Co-Intelligence

Ever since GPT-2 arrived, I’ve kept a small ritual. Each time a major model is released, I clear my diary and stay up late, testing it. What can this one do that the last could not? What can it do that I cannot? And—more sobering—what can I still do better? Ethan Mollick, the Wharton professor behind Co-Intelligence, calls this period the “three sleepless nights.”

Mollick is not your usual management academic. While others wrote white papers about LLMs, he was experimenting with them in public—one of the first “LLM mages” on Twitter—poking at their edges and uncovering their oddities. His early trials shaped how many of us came to think about AI, myself included.

If you make a living from knowledge and haven’t yet spent long evenings exploring AI, it’s time to wake up. The landscape has shifted. But not evenly. Mollick calls this unevenness the Jagged Frontier—the messy border between competence and collapse. LLMs can write competent code and tolerable poetry, yet struggle to count the “r”s in “strawberry.” The only way to map this frontier is to traverse it yourself.

That process rewires how you think. It replaces naïve awe with a working mental model. When colleagues ask how to approach AI, I tell them this: treat it like an over-eager, well-read intern. The intern is bright, fast and keen to impress—but unreliable, easily confused, and occasionally delusional. They know trivia in depth but lack any sense of context or craft. They can tell you who painted the Sistine Chapel but not what it smells like inside.

Mollick captures this duality well. Co-Intelligence doesn’t deal in vague futurism. It’s a manual for the present. He argues that hallucinations—the AI’s tendency to invent—aren’t simply flaws but signals that humans must remain in the loop. The danger creates the partnership.

That is the core of Co-Intelligence: a framework for collaboration between human and machine. AI, Mollick suggests, is a force multiplier for creativity and pattern recognition. He offers practical methods for using it to brainstorm, analyse and build. His research into technological innovation gives his argument weight: disruption tends to defeat organisations but empower individuals. Those who are willing to experiment—to trace the Jagged Frontier and merge human judgment with machine speed—will outpace their peers tenfold.

History offers perspective. I am typing this on a computer, yet only decades ago “computer” referred to a person—someone like Dorothy Vaughan, the human mathematician whose work at NASA helped launch rockets. The profession vanished quietly. Others will follow. “Doctor.” “Lawyer.” “Trader.” Words with long pedigrees may not always denote people.

Mollick’s message is blunt but hopeful: you can drift with the current, or you can steer. The tools exist now. The question is whether you’ll learn to wield them before they render you obsolete.

Co-Intelligence therefore is a fine place to start.