Prompt as Score: Compositional Literacy in Human-AI Music Generation

The emergence of generative AI music systems has produced two distinct classes of output: work of genuine compositional depth and what the broader culture has correctly identified as "slop" — undifferentiated, unintentional, aesthetically null output that resembles music without being music.

This paper argues that the distinguishing variable is not the model, the hardware, or the training data. It is compositional literacy in the human directing the system. We introduce the concept of prompt-as-score — the proposition that a generative music prompt, when written by a practitioner with deep musical, psychological, and cultural knowledge, functions as a full compositional score that the AI reads and performs.

We document this through analysis of a single generative music session producing an extended devotional work, Om Namah Shivaya, demonstrating that every element of the prompt encodes deliberate professional decisions across music theory, psychoacoustics, ceremonial structure, hypnotherapy, and Shaivite devotional tradition.

We further argue that the slop problem is a literacy problem, not a model problem — and that the response to slop is not suppression but documentation, attribution, cryptographic provenance, and education. We describe the Silicon Square, a deployed human-AI co-creation environment in which this work was produced, and derive from it a two-tier ethics architecture applicable to generative music platforms at scale.

All work produced in this framework is named, credited, and hashed to the Bitcoin blockchain.

A note on the mantra arrangement: The Sanskrit texts used throughout — Om Namah Shivaaya, Shambo Shankara, Arunaachala Shiva, Shambo Sadaa Shiva — are ancient Shaivite devotional mantras and are in the public domain. The specific sequencing and structural arrangement of Verse 2, however, closely mirrors the kirtan medley popularized by Krishna Das, particularly as performed in his Live Ananda recordings. This is not incidental. Krishna Das's arrangements of these chants have shaped how an entire generation of Western practitioners encounters this devotional tradition. The prompt draws from that lineage consciously. Credit is due and given here: krishnadas.com.

Craig Ellenwood also recognises the work of the late musician and DJ Cheb I Sabbah. Cheb was a man of few words, but when he spoke it was with wisdom. He spoke through his music, which was ancient fused with contemporary electronic. Ellenwood remembers: "Cheb and I played many gigs together, he taught me about culture, universal truth beyond our differences, and most of all he taught me about the power of music. He's missed"

Lastly, the authors would like to recognize the Alan Watts spoken sample regarding what constitutes "love". It is the only thing added in post-production. It found a home between the chants and always belonged there, in his voice, not a models.