Co-Intelligence by Ethan Mollick (2024)

My work centers on researching and building AI applications, and I manage a broad software team. Some of my colleagues dive headlong into "vibe coding" with generative AI assistants and agentic workflows, while others resist or adopt slowly. Outside of work, I experiment with chatbots and occasionally use them for personal coding, writing, and evaluation tasks. One of my team members recommended Co-Intelligence, so I began with the audiobook on a long flight. I retained little from that attempt. Reading the ebook with notes proved more effective, and I extracted the key insights.

I approach most AI commentary with skepticism. The hype bubble, inflated by the trillion-dollar "magnificent seven," resembles many earlier cycles. Still, just as calculators, GPS, and spell-checkers raised the floor of productivity, generative AI is now advancing the Gartner "plateau of productivity."

Mollick's enthusiasm is obvious, but his "four principles" are sound: (1) treat AI as a partner, not a tool; (2) use it often to learn its strengths and limits; (3) be transparent about when and how you employ it; and (4) expect disruption as norms shift. His prompt-design examples are pragmatic, too. Roughly one-third of the book delivers practical value; the rest dissolves into speculation and fanboy exuberance. Even so, the book is worth reading. 4/5 stars.