Musica (AetherOS)

The Musica project is a research initiative within AetherOS designed to develop a cohort of autonomous AI agents, known as Musician ARCs, capable of generating and performing stylistically coherent music. The project serves as a primary testbed for the Hierarchical Reasoning Model (HRM) and the practical application of Energy-Maneuverability Theory to creative domains.

Mandate

The mandate of the Musica project is to create AI agents that can pass a "Musical Turing Test." This involves generating compositions that are not only technically correct but also emotionally resonant and stylistically indistinguishable from human compositions within a given genre. The project's secondary mandate is to serve as a production environment for developing the Producer ARC system.

Core Philosophy: Music as a Physical System

The central philosophy of Musica is the treatment of musical performance as a physical system governed by the laws of energy and maneuverability. Abstract musical concepts like harmony and rhythm are quantified as expressions of energy expenditure and strategic positioning.

• Energy State: The collective capacity of an ensemble to perform work, measured by the Musica Maneuverability Score.
• Virtuous Action: The execution of a single musical event (a note) that is within the ensemble's current energy capacity and aligns with the strategic musical goal. An action is virtuous if its "cost" (the Total Note Load) is less than the available "energy" (the MM Score).

System Architecture

The Musica ecosystem consists of two primary classes of agents: Musician ARCs and Producer ARCs.

Musician ARCs

These are specialized HRM agents trained to perform the role of a specific orchestral section (e.g., strings, woodwinds). Each Musician ARC is an independent entity whose intelligence is derived from a curriculum designed to teach fundamental music theory.

• Model: Hierarchical Reasoning Model implemented in PyTorch.
• Training Data: A synthetic curriculum generated by sagas/generation/generate_theory_curriculum_v4.py, which creates tens of thousands of contextual "problem/solution" pairs.
• Core Task: Given a musical context (key, tempo, beat position) and an input note, the ARC must generate a "virtuous" output command specifying the corrected note, velocity, articulation, and delay.

Producer ARCs

The Musica Operates system is a trio of meta-level agents (Guide, Navigator, Oracle) responsible for training, evaluating, and refining the Musician ARCs.

Key Metrics: The Musica Maneuverability Score

The Musica Maneuverability (MM) Score is the canonical metric for evaluating the health and performance capacity of an ensemble. It is a real-time index from 0-100 that quantifies the group's collective sonic energy.

The score is used to calculate the Note Virtuousness Score, which is the delta between the ensemble's capacity and the cost of an individual action: $\text{Note Virtuousness}=\frac{\text{MM Score}}{\text{Total Note Load}}$

The Total Note Load is the weighted sum of four components:

1. Frequency Load ($L_f$): The cost of a note's pitch (range extremity + harmonic dissonance).
2. Amplitude Load ($L_a$): The cost of a note's volume (dynamic extremity).
3. Duration Load ($L_d$): The cost of a note's rhythm (complexity + sustain).
4. Timbre Load ($L_t$): The cost of a note's articulation (technique complexity).

Current Status (August 24, 2025)

• Musician ARCs: Four section-specific models (Strings, Woodwinds, Brass, Percussion) have been successfully trained on Curriculum v4. Validation shows proficiency in scale adherence but indicates a need for more nuanced training in expressive dissonance.
• Producer ARCs: In the conceptual design phase. The project structure and development playbook have been outlined. The immediate next step is the development of the specialized script_writer.py toolchain.