I get that this guy tries to take a perspective based on error reporting, but it really has nothing to do with the difference between interpreted and compiled.

I get that this guy tries to take a perspective based on error reporting, but it really has nothing to do with the difference between interpreted and compiled. I've built a scripting language with a full lexer and parser. It is capable of giving errors in both stages. It does not make it compiled or even interpreted. It is first when you execute the instructions of the language it becomes an interpreted language. It is a good point that "compiled" is not a boolean property. A language like C# can be interpreted, JIT compiled and AOT compiled - it all depends on the toolchain used. However, usually developers refer to compiled languages as those that gets translated down to machine instructions, and to my knowledge, CPython does not do that.