The original Smalltalk in 1972, the language of which Kay designed, indeed had some kind of message passing (even though it was synchronous, the receiving object interpreted messages composed of tokens). Smalltalk-76, essentially designed by Ingalls, who was also the sole author of the 1978 publication, made a fundamental shift towards compiled virtual methods, essentially as it was done in Simula 67 and adopted by C++ (though much less efficient). So yes, it makes pretty much sense when Kay claims that he didn't have C++ nor Smalltalk in mind when talking about OO. See also https://dl.acm.org/doi/abs/10.1145/3386335.