The speaker will analyze the concrete modalities of the production and recognition of a logical theorem in the field of artificial intelligence in the 1990s. Ethnographic observations, interviews, and textual analysis, reveal the impact of a heterogeneity of practices of evaluation and other forms of interaction between the author of the theorem and a varied community of interlocutors, especially during the draft stage of the theorem. Individual and collective representations of the theorem were structured by the proliferation and polysemy of its reformulations, by the imperfect access to proofs and counterproofs, and by the coordination of action within opposed groups. The stabilization of debates over, and the certification of, the theorem were not based on the simple victory of one side over another but on relatively unified responses by critics and by the author’s responses to critiques—responses that tended to allow for multiple interpretations. This ethnography of logic in development illustrates why social scientists should not consider logic to be just a methodological tool; it is also a privileged object that enables exploration of the material and social forms of intellectual work, including the building of credibility.