurring despite adoption still being in its infancy.
Hardest push from Joe ▶ 36:36 Host challenges continuous exponential LLM scalingJoe explicitly refuses Elad's assumption of continuous exponential capability gains, challenging him on whether current transformer architectures are destined to hit a local maximum.
Biggest teaching moment ▶ 34:52 Dissecting transformers versus 20 years of legacy MLElad educates the audience and frames the technical distinction between previous statistical association engines (CNNs, RNNs) and current generative transformer architectures.
Joe holds their own ▶ 34:00 Host details precise addressable numbers for services automationJoe demonstrates his domain mastery by breaking down exact figures ($2.1 trillion across logistics, healthcare, and auditing) and explaining how human-machine symbiosis is already doubling productivity.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Joe as informed peer | Guest teaching | Guest disagreement | Joe pushing back | Why |
|---|---|---|---|---|---|---|
| Elad Gil's Background and the Dot-Com Bust | 3 | 2 | 1 | 1 | Joe asks about Elad's origin story in Silicon Valley right during the dot-com bust. Elad details living off cheese sandwiches and doing startup cash-burn math to anticipate layoffs. The exchange is warm, conversational, and biographical. | |
| Building Teams and Culture in Google's Golden Era | 4 | 3 | 1 | 1 | Elad explains Google's golden era when Larry Page eliminated middle managers, creating a free market for talent to build mobile. Joe contributes observations comparing Google's cash gusher to modern high-growth companies like Ramp. | |
| Twitter's Early Days and Product-Market Fit | 5 | 3 | 1 | 2 | Elad describes Mixer Labs' acquisition by Twitter and fixing basic infrastructure amid the 'fail whale' era. Joe shares hiring experiences at Palantir and probes Twitter's cultural transition from early free-speech principles. | |
| Founding Color, Health Tech, and Longevity Innovations | 6 | 4 | 1 | 1 | Joe shares a personal story of how Color's genetic testing caught his mother-in-law's stage-three cancer early. Elad contextualizes systemic healthcare incentives and the surprising limitation of curing single major diseases on overall life expectancy. | |
| Founder Archetypes, Business Defensibility, and Talent Clusters | 7 | 3 | 1 | 2 | Elad categorizes three super-founder archetypes and discusses defensibility through scale and network effects. Joe actively enriches the conversation with concrete examples from Palantir's international deployments and Anduril's rapid contracting speed. | |
| High Growth Handbook and Regulatory Challenges | 6 | 3 | 1 | 3 | Joe asks about Elad's interview with Lina Khan, criticizing FTC aggressiveness in biotech M&A and comparing NIH funding sclerosis to the youth-focused Damon Runyon foundation. Elad explains his curiosity regarding the broader societal decline in empowered young leaders. | |
| Current AI Investment Landscape and Cognitive Units | 4 | 4 | 2 | 2 | Elad counters the prevailing narrative of an AI bubble by arguing AI is dramatically under-hyped given massive early enterprise revenue and workforce transformations like Klarna. He introduces the framework of AI as 'units of cognition.' | |
| Quantifying the Transformation of the Services Industry | 7 | 3 | 1 | 2 | Both investors quantify the addressable services market, with Joe presenting his estimate of $2.1 trillion in US headcount that AI can address via human-machine symbiosis. Elad highlights how converting even 10% of services payroll into SaaS would duplicate total enterprise market cap. | |
| The Technical Shift: Transformers versus Legacy Machine Learning | 6 | 5 | 2 | 5 | Elad distinguishes the transformer paradigm from legacy statistical machine learning, citing vertical breakthroughs like Harvey. Joe explicitly challenges whether LLMs will continue on an exponential trajectory or plateau at a local maximum. | |
| Evaluating the Five Layers of the AI Value Stack | 7 | 3 | 1 | 2 | Joe outlines his five-layer AI value stack model and vertical buyout strategy, which Elad validates and mirrors with his own buyout plays. They discuss long-term durable assets and customized 1-on-1 AI tutoring for future generations. | |
| Reviving Civic Ambition: Constructing Inspiring Public Monuments | 6 | 2 | 1 | 1 | Elad advocates for constructing inspiring large-scale civic monuments as historic indicators of civilizational vitality. Joe connects this to his own philanthropic involvement with neoclassical monument foundations. | |
| AI for Global Good: Healthcare Access and the Modern Library of Alexandria | 5 | 4 | 1 | 1 | Elad highlights positive global applications of AI, from Med-PaLM surpassing human physicians in triage to his project creating an interactive, multilingual 'Library of Alexandria' for 1,000 classic texts. Joe strongly endorses this optimistic outlook. |
Statements from this episode (26)
Gil: Larry Page eliminated three management layers at early Google
“Larry Page was still very active at Google, and he decided to get rid of all middle managers. And so when I joined, he removed like three management layers, and suddenly every director or VP level person had 50 to a hundred people working directly for them.”
Gil: Recruited Google's early mobile team secretly without their managers knowing
“I recruited a lot of the early mobile team without their managers knowing because their managers didn't know what they were doing. So they'd just come and start working with me on mobile.”
Gil: Google used to maintain a webpage of things it would never do
“Google used to have this web page up that was like, here's all the things we're never going to do. And I said, we're never going to do IM or chat. We're never going to do a browser. We're never going to end up doing all those things, right? We're never going t…”
Gil: Golden age tech companies concentrate talent that later fuels Silicon Valley
“And I feel like at any given moment in time, there's one or two companies that are going through this kind of golden age period where you're expanding rapidly. You're soaking in all the best talent. You're launching multiple product lines. You have this, Diver…”
Gil: Twitter couldn't deploy code for weeks during MixerLabs acquisition
“At the same time when we were bought they couldn't what's known as deploy code, they couldn't update the website. And they hadn't been updating it for weeks. They just didn't know how to update the code.”
Gil: Consumers resist paying out of pocket for preventative healthcare
“Honestly, people just don't want to pay out of pocket for their health. So to your point, you could pay a 102 hundred dollars and get information that may be life-saving. People don't want to do that, right? They're used to their employer paying for it or the …”
Gil: Curing all cancer and heart disease would only add 5-7 years
“If you got rid of all of cancer and all of heart disease, I think you add something like five to seven years”
Lonsdale: Anduril integrated with Saronic in three days versus legacy primes
“So we have this company, Sironic, I think you're also just invested in them, that we helped, I helped build, and it's growing really fast for the Navy, and the people, they asked us to weaponize some of the autonomous vessels, and they, and so we reached out t…”
Gil: Stanford, MIT, and Harvard founders created most recent startup value
“Somebody on my team actually went back and asked what are all, There are common schools that all the founders of the biggest companies went to, or the last decade, for example, right? And there's basically three schools that are like a huge chunk of the overal…”
Lonsdale: FTC Chair Lina Khan's aggressive enforcement is preventing biotech cures
“In my experience, there's been things in biotech that are being really, like, hurt and set back, and things are not being cured because she's being so aggressive and blocking things.”
Lonsdale: Damon Runyon foundation is 100x more effective per dollar than NIH
“By all accounts, it's like a hundred times more effective per dollar than what the NIH does in these areas.”
Gil: Fewer young breakout startup founders exist today compared to a decade ago
“Generationally, you go back 10 years, and it was Patrick, and it was Zuck, and it was Dylan from Figma, and you can kind of come up with a list. Today, there's many fewer, and I'm starting to see some of them in AI emerge again, right?”
Lonsdale: A $500k tech job at 21 would have prevented him founding Palantir
“It would have been tough if someone offered me half a million dollars a year when I was 21. I probably wouldn't have gone to start Palantir, I have to be honest.”
Gil: AI is dramatically underhyped despite current massive investment levels
“AI is in my opinion, dramatically under hyped right now. And I think part of that is number one, we're seeing massive actual revenue and impact without that much adoption, right?”
Gil: The ultimate end product of AI is rentable units of cognition
“I think what people are really misunderstanding is what is going to be the end product. And I think the end product is units of cognition. Right. You're going to be effectively paying for or renting something that's going to think or do things on your behalf.”
Gil: Machine intelligence will eventually dominate total brain equivalents
“There'd be like three eras of mankind in terms of, or three eras of intelligence, right? And the first era is basically people. And the second era is some hybrid era where it's a you know, if you look at the number of intelligence units or whatever you want to…”
Gil: Capturing 10% of services headcount recreates the entire enterprise software market
“If you say take a 10% cut, And transform employee headcount costs into SaaS software, you're back to half a trillion, and you've recreated the entire enterprise software market cap.”
Lonsdale: AI can already double productivity across $2.1T in US services
“I think in the 2.1 trillion that we think could be addressed right now, stuff like healthcare billing, logistics billing, auditing, whatever, legal, some legal stuff, we think already you can at least double the productivity, which would be kind of a trillion …”
Gil: LLMs will eventually advance to perform all human services
“You have these steps, and so it's this ladder that you're climbing, and eventually you're going to hit a point where it can do all of human services, right? And I don't know if that's GPT seven equivalent or eight or six or 10 or whatever, but we're going to g…”
Gil: Owning vertical workflows insulates AI apps from model commoditization
“The things that I'm bullish on are where you own a workflow for a vertical. And so everybody in that vertical is using that workflow. And so if you upgrade the model, it just makes that workflow better. And so it doesn't matter what the model is.”
Gil: Was totally wrong investing in two Nvidia competitors eight years ago
“The bottom side of it, the chip side actually invested in two companies seven, eight years ago thinking that there would be Nvidia competitors and I was totally wrong.”
Gil: Backing AI-driven buyouts to reshape cost structures and roll up verticals
“The other thing I've been doing is actually been backing. I've now backed two AI driven buyouts where the idea is you buy the asset itself and you can radically change the cost structure or increase the leverage of our organization by putting in AI in a deep w…”
Gil: Religious upbringing protects kids from adopting fake modern substitute religions
“Also I feel that some form of religious upbringing is useful if only to protect them from fake religions sort of coming in or modern religions.”
Gil: Peak civilizations always build monuments to inspire future generations
“I think if you look at every society at its peak, or at least on its way up, you'd always have these large scale Inspiring monuments. You celebrate wins, but you also inspire the future. You inspire the next generation of people.”
Gil: Modern society has lost its aspiration to create beautiful artifacts
“I mean, we've kind of lost some form of beauty or sense of beauty in society, right? We don't aspire to make giant beautiful artifacts.”
Gil: Google's Med-PaLM 2 outperformed human physician experts
“Google released a model, I think two years ago called MedPalm too, where if you compare the answers from that model relative to physician experts, it outperformed them. In other words, if you use physicians to train the model will get worse.”