So, if instead of designing one "Dog" class with "Bark", "Sleep" and "Eat" methods, I must design "AnimalWhoBarks", "AnimalWhoSleeps", "AnimalWhoEats" classes, etc? Why? How does that make my code any better? How am I supposed to simply implement the fact that my dog will not go to sleep and will bark all night if he hasn't eat?

"Split big classes in smaller ones", is a fine practical advice, but "every object should have a single responsibility" is an absolute OOP dogmatic nonsense.

Imagine the .NET framework was written with SRP in mind. Instead of 40000 classes, you would have millions.

Generalized SRP ("Generalized" is the important word here) is just a crime IMHO. You can't reduce software development to 5 bookish principles.