This manifesto proposes and systematizes a new discipline: AI transhumanities — the projective humanities that employ artificial intelligence as an organ for the expansion of humanistic thinking. The term unites two genealogies: the transhumanities, a term the author first used in print in Transcultural Experiments (1999; drafted in 1994), restated in the 2006 manifesto "Towards the Techno-Humanities," and systematically formulated in The Transformative Humanities (2012) as the constructive branch of the human sciences; and the transformer architecture (2017), which underlies the dominant family of large language models — the first general-purpose technology to operate generatively, at civilizational scale, on humanity's accumulated language. The discipline is demarcated from transhumanism, from Boltuc's "transhumanities" as humanities for advanced AI, and from the currently prominent "humanities for AI," whose growing reciprocity it radicalizes by making humanistic invention itself a primary scholarly aim. Four branches are outlined, each grounded in existing practices and publications: instrumental humanities (formalized interpretive indices of the interesting, of plasticity, of fatum, and other cultural qualities); creatoric humanities (machines of concept-formation such as the Kreator and the Nootron); potentiological humanities (generative poetics: reconstructions and alternative versions of classical texts); and noontological humanities (the philosophy of alter-intelligence). The unifying principle of the discipline is modal: to the indicative mood of traditional scholarship the transformative humanities added the imperative; AI now makes the subjunctive operational — the humanities of the possible. The manifesto concludes with a definition of the discipline and seven theses. ( direct link )