Bad news: you don't have much of a choice when it comes to authoritative textbooks on information theory. Good news: Cover is not a bad book to be stuck with.

Every chapter, theorem, idea, and problem is well-motivated. Landau-lovers of terse textbooks might bawk at the equation-less pages preceding some results, but I greatly appreciated the intuitive prefaces to mathematical formalism.

This book also has one of the finest problem sets I've seen. Each problem is designed to convey a new idea or point out a subtlety of one mentioned in the text. Striking just the right balance of guidedness with openness, doing Cover problems feels more being in a sandbox than a forced labor camp.

I laughed, I cried, I learned.

Two criticisms though. First, Cover has a tendency to mention vocabulary and use notation that he has not yet introduced and this can make for some frustrating battles of interpretation. Second, some of the problems require a good amount of knowledge not found in this book. I took a course based on this book with zero background in electrical engineering (I'm get my kicks from physics and maths) and at times, I felt like I was trying to crack the Rosetta Stone.