Patrick Collison on AI, Dropouts, and Stripe's Early Lessons

Patrick Collison, co-founder of Stripe, shared insights on entrepreneurship, the impact of AI, and the early days of Stripe at Startup School. He reflected on his journey, the decision to drop out of college, and the current landscape for startups.

The Value of Foundational Knowledge in an AI World

Collison discussed the balance between leveraging AI tools and retaining core knowledge. He likened cognitive knowledge to an L1 cache, emphasizing that direct recall is significantly faster than querying an AI agent. While AI models are capable of monumental feats, he noted their current deficiencies in nuanced, interpersonal communication and writing. He personally still writes all his own communications, finding the output of AI models to be lacking in compelling quality, possibly due to the difficulty in defining a suitable utility function for reinforcement learning in that domain.

He highlighted that companies like Stripe and leading AI labs still place an enormous premium on cognitive ability, suggesting that renouncing the pursuit of deep understanding would be premature.

Dropping Out of College: A Personal Experience and Advice

Collison has the unusual distinction of having dropped out of college twice to start companies. His first venture was with his co-speaker, Harge, and the second was Stripe. He emphasized that dropping out is not a "trapdoor" decision, as one can always return.

He initially envisioned an academic career in physics but discovered the world of startups, which was less known in Ireland and even on campus during his college years. While he felt an "unnecessary sense of urgency" to start a company, he advises that if one enjoys college, there's no harm in finishing. However, if college isn't captivating, the perceived risks of dropping out are often exaggerated, as "nobody has ever cared" about his drop-out status in his professional life.

His urgency stemmed from a desire to "speedrun" life and a belief that startup opportunities were ephemeral. In hindsight, he realized this was a "poor intuition," as Silicon Valley has consistently offered a "surfeit of opportunities" over many decades.

Debunking the "Permanent Underclass" Fear

Addressing the common student fear of being left behind if they don't immediately drop out to start a company and make money, Collison dismissed this as a "millenarian" view. He compared it to past societal transformations, like the advent of aviation, where people believed everything would change fundamentally. While aviation was significant, it didn't lead to the complete "sociological rewriting" some proponents imagined. He advised against believing that the current moment is the "last couple of years" to create a company.

The Genesis of Stripe: An "Obviously Good, But Also Bad Idea"

Stripe's initial concept—combining the internet and money—seemed like an "obviously good idea" because existing payment methods were universally disliked, antiquated, and cumbersome. However, it also seemed like a "bad idea" because two young founders attempting to start a financial services business without prior experience was seen as improbable. "Fintech" as a sector didn't even exist as a term at the time.

Collison described feeling like "squirrels in a trench coat" trying to masquerade as a real business. Many potential partners and banks were skeptical, almost "looking for the button to call security." The fact that Stripe was grounded in a "concrete, actual, real user problem" ultimately saved them.

He recounted the moment he and his brother John decided to start Stripe after attending Startup School in 2009, thinking, "Yeah, you know, we might as well because it probably won't be that hard." This highlights the often-underestimated complexity of building a financial services company.

Stripe's "Slow" Launch and the Power of Early Customer Feedback

Stripe took almost two years from the first lines of code to its public launch in September 2011, a timeline that would typically be frowned upon in the YC world. This extended development period was necessary due to the extensive infrastructure required for security, partnerships, money movement, and reliability.

However, Stripe had production users from very early on. Their first live customer in January 2010, Ross Buché of 28 North, provided continuous feedback. This "just-in-time development" approach, where features like dashboards, refunds, and payouts were built in response to actual user needs, was crucial. This constant stream of customer feedback and learning from reality, rather than hypothesized conceptions, allowed them to delay a public launch without being "in the wilderness."

AI's Impact on Startup Strategy: Beyond Lean Startup

Collison pondered whether the rise of AI and the ability to build software cheaply and quickly should encourage more ambitious initial product launches. He suggested that the traditional "lean startup" doctrine of finding small niches and iteratively expanding might become more competitive and harder to execute in the AI era.

Instead, he proposed that startups might need to "more aggressively decorrelate" and pursue "really divergent starting points" where competition is less fierce. He noted that many successful companies of the last decade, including AI labs and companies like Anduril, have not strictly followed the lean startup model. He believes that with AI, it's now possible to "start these much more aggressive and ambitious things up front."

The "Schlep Blindness" and Intellectual Rewards at Stripe

Collison addressed the concept of "schlep blindness"—the tendency to avoid unglamorous but necessary tasks. While setting up payroll or dealing with arcane financial services might seem unrewarding, he feels "extremely lucky" with Stripe. He emphasized the importance of considering "what if you succeed?" when starting a company, asking if one would enjoy working on it for decades.

For Collison, Stripe has been intellectually rewarding because it involves working with "the world's most interesting and innovative companies." He finds every business to be an "applied theory on how some aspect of the world works," and he has "never met a Stripe customer and thought that's boring." This makes the overall business the "opposite of the schlep blindness instinct."

AI and Decentralization: A World of Many Winners

Addressing concerns that large AI labs will centralize power and stifle smaller startups, Collison offered a more optimistic outlook. He recalled similar fears about Google's omnipotence 20 years ago, noting that even with immense resources, human organizations struggle to prosecute 100 different priorities effectively.

While AI capabilities will undoubtedly obviate some specific verticals, Stripe's data suggests a different trend. The number of new businesses starting is significantly higher than a year ago, with a nearly 2x year-over-year increase, the largest relative jump observed. Furthermore, the median business is performing better, and the probability of reaching revenue thresholds like $1 million or $5 million is increasing. The time to revenue for new companies incorporated with Atlas is also declining.

Collison concluded that based on current data, it's "a better time than ever to start a business." He believes that the "hunger and intensity" with which companies are leveraging new AI capabilities, both new and existing, points towards a "more decentralized world and one with more broad-based prosperity," with "many thousands of winners."

He also noted that enterprises are more willing to buy from startups, driven by a "real terror of being left behind with archaic and antiquated ways of operating." This makes it an opportune time for startups to sell and achieve meaningful scale quickly. Consumers, too, are "intrigued by the products" and open to experimenting with new AI-powered offerings.