What this covers
Patrick interviews Michael Seibel, who helps run selection at Y Combinator, about what reading tens of thousands of applications has taught him. The scale is concrete: a little under 30,000 applications across the Winter and Summer 2020 batches, about 10 percent interviewed (partly over video because of COVID), and roughly 475 companies accepted. Selection runs on a few rules: the Don't Be Too Smart Heuristic, Trajectory Over Starting Position, and a strong yes policy adopted after companies admitted on three weak yeses underperformed. He also argues that investors cannot reliably rank the bottom of a batch.
The middle of the conversation covers specific business terrain. On the two-sided marketplace, he recommends hand-recruiting suppliers and says the real customer is usually the side that needs the service most often, such as Airbnb hosts. He argues Airbnb won by building better tools, not by having a novel idea. Several claims are contested: that software is not commoditized, and that Silicon Valley's funding system favors software, leaving Tesla and SpaceX largely government-funded. The Frequency Intensity Willingness to Pay heuristic gets a walkthrough. The Promise story, in gov-tech, shows a company discovering that its customers' incentives ran against its mission.
The last stretch turns to founders. Seibel describes the self-deception behind product risk and execution risk, the co-founder relationship as a breaking point, and the need to sit several standard deviations from average. He credits the peer batch, not YC's advice, as the program's main asset. He closes on why early customer traction gives founders leverage over investors.
Seibel argues that successful startup selection and building depend on judging trajectory, customer pain, and founder execution rather than cleverness, and that founders gain leverage by validating demand with real customers before approaching investors.
- Early-stage evaluation should favor demonstrated forward motion and a deep understanding of the problem over pedigree or idea quality
- Two-sided marketplaces are often best understood by identifying who needs the product most frequently, which is frequently the supplier
- Founders with customer traction have more leverage with investors than founders with only a pitch
Software is leveraged technology, and companies increasingly encode everything into software.
- Patrick reports that Peter Thiel defined technology as doing more with less, and that software is literally leveraged, which connects to Seibel's emphasis on founder leverage.
“I think that's Peter Thiel's definition of technology. Like, just do more with less.”
- Patrick cites Jay Kreps of Confluent for the observation that companies that are essentially software or technology companies are increasingly encoding everything into software.
“I think it was Jay Kreps at Confluent that wrote that great post about how every company”
Andy Rachleff's dogs-eating-dog-food idea is an appealing basic test of product-market fit.
- Andy Rachleff's idea that the dogs are eating the dog food is a basic way to describe product-market fit, which Patrick finds appealing as a test for builders.
“I've always just liked Andy Rachleff's basic idea that the dogs are eating the dog food.”