Co Intelligence
Ethan Mollick
one sitting
I read it in one sitting. The clearest thinking I've found on working with these models instead of fighting them.
Ethan Mollick teaches at Wharton and has spent the last few years putting AI into his teaching and research. This book is what he's learned from doing that, and it's the most practical book on AI I've read.
He gives four rules. Always invite AI to the table: try it on everything, because you won't know where it helps until you do. Be the human in the loop, checking and deciding. Treat it like a person, and tell it exactly what kind of person you need it to be. And assume this is the worst AI you will ever use, because the next one will be better.
The idea I think about most is what he calls the jagged frontier. AI is brilliant at some tasks and oddly bad at others that look just as hard, and you can't tell which is which without testing. He describes a study with consultants using AI, where the ones working inside the frontier did better work, faster. The ones who trusted it on tasks outside the frontier did worse than people with no AI at all.
I read it in one sitting because it described exactly what I'd been doing. Vibe Link got built by vibe coding my way through a stack I'd never used, and that's the jagged frontier in practice: learning where the model is sharp and where I have to take over. AetherMem exists because I kept running into the edge where it forgets.
