Episode Summary
Executive Summary: Elad Gil argues AI is still early and underhyped, with transformer-based systems creating a new curve centered on units of cognition and labor rather than traditional machine learning. He compares AI's trajectory to stacked S-curves, discusses open vs. closed models, the AI/crypto overlap, and why tech has become more political amid regulatory conflict. He sees huge opportunity, but warns that winners will depend on application, defensibility, and market structure.
Main Topics: AI as a new technology curve (Priority: 5/5): Gil explains that transformers created a fundamentally different AI trajectory than earlier machine learning, with ChatGPT marking the moment mainstream awareness caught up. He frames current AI as generative systems capable of manipulating language, code, and knowledge. AI value creation as units of cognition and labor (Priority: 5/5): He argues the end product of AI systems is not software per se but purchasable units of cognition, labor, or eventually robot time, illustrated by customer-support and coding automation examples. Where AI is on the S-curve (Priority: 5/5): Gil says AI is still early in both technology and adoption S-curves. He emphasizes continued runway in training, post-training, and inference, plus new markets unlocked by better models. Open source vs. closed source market structure (Priority: 4/5): He believes both open and closed models will coexist. Open source lowers access and broadens adoption, while closed systems may retain major commercial advantages, especially at the infrastructure and hyperscaler level. AI and crypto intersection (Priority: 4/5): Gil sees overlap in identity, payments, censorship resistance, and agentic systems, but is skeptical of some decentralized compute claims. He thinks the most interesting crypto-AI frontier may be AI agents using blockchain-native economic games and wallets. Crypto’s maturation and role in tech investing (Priority: 4/5): He frames crypto as a durable technology with clear use cases like store of value, DeFi, and cross-border portability, but says many imagined applications failed to materialize. He advises thinking in terms of capabilities and use-case fit. Politics, regulation, and tech (Priority: 4/5): Gil argues tech became more political because regulation, enforcement, and ideological conflicts increasingly affected innovation in crypto and AI. He supports reform that removes overreach while preserving essential protections.
Key Arguments: AI is underhyped despite intense attention because it sells units of cognition and labor, which can directly replace or augment expensive human work. The current AI wave is fundamentally different from older ML because transformer architectures enable language, code, and knowledge generation at scale. There are multiple stacked S-curves in AI: model capability, inference compute, training/data, and adoption; all still have room to run. The biggest near-term AI value is in digital labor; robotics and humanoids are a later, separate curve tied to physical-world labor. Open source and closed source will both matter, but the commercial dynamics differ from crypto because AI open source lacks token-based monetization. A lot of decentralized AI rhetoric overstates decentralization; the key question is openness, access, and data/model availability. The AI/crypto overlap is most promising in identity, payments, censorship resistance, and agentic systems with wallets, not in naive decentralized compute. Crypto’s strongest use cases are those that uniquely need permissionless, censorship-resistant, always-on value transfer and ownership. Tech became political because governments increasingly targeted innovation through regulation, enforcement, censorship, and antitrust, pushing founders and investors to engage. For investors, the best approach is to buy “index” exposures to major tech waves and favor businesses that are durable even as AI changes the world.
Data Points: Unicorns backed: 40 - Gil is described as having backed about 40 unicorns, including Airbnb, Coinbase, Figma, and Stripe. Customer support workforce reduction at Klarna: 700 people - Used as an example of AI replacing support labor with an OpenAI-based system. Repeat query reduction at Klarna: 25% - Gil cites the AI system reducing repeat customer queries. Microsoft quarterly revenue: $28 billion - Example of large-cloud revenue where AI contributed meaningfully to quarterly growth. AI contribution to Microsoft quarter: 15% - Gil says Microsoft publicly attributed about 15% of quarterly lift to AI. AI-related quarterly lift at Microsoft: $3.5–4 billion - Derived from the cited 15% of a $28B quarter. SaaS and enterprise software market size: $500 billion/year - Used to compare against the potential AI market for white-collar software/services. White-collar payroll addressable by AI: $3.5–5 trillion - Gil estimates employee-salary spending in AI-exposed sectors. Potential conversion of payroll to SaaS revenue: 10% - He argues converting 10% of labor cost into SaaS revenue would recreate the scale of enterprise software. Open source model user threshold: 700 million users - Gil references Meta’s Llama licensing trigger for very large users/hyperscalers. Prime Intellect decentralized training claim: 400 million to 10 billion parameters - Referenced as an example of decentralized AI training making a big scale jump. Bitcoin price milestone: $100,000 - Mentioned as a sign crypto is moving deeper into its adoption curve.
Pivotal Quotes: "AI is still underhyped despite being incredibly hyped" — Elad Gil: His core thesis on the state of AI investment and public perception. "You’re selling pieces of thought or ability to do things" — Elad Gil: He explains the real economic product of modern AI systems. "I do think that to some extent, one could argue that there will have been three ages of humanity" — Elad Gil: He frames human history as an era of human compute, a hybrid human-machine era, and a future machine-intelligence era.
Implications: Listeners should expect AI adoption to keep accelerating, with major value accruing to infrastructure, apps, and durable index-like businesses. Crypto’s best future may be as a base layer for AI agents, identity, and payments, while politics and regulation will remain central to both sectors.