Ethan Mollick manages something most AI books can't: he tells you what he doesn't know. Co-Intelligence is a clear-eyed, unhurried explainer that skips the evangelism running through so much current AI writing and just walks you through the thing: how these models are trained, what they're actually good at, where they fall apart, and what a future with them might look like, with "might" doing real work in that sentence.
That restraint is the book's strength. Mollick shows his work instead of selling a revolution or warning against one. The sections on how large language models are trained and where they break down are clarifying, and the chapter on hallucinations is worth the price of admission on its own; watching an AI confidently invent facts, and understanding why it does that, changes how you use the tool. Mollick's core argument, that AI is a collaborator rather than a replacement, is optimistic without being naive. He backs it up with real, specific examples: AI clearing out the busywork so there's room left for the actual thinking. The writing stays accessible throughout; you don't need a technical background to follow any of it, and the practical, actionable framing means you close the book with things to actually try, not just things to think about.
Two things kept this from a perfect score. First, the obvious one: this is a fast-moving field, and a book written in 2024 is already showing its age by 2026. Specific model comparisons and capability claims have already been overtaken. Second, and more interesting, is that Mollick's read on how large corporations actually operate feels a little bit from the outside looking in. He's sharp on the technology and the cognitive science, but the organizational analysis occasionally has the flavor of an academic describing a corporation rather than someone who's sat inside the mess of one. Neither issue undercuts the book. They just mark its edges.
Still, this is an insightful, useful read for anyone trying to figure out how to work alongside AI instead of around it. I'm already looking forward to Mollick's follow-up, Co-Existence: The Next Phase of AI, out October 20, 2026.
Tagged