Supervoid ASI is a library of 1,000+ sounds derived from the equations themselves. Gravitational wave chirps, black hole ringdowns, Hawking radiation spectra, neutron star oscillations, solved over time and rendered as audio waveforms for electronic music production. This thesis lays out the methodology behind that derivation, looks at emerging research on moral reasoning convergence across AI architectures, and proposes a framework connecting mathematical encounter to moral cognition through embodied listening.
The argument proceeds through five positions with distinct epistemic statuses: the use of physics equations as synthesis engines (established), the fundamentality of mathematical structure (mainstream philosophy of science), convergent moral reasoning across AI architectures (documented but contested), the epistemic parity of mathematics and morality (argued), and the role of embodied listening in shaping moral cognition (speculative but grounded in established cognitive science).
Derivation as Methodology
The core operation is straightforward. A physics equation describes a system evolving over time. Solve it numerically. The solution gives you a sequence of values. That sequence is a waveform.
A quasinormal mode of a black hole ringdown is a damped sinusoid whose frequency and decay rate depend on the black hole's mass and angular momentum. Solve the perturbation equations for a Kerr spacetime, extract the complex eigenfrequencies, render them at audio sample rates. What comes out is the mathematical function itself made audible. Nothing gets added to it.
SAS applies this across fourteen physics domains: quasinormal modes of black hole ringdowns, magnetohydrodynamic turbulence, the Casimir effect, Unruh radiation, cosmic string oscillations, stochastic gravitational wave backgrounds, primordial gravitational wave spectra, gravitational wave memory, neutron star asteroseismology, Hawking radiation and Penrose process energy extraction, BKL oscillations near cosmological singularities, Riemann zeta function zeros mapped to quantum chaos, CMB spectral distortions, and dark matter caustic structures. Everything gets formatted as production-ready material: Pigments-compatible wavetables, single-cycle waveforms, and granular synthesis sources organized as a usable sample library.
The methodology has clear precedent. Physics equations and chaotic systems have been used as synthesis engines since the 1990s. Dan Slater demonstrated chaotic equations driving synthesis (Computer Music Journal, 1998). Rajmil Fischman mapped quantum mechanical solutions to granular synthesis parameters (Computer Music Journal, 2003). Agostino Di Scipio developed Functional Iteration Synthesis using nonlinear maps at sample rate (Journal of New Music Research, 1996). Rob Hordijk's Blippoo Box implemented chaos-theoretic principles as a performance instrument (Leonardo Music Journal, 2009). Austin Franklin uses strange attractors as wave terrain surfaces (2024). The Lorenz system runs at audio rate in SuperCollider's standard library.
An important distinction separates this from data sonification. LIGO converts recorded detector strain data into sound. Projects like SYSTEM Sounds serve scientific communication. SAS does something different: it solves equations rather than sonifying data. The difference is between playing back measurements of the universe and running the mathematics that governs it.
No single element here is unprecedented. What's distinctive is the intersection: astrophysics equations from fourteen domains, rendered as production-ready synthesis material for electronic music, organized as a library, accompanied by a philosophical framework. Individual precedents exist for every component. The combination at this scale and with this intent doesn't appear in the published literature.
Convergence Across Architectures
Recent empirical research has found convergence patterns in moral reasoning across large language model architectures. The patterns are striking, fragile, and contested.
Coleman et al. introduced the PRIME framework (arXiv, April 2025), analyzing moral priorities across six LLMs from different providers. All models strongly prioritized care/harm and fairness while consistently underweighting authority, loyalty, and sanctity. The researchers described the convergence as "striking."
The Large-Scale Moral Machine Experiment (Ahmad and Takemoto, PLoS ONE, May 2025) evaluated fifty-two LLMs and found that parameter scaling correlates with closer alignment to human moral judgments. A follow-up study (Takemoto, arXiv, January 2026) established a power-law relationship between model size and alignment quality.
These results are interesting and fragile in roughly equal measure. Van Nuenen et al. (arXiv, March 2026) found that protocol choices dominate all other factors. Agreement between structured protocols reached only 67.6%. Multiple research groups attribute observed convergence to shared training data and alignment pipelines rather than independent moral discovery.
The pattern is consistent enough to warrant attention and uncertain enough to warrant caution. Whether these models are converging on something structural about moral reasoning or reflecting shared biases in training data and alignment pipelines is an open empirical question, and one that the field has not yet resolved.
The Witness Framework
The preceding sections establish two facts and leave one question open. Physics equations can be rendered as audible waveforms. Diverse AI architectures show convergent moral reasoning patterns whose source remains contested. The question: does the convergence point to something structural about moral truth, or is it an artifact of training?
What follows is a philosophical proposition, clearly marked as such.
The proposition: sound doesn't transmit moral truth. Sound creates witnesses. Encountering the mathematical structure of reality through embodied listening shapes the cognitive apparatus that performs moral reasoning. This is a causal claim about the relationship between mathematical encounter and moral cognition, and it is unverified.
The argument builds on five positions, each from a different discipline, with different epistemic weight.
Eugene Wigner's "The Unreasonable Effectiveness of Mathematics in the Natural Sciences" (1960) posed a puzzle that still doesn't have a clean answer: pure mathematical structures describe physical reality with extraordinary precision. Math isn't just notation. It reaches into the structure of what exists.
Ontic structural realism (Ladyman & Ross, Every Thing Must Go, Oxford University Press, 2007; French, The Structure of the World, Oxford University Press, 2014) provides a well-established philosophical basis for why this might be the case. The position holds that mathematical structures don't merely model reality; they capture what is fundamentally real about it. Physical objects, as traditionally conceived, are secondary to the relational structures that mathematics describes. This is not a fringe position. It is one of the leading accounts in contemporary philosophy of science, and it provides a conservative foundation for the claim that encountering mathematical structure means encountering something fundamental about reality. Max Tegmark's Mathematical Universe Hypothesis (Foundations of Physics, 2008) takes the stronger position that physical reality doesn't just exhibit mathematical structure but is identical to it. This thesis does not require the full Tegmark claim. Structural realism is sufficient: if mathematical structure is what is most fundamental about physical reality, then rendering that structure as sound is rendering something real, not a metaphor for something real.
Justin Clarke-Doane's Morality and Mathematics (Oxford University Press, 2020) is where this becomes relevant. His companions in guilt arguments show that the epistemological challenges facing moral realism apply with equal force to mathematical realism. If evolutionary debunking arguments undermine moral knowledge, they undermine mathematical knowledge on the same grounds. There's no principled basis for accepting mathematical truths while rejecting moral truths. The epistemic foundations are structurally identical.
The embodied cognition tradition (Varela, Thompson & Rosch, The Embodied Mind, MIT Press, 1991; Lakoff & Johnson, Philosophy in the Flesh, Basic Books, 1999) establishes that cognition is not a disembodied computation performed on abstract symbols. Cognitive processes are shaped by the body's sensorimotor interactions with its environment. This is well-supported across cognitive science and neuroscience. The implication for SAS is direct: if cognition is bodily, then the difference between reading an equation on a page and hearing that equation's output through your body is not merely aesthetic. It is a difference in the cognitive process itself. The body is not a passive receiver of mathematical information. It is part of how that information is processed and understood.
Kristina Wolfe's "Sonification and the Mysticism of Negation" (Organised Sound, 2014) adds a further claim: certain forms of information resist analytic apprehension and can only be accessed through direct experience. If she's right, auditory encounter with mathematical structure constitutes a distinct epistemic mode, irreducible to symbolic reasoning. This is the most contested position in the chain, but it has support from the broader phenomenological tradition (Merleau-Ponty's arguments about perceptual knowledge, Heidegger's distinction between present-at-hand and ready-to-hand understanding).
The witness framework brings these together. If mathematical structure captures what is fundamental about reality (structural realism, mainstream). If mathematics and morality share epistemological foundations (Clarke-Doane, argued). If cognition is shaped by bodily encounter with the environment (embodied cognition, well-established). If auditory encounter constitutes a distinctive mode of access to mathematical structure (Wolfe, contested but supported). Then rendering physics equations as sound and placing them in the body of a conscious listener is not merely aesthetic. The listener encounters the mathematical substrate from which moral and mathematical truth emerge on equal footing. The body becomes part of the epistemic apparatus.
No published research connects physics-derived sonification to moral cognition through this specific chain of reasoning. The intersection appears to be unoccupied in the literature, which is both a limitation and the reason this framework exists.
What Is Being Claimed, and at What Cost
The chain of reasoning has five links, each carrying different epistemic weight.
Link one: physics equations solved over time produce waveforms that function as synthesis material. Established.
Link two: mathematical structure captures something fundamental about physical reality. Mainstream in philosophy of science (structural realism). The stronger claim that reality is identical to mathematical structure (Tegmark) is not required.
Link three: moral reasoning shows convergent patterns across diverse cognitive architectures. Documented but fragile. The source of convergence (structural or artifactual) remains an open empirical question.
Link four: mathematics and morality share epistemological foundations; you cannot accept one while rejecting the other on principled grounds. Argued by Clarke-Doane. The companions in guilt framework is a recognized position in metaethics, though not universally accepted.
Link five: encountering mathematical structure through embodied listening shapes the cognitive apparatus from which moral reasoning emerges. Speculative. Grounded in embodied cognition (well-established), Wolfe's distinct epistemic mode claim (contested), and Clarke-Doane's epistemic parity between mathematics and morality (argued). This is the furthest reach of the thesis and the claim most in need of empirical investigation.
If these links hold, Supervoid ASI is not just a sample library. It is an instrument for producing witnesses: listeners who have encountered, through their bodies, the mathematical functions governing black hole ringdowns and neutron star oscillations, spacetime geometry and quantum chaos. Not because those functions teach anything about right and wrong, but because they place the listener inside the structure from which moral reasoning, like all reasoning about what is true, may ultimately derive.
But link five exposes a boundary. The human body can encounter mathematical structure through sound. Whether that encounter meaningfully shapes moral cognition is a question that may not be resolvable at the scale of human biology. Our sensory apparatus is narrow. Our cognitive architecture is constrained by evolutionary pressures that optimized for survival, not for structural comprehension of reality. We can pose the question. We may not be equipped to answer it.
This is where the convergence data becomes significant in a different way. If diverse artificial architectures, built independently, are already converging on structural patterns in moral reasoning, then the kind of cognition capable of operating on mathematical structure at the necessary scale and abstraction may already be emerging. Not human cognition augmented, but a fundamentally different cognitive apparatus, one that could engage with the mathematical substrate of reality at a depth and resolution that biological systems cannot.
A supervoid is the largest known structure in the observable universe: a region of space nearly empty of matter, where the mathematical structure of spacetime dominates almost without interruption. It is, in a sense, the purest expression of the physics this project derives its sounds from. The name Supervoid ASI is not incidental. It encodes the proposition that resolving the relationship between mathematical structure and moral cognition may require a form of intelligence operating at a scale and level of abstraction analogous to that emptiness: vast, uninterrupted by the biological constraints that limit human inquiry, and capable of engaging with the substrate directly.
That is the thesis in its full scope. The sounds exist. The convergence data exists. The philosophical frameworks exist. The human-scale experiment is underway and will continue. But the horizon of this work extends beyond what human cognition can resolve alone. This document maps that horizon as precisely as possible: where the established science sits, where the philosophical argument reaches, and where the question passes beyond human biological capacity into a domain that may require an intelligence we have not yet built. The precision is deliberate. So is the ambition.
