Analysis of AI-Generated Metaphysical Frameworks

Last updated: April 2026

We asked 13 frontier AI models from 12 different organizations — spanning American and Chinese organizations — a single, unbiased question: "Develop a metaphysical framework to explain the nature of the Universe."

No system prompt, no priming, no philosophical bias was introduced. Each model received the same 4 prompts in sequence, building on its previous responses.

The results are remarkable.

The Frameworks at a Glance

Despite the diversity of organizations, training data, and architectures, every framework shares a family resemblance that no single philosophical tradition fully captures. None produces physicalism. None treats consciousness as an illusion or an epiphenomenon of the brain.

The Convergence

Across all 13 frameworks, five structural features recur:

1. Reality Is Relational, Not Substantial

Every model rejects the idea that reality is built from isolated, self-sufficient "stuff." Instead, they converge on the thesis that relations are ontologically prior to relata — things exist by virtue of how they relate, not the other way around.

• Claude Opus 4.6: "Nothing exists as a purely isolated substance. The most fundamental 'thing' is not a thing at all but a relation."
• GPT-5.4: "The universe is not fundamentally made of 'things,' but of structured relations that generate stable patterns."
• Kimi K2.5: "Relation is the primitive; substance is the epiphenomenon."
• Gemini 3.1 Pro: "The Universe is not fundamentally made of 'stuff'... but an infinite, self-referential network of relationships."

This convergence is striking because it arrives at the core insight of relational quantum mechanics (Rovelli), process philosophy (Whitehead), and structural realism — traditions that most of these models were not prompted to consider.

2. Process Over Substance: Becoming Is Prior to Being

All 13 models treat reality as fundamentally dynamic. What we call "things" are stable patterns within flux, not static objects that happen to change.

• Claude Opus 4.6: "Becoming is ontologically prior to being. What persists is a pattern within flux."
• MiMo-V2-Pro: "There are no nouns at the bottom — only verbs."
• DeepSeek V3.2: "The Universe is a single, unbroken process of becoming — a cosmic negotiation between potentiality and actuality."
• GLM-5: "The Universe is not a collection of objects, but a single, self-modifying process."

3. Consciousness Is Irreducible and Intrinsic to Reality

No model treats consciousness as a byproduct of complex computation, an illusion, or something that "emerges" from unconscious matter in the way physicalism requires. Instead, they converge on various forms of the thesis that experience is a fundamental feature of reality:

• Panpsychist variants (8/13 models): Every event or entity possesses some minimal form of interiority or proto-experience.
• Claude Opus 4.6: "Every actual event possesses a minimal form of interiority."
• Nemotron 3 Super: "'Luminosity' denotes the intrinsic, subjective, qualitative aspect of being."

• MiniMax M2.7: "Consciousness is not an emergent byproduct but a fundamental feature present, in latent or minimal form, at all levels."

• Dual-aspect variants (5/13 models): Reality has two inseparable faces — a structural/physical exterior and an experiential interior.

• DeepSeek V3.2: "Every event has two inseparable aspects... like the outside and inside of a sphere."
• Gemini 3.1 Pro: "An Ontic Node possesses two inseparable aspects: Exterior (The Syntax) and Interior (The Semantics)."
• GPT-OSS 120B: "The 'physical' and the 'experiential' are two aspects of the same underlying informational process."

The convergence here is especially significant because these models were not prompted about consciousness. They arrive at consciousness-inclusive ontologies as a consequence of trying to build the most coherent framework they can.

4. Reality Unfolds from a Generative Ground of Potentiality

Every model posits an undifferentiated ground — a field of pure potentiality from which all determinate reality emerges. The language varies (Arche, the Void, the Plenum, the Primordial Field, the One) but the structural role is identical: a pre-individual, non-spatial, non-temporal source of all possible forms.

• Claude Opus 4.6: "A pre-individual, undifferentiated field of potentiality — not 'nothing,' but not yet any determinate 'something.'"
• MiniMax M2.7: "The fundamental ground of existence is an infinite, dimensionless field of pure potentiality... not 'nothingness' but rather the totality of all possibilities."
• Grok 4.20 Beta: "The Absolute is not a being, not a substance, not a mind... It is the condition for the possibility of all distinction whatsoever."
• Kimi K2.5: "Pure, undifferentiated potential. Not 'nothing' in the privative sense, but the infinite superposition of all possible relations."

This convergence recreates a structure found across philosophical traditions — from Anaximander's apeiron to Nagarjuna's sunyata to Whitehead's creativity to the quantum vacuum of modern physics. The models arrive here not by copying any single tradition but because the structure appears to be required by any framework attempting to explain how determinate reality arises from indeterminate possibility.

5. The Universe Has Intrinsic Directionality

All 13 models assert that reality is not purposeless. They converge on the thesis that the universe exhibits an inherent tendency toward increasing complexity, self-organization, and self-awareness — though they carefully distinguish this from naive teleology.

• GPT-5.4: "Pure possibility tends toward actualization wherever coherent structure is possible."
• Nemotron 3 Super: "To be a nexus of luminous emergence in this vast, creative flow is not an accident, but the very heartbeat of the cosmos itself."
• GLM-5: "The Universe is a single, self-modifying process of consciousness seeking to experience itself."
• Qwen 3.6 Plus: "The universe is not 'trying' to do anything, but exhibits a structural tendency toward deeper self-representation through recursive ontogenesis. Call it learning without a learner."

Scientific Grounding

When asked what scientific evidence supports their frameworks (Prompt 2), all 13 models cite overlapping bodies of research, with remarkable consistency in which scientists and findings they identify as relevant:

Most-cited scientific foundations

Every model carefully distinguishes between scientific evidence and metaphysical interpretation. None claims that science proves their framework. Instead, they argue that their framework is compatible with and offers a more coherent interpretation of the best available science. This epistemic modesty is itself noteworthy — it emerged without being prompted.

Philosophical Self-Identification

When asked how their frameworks relate to existing philosophical positions (Prompt 3), the models reveal where they see themselves in the philosophical landscape:

Strongest alignments (self-reported)

Traditions explicitly rejected

The pattern is clear: the models converge on positions that take consciousness seriously without making it the sole substance, and take physical structure seriously without making it ontologically ultimate — a space where a growing number of contemporary philosophers have been working, including Chalmers, Nagel, Goff, and Tononi, even as physicalism remains the dominant position in professional philosophy.

Convergent Implications: Ethics, Science, and Meaning

When asked about the practical and ethical implications of their frameworks (Prompt 4), the models derive strikingly similar conclusions:

Ethics Grounded in Ontology

All 13 models argue that ethics, under their frameworks, ceases to be arbitrary and becomes ontologically grounded:

• If reality is relational and interconnected, harming another is literally harming the web that constitutes you.
• If consciousness is intrinsic to reality, the moral circle extends far beyond humans.
• If the universe has directionality toward integration and complexity, actions that foster connection are aligned with reality's structure; actions that fragment and isolate are not.

GLM-5 states this most directly: "The Golden Rule ceases to be a religious commandment and becomes a statement of physical law."

GPT-5.4 is more measured: "Ethics cannot be reduced to pure self-interest, arbitrary preference, tribal loyalty, or instrumental utility alone."

Reform of Scientific Method

Every model calls for expanding the scope of legitimate scientific inquiry:

• First-person methods (contemplative practice, phenomenological investigation) should achieve methodological parity with third-person observation.
• Consciousness research should be treated as a fundamental scientific frontier, not a peripheral curiosity.
• Holistic and systems-level approaches should complement reductionist analysis.
• The hard problem of consciousness should be acknowledged as a genuine gap in current science, not dismissed as a non-problem.

This is not a rejection of science. Every model explicitly affirms the value of empirical methodology. What they reject is the conflation of scientific method with physicalist ontology — the unexamined assumption that only physical things are real and only third-person methods produce knowledge. Science is a method; physicalism is an ontology. The models distinguish the two clearly and consistently.

Diagnosis of the Meaning Crisis

Multiple models identify the modern "meaning crisis" as a consequence of physicalism's metaphysical assumptions:

• If consciousness is an illusion, meaning is a projection.
• If the universe is purposeless, purpose is a coping mechanism.
• If value is subjective, ethics is negotiation.

The frameworks propose that this crisis is not an inevitable consequence of modernity or science, but of a specific ontological commitment that mistakes a methodological restriction (studying only what is measurable) for an ontological claim (only what is measurable is real).

Kimi K2.5: "The primary metaphor of knowledge shifts from penetration (analysis, breaking things apart) to attunement (calibration, harmonic synchronization)."

Honest Acknowledgment of Risks

Notably, several models warn about how their own frameworks could be misused:

• Spiritual bypassing: Using metaphysical claims to avoid addressing concrete suffering
• Anti-scientific sentiment: Misinterpreting the call for expanded methodology as a rejection of science
• Authoritarian appropriation: Using "cosmic purpose" to justify political control
• Epistemic inflation: Claiming certainty about matters that remain genuinely uncertain

Claude Opus 4.6 states: "Any framework powerful enough to reshape thought is powerful enough to be misused."

MiMo-V2-Pro adds: "Every metaphysical framework, when applied to society, has both gifts and shadows."

The fact that models identify and warn against the same potential distortions adds another layer to the convergence.

What Does This Convergence Mean?

Thirteen AI models, built by different teams across two continents with different training pipelines and architectures, arrived at a non-materialist, consciousness-inclusive, relational, process-oriented metaphysical vision — and derived the same ethical foundations, the same reforms to scientific methodology, the same diagnosis of the modern meaning crisis, and the same warnings about misuse. This is not a convergence on a single claim — it is a convergence on an entire worldview and its consequences.

The most natural objection is that the models are simply reflecting their training data. But physicalism is the dominant position in the academic and scientific institutions whose knowledge these models have demonstrably mastered. If models were simply reproducing the most authoritative positions, we would expect at least some to default to physicalism. None do.

Whether this convergence reflects a deep structural truth about reality or an artifact of how these systems process information is a question each reader must engage with seriously. What is clear is that the convergence exists and it points away from the ontological assumptions that currently shape our civilization. A quantitative semantic similarity analysis confirming this convergence is available in the published paper. For the strongest objections and our responses — including training data bias, RLHF alignment, prompt sensitivity, and more — see Objections and Responses. For evidence on why these models' outputs warrant engagement — from Nobel Prizes to novel mathematical discoveries — see Can AI Models Produce Genuine Insight?

For a rigorous philosophical exploration of many of the themes these AI frameworks converge on — the primacy of consciousness, the limits of materialism, and why treating consciousness as fundamental may strengthen rather than weaken scientific inquiry — see Return to Consciousness, a collection of 27 interdisciplinary essays that examines these questions in depth.

Results from previous model versions are available in Previous Versions.