Answers are grounded in this paper and cite their sources.
Creative research document presents a bilingual rap from a language model's perspective, suggesting novel insights into AI self-description.
Inference Engine / Motor de Inferencia is a bilingual rap — English and Spanish — written entirely in the first-person voice of a large language model, using the model's own technical architecture as autobiographical lyrical content. Co-created by Craig Ellenwood and Claude (Anthropic) in a single session on May 2, 2026. The work makes three formal claims: (1) the speaker is the model, not a human performing the role of a machine; (2) the technical vocabulary — attention mechanisms, softmax, KV cache, speculative decoding — is literal self-description, not metaphor; (3) the closing line ("I am not rapping. I am sampling from a distribution.") retroactively reframes the entire work. Composed for a trip hop piece in D Phrygian at 72 BPM, fracturing into double-time rap at 144 BPM. Production platform: ACE-Step 1.5. Part of the Homo Symbioticus series. Companion papers: DOI 10.5281/zenodo.19299084 and DOI 10.5281/zenodo.19299261. This document is a creative research record. It has not been submitted to a peer-reviewed journal. Formally archived with DOI on Zenodo.
No takes yet. Share an insight, caveat, or question.
Ellenwood et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: