What is Sutskever's List?
In late 2023, a question ricocheted online: "What did Ilya see?" The backdrop was a dramatic power struggle at OpenAI, where Ilya Sutskever had reportedly been alarmed enough to support ousting the CEO. Elon Musk's public musings on the matter fueled speculation. Amid the commotion, one intriguing clue to Sutskever's mindset resurfaced: a personal reading list of research papers he had shared with John Carmack, informally known as "Sutskever's List." Sutskever allegedly claimed it captured "90% of what matters today" in AI. The notion that a single collection of papers could hold the keys to modern AI lent the list an aura of mystery and importance.
I first learned of Sutskever's List through these whispers and online treasure hunts. Enthusiasts pieced together clues, and eventually a reconstructed version of the list appeared, quickly gaining almost a million views. It became a cultural touchstone. To ask, "Have you read Sutskever's List?" was shorthand for claiming a handle on the fundamentals of contemporary AI. And yet, this question was often more signaling, as many knew of the list without truly engaging with the papers on it. In a field racing forward with speed, the idea of a stable canon of ideas felt refreshing and necessary. I realized that exploring these works and the context in which they arose could provide invaluable insight into how AI evolved into what we see today.
This book grew out of that realization.
The goal of Sutskever's List is to illuminate the connections between key breakthroughs, placing each into a narrative that spans the deep learning revolution of the 2010s and early 2020s. Rather than treating each paper as an isolated achievement, the chapters weave them together to tell a larger story that connects technological innovation with shifting paradigms, cultural milestones, and evolving philosophies within AI. In these pages, you will discover what problems each solved, what doors they opened, and even what controversies or questions they raised.
By examining the landmark research through the eyes of Ilya Sutskever, the book offers a cohesive framework for understanding how we arrived at today's state of AI, and where we might be heading. You will meet the researchers who carried the torch of deep learning, the engineers who scaled algorithms, and even the critics and skeptics who challenged the hype. We also highlight how ideas build on one another: how a trick in one experiment enabled a breakthrough in the next, or how a theoretical insight found its way into practical systems. In doing so, Sutskever's List serves both as a map of modern AI's intellectual terrain and as a commentary on its journey.
Sutskever's List is the kind of book people might actually enjoy reading and finish, which is more than can be said for most books in machine learning and artificial intelligence, which are generally written to be referenced, not read. Specifically, this book is written for a broad but technically curious audience. Its primary audience is software engineers, data scientists, and machine learning practitioners. If you build or work with AI models, this book will enrich your understanding of why those models exist in their current form and the key engineering patterns that enabled them.
However, Sutskever's List is also meant to be accessible to motivated general readers. We assume only a basic familiarity with computing and math. If you have ever read popular science accounts of technology or enjoyed learning the stories behind scientific breakthroughs, you will find a similar narrative approach here, albeit one that does not shy away from the technical heart of each idea. In short, this book welcomes anyone eager to learn how AI evolved, offering different layers of insight for different readers. An engineer might appreciate the finer technical points or historical references, while a non-specialist will come away with a solid conceptual understanding.