tl;dr:The intention behind the Code Review Pyramid is to help put focus on the parts which matter the most during a code review, and what can be automated. The peak of the pyramid are matters that require small effort to change later on i.e. code style and tests, and the bottom require significant effort i.e. documentation, implementation and API semantics.
tl;dr:Three pieces of information should be conveyed by an error message: (1) Context: What led to the error? What was the code trying to do when it failed? (2) The error itself: What exactly failed? (3) Mitigation: What needs to be done in order to overcome the error? Gunnar also highlights best practices: having a uniform voice and style, consistency in terminology, and more.