Patrick Collison: "What If You Succeed?"

Patrick Collison, CEO of Stripe, discusses his entrepreneurial journey and philosophy in conversation with a YC partner. He reflects on dropping out of college twice—first after freshman year and again to start Stripe—noting that the cost of dropping out is minimal and the decision is reversible. He emphasizes that while he felt urgency to build at the time, in hindsight that sense of urgency was somewhat unnecessary.

On the topic of AI and learning, Collison uses the metaphor of 'cognitive L1 cache' to argue that deep knowledge remains valuable even with powerful AI tools available. He contends that keeping information in one's own memory is faster than repeatedly querying AI systems, and that companies still place enormous premiums on cognitive ability. He personally refuses to use AI-generated writing suggestions, believing that writing and interpersonal communication remain fundamentally important skills.

Collison details Stripe's founding, explaining that the idea emerged during a walk back from sushi after attending Startup School in 2009. He and co-founder John Collison identified a concrete problem—the difficulty of handling online payments—that was viscerally felt by potential customers. Notably, Stripe waited nearly two years before public launch (September 2011), departing from YC's typical 'launch early' philosophy. However, this approach worked because they had production customers from very early on, starting with Ross Buché at 28 North in January 2010, providing real-world feedback that grounded their development.

On the risk that big tech companies will monopolize AI opportunities, Collison argues the track record suggests otherwise. He points to Google's history as a cautionary tale—despite immense resources, Google hasn't successfully executed across all possible domains. Organizational complexity and competing priorities limit even well-resourced companies.

Most significantly, Collison presents Stripe's data showing that more businesses are being started now than ever before, with the year-over-year growth rate at roughly 2x (the largest relative jump ever recorded). Importantly, these aren't just more businesses—median business revenue is higher, and probability of reaching revenue thresholds ($1M, $5M, $10M) is improving. This suggests a more decentralized outcome where many thousands of companies succeed, rather than hegemonic centralization by a few large AI companies.