![Sarah Guo - What the 250 People Building AI Believe - [Invest Like the Best, EP.489]](https://megaphone.imgix.net/podcasts/ef669774-cccd-11ed-889b-c36caad6646f/image/158efdddfb983d2678b3530d484e8aa2.jpg?ixlib=rails-4.3.1&max-w=3000&max-h=3000&fit=crop&auto=format,compress)
Sarah Guo's firm, Conviction, centers its strategy on identifying and supporting the roughly 250 key individuals—a mix of entrepreneurs and researchers—who are actively pushing the boundaries of AI. This hyper-focused, network-centric approach is core to their investment thesis for finding frontier companies.
Many leading AI researchers now believe progress is primarily driven by massive compute resources, not individual breakthroughs. This leads to a sense of disempowerment, as they feel their personal work is less impactful on the overall outcome when competing with billion-dollar training runs.
Instead of a traditional customer-first approach, Conviction successfully invested in companies like Harvey by first identifying workflows technically well-suited for AI models' capabilities. For law, this was its nature as "structured language," a technology-forward, first-principles approach to finding product-market fit.
A significant belief shift has occurred among top AI researchers in the last 12 months. Many now feel that recursive self-improvement in models means exponential intelligence is just a couple of years away, a notable acceleration from previous, longer timelines.
The primary obstacle to meeting AI's future compute demand is not a failure of technology or capital markets. Instead, it's a regulatory and public alignment problem that slows the construction of necessary infrastructure like data centers and nuclear power plants.
A major risk in the current AI landscape is investors proxying judgment to founder pedigree. Making large bets without a fundamental, intuitive grasp of the technology or business model is a dangerous strategy that will likely lead to poor outcomes.
The founders of Sunday Robotics are making strides by treating data acquisition as a core technical problem. Their innovation lies in developing cheap, practical methods to collect data that mirrors real-world environments, a key and often overlooked challenge in robotics.
When raising her first fund, Sarah Guo deliberately avoided a polished, detailed strategy presentation. She pitched LPs on her background, team, and a commitment to "execute like hell and figure it out," believing a strategy is only valid after contact with the market.
Previously, it was believed the only path to value in biotech was developing a drug. Now, AI models are proving so effective at accelerating R&D that traditional software and platform business models are becoming highly valuable and viable for serving large pharma customers.
Given the high cost of labor and specific skills gaps, reshoring US manufacturing is not feasible without massive investment in automation and AI. This positions AI not just as a technology trend but as a critical component of national economic and security strategy.
Effective investing is less about being contrarian for its own sake and more about identifying a truth the market has mispriced. The skill lies in acquiring better, asymmetric information and then confidently maintaining that position until the market catches up.
Sarah Guo argues investors waste energy debating grand strategic frameworks like which layer of the AI stack will capture the most value. A more productive focus is on the "next 99% of diffusion"—the countless specific ways AI will be adopted across the economy.