Steve Russell comment is about writing the actual interpreter for Lisp, so the comment is correct on that Lisp started out as an interpreted language, and even mentions that in the comment that you quote.
The interpreter (i.e. Lisp READ-EVAL-PRINT loop) was written in IBM 704 machine code, which just was a reimplementation of the EVAL function as JMC described it in his paper.
The part that "wasn't supposed to be implemented" was about the form of Lisp, M-expressions vs. S-expressions not Lisp it self. Which was supposed to be implemented as a programming system from day one.
The interpreter (i.e. Lisp READ-EVAL-PRINT loop) was written in IBM 704 machine code, which just was a reimplementation of the EVAL function as JMC described it in his paper.
The part that "wasn't supposed to be implemented" was about the form of Lisp, M-expressions vs. S-expressions not Lisp it self. Which was supposed to be implemented as a programming system from day one.