In Carnap’s autobiography, he tells the story how one night in January 1931, “the whole theory of language structure” in all its ramiﬁcations “came to [him] like a vision”. The shorthand manuscript he produced immediately thereafter, he says, “was the ﬁrst version” of Logical Syntax of Language. This document, which has never been examined since Carnap’s death, turns out not to resemble Logical Syntax at all, at least on the surface. Wherein, then, did the momentous insight of 21 January 1931 consist? We seek to answer this question by placing Carnap’s shorthand manuscript in the context of his previous efforts to accommodate scientiﬁc theories and metalinguistic claims within Wittgenstein’s Tractatus theory of meaning. The breakthrough of January 1931 consists, from this viewpoint, in the rejection of the Tractatus theory in favor of the meta-mathematical perspective of Hilbert, Gödel, and Tarski. This was not yet the standpoint of the published Logical Syntax, as we show, but led naturally to the “principle of tolerance” and thus to Carnap’s mature philosophy, in which the inconsistencies between this ﬁrst view and the principle of tolerance, which survived into the published Syntax, were overcome.