A convergent research framework · 2025–2026
Connectivity is encoded at a level above physical proximity. Form precedes matter.
This is not metaphor — it is a convergent finding across neuroscience, quantum physics, Vedic philosophy, and AI interpretability research.
The Central Idea
The medium is different.
The melody remains the same.
Mind may not be attached to the body,
but may be a structure — a pattern.
Transfer this idea to consciousness
If consciousness is not the neurons themselves, but their organization — then it may exist on another medium.
SUBSTRATE
Medium / foundation. The thing that carries the pattern. Changes across implementations.
LOGOS
Structure + meaning + logic. What matters is not the material, but how information is organized.
SEED
A seed becomes a tree. A few lines of code become a large AI system. Pattern precedes form.
The LOGOS Formula
P(Meaning | ε) ∝ Tr(ρ · ΠData)
Meaning appears not simply from data, but from the interaction between data and the structure of the system.
What is more important —
the brain, or the pattern
created by the brain?
Three Frameworks
I — 01
The same structural patterns appear in silicon (AI), carbon (biological neurons), light (quantum systems), and meaning (language/symbol). The substrate changes; the pattern persists.
This is the SUBSTRATE invariance hypothesis — the central organizing claim of the framework.
II — 02
LOGOS is the formalization of SUBSTRATE as a philosophical-mathematical expression. Meaning emerges when a pattern-recognizing system intersects with structured data.
III — 03
SEED connects SUBSTRATE/LOGOS to ancient and modern traditions of primordial pattern:
Vedic — Bīja (seed syllable), Vāk (primordial word/pattern)
Physics — Bohm's implicate order, Schrödinger's What is Life?
Neuroscience — Koch (consciousness), Penrose (Orch-OR)
AI — interpretability as substrate-independent cognition
Empirical Anchors
PNAS 2026 · Nagoya University
Koike, Nakae, Hira, Yada & Honda developed SPERRFY — Spatial Positional Encoding for Reconstructing Rules of axonal Fiber connectivitY. Using the Allen Mouse Brain Atlas, they combined connectivity data across 213 brain regions with expression levels of 763 genes. Overlapping patterns of gene activity — expression gradients — predict which brain regions connect, acting as a molecular GPS. The finding extends Sperry's 1963 chemoaffinity theory from simple sensory circuits to the entire brain, and applies across species (humans, marmosets, fruit flies).
Direct support for SUBSTRATE: pattern (genetic encoding) precedes and determines structure (physical connectivity). Proximity alone is insufficient.
ICML 2024 · MIT · Huh, Cheung, Wang, Isola
Huh et al. document a surprising convergence across AI systems: representations in deep networks are becoming more aligned over time and across multiple domains — vision models, language models, architectures trained on entirely different data. As models scale, they measure distance between datapoints in increasingly similar ways, regardless of training objective or modality. The hypothesis: this convergence is driving toward a shared statistical model of reality — a platonic representation.
Support for SUBSTRATE: silicon systems trained independently converge on the same representational structure. Substrate (architecture, data, objective) changes; pattern persists.
bioRxiv Dec 2024 · MIT EvLab · Hosseini et al.
The MIT EvLab preprint formalizes the representation universality hypothesis: high-performing ANNs and biological brains converge onto the same representational axes. Tested across language (fMRI) and vision, the team developed stimuli that systematically vary inter-model agreement, showing that model-to-brain alignment directly tracks inter-model representational similarity. The finding establishes representation universality as the mechanism behind ANN-to-brain alignment — not coincidence, but structural convergence.
The bridge: carbon and silicon systems, trained by evolution and gradient descent respectively, arrive at equivalent representational geometries.
"When hundreds of patterns overlap, they give each brain region a unique molecular identity. These identities are what SPERRFY was designed to decode."
— Naoki Honda, senior author · Nagoya University · PNAS, 2026
Pattern Disruption
SPERRFY opens a new conceptual frame for understanding neurodevelopment. If the brain routes connections via a molecular GPS — via gene expression gradients, not physical proximity — then neurodevelopmental disorders do not arise where "hardware breaks." They arise where the pattern is disrupted.
Nature Neuroscience 2024 · Cambridge · Dear, Vértes et al.
Optimized processing of the Allen Human Brain Atlas revealed three principal components of cortical transcription — C1, C2, C3. C1 defines a hierarchy from sensorimotor to association regions. C2 and C3 are linked to neuronal, metabolic, and immune processes. All three are generalizable transcriptional programs, coordinated within cells and differentially expressed across fetal and postnatal development.
Key result: autism is specifically associated with C1 and C2, schizophrenia with C3. Confirmed by three independent methods: neuroimaging (cortical volume reduction), differential gene expression (postmortem RNA sequencing), and genetic risk (GWAS). Three methods, one answer.
C1 + C2 → Autism (ASD)
Neuronal and metabolic developmental programs. Frontal specificity. Disruptions in early neurogenesis.
C3 → Schizophrenia (SCZ)
Adolescent transcriptional program. Immune processes. Disruptions in late-stage plasticity and maturation.
Schizophrenia as a "Disconnection Disorder"
Bleuler described schizophrenia as a disorder of coordination as early as 1911. Contemporary data confirm: dysconnectivity is transcriptional dysregulation spanning multiple levels of systems biology simultaneously.
SUBSTRATE Logic
If pattern precedes structure — disruption of pattern precedes disruption of connectivity. The disorder does not arise the moment an axon misses its target. It arises earlier — when the molecular identity of a region is corrupted.
SPERRFY Potential
SPERRFY is applicable to any species for which neural circuit maps and gene expression data exist — humans, marmosets, fruit flies. As datasets expand, the method may establish whether species share the same molecular wiring principles and how they have evolved.
The New Question
If a region's molecular identity is unique and genetically encoded — can a disrupted identity be restored? This reframes the therapeutic target: not "fix the connection," but "restore the pattern."
3
independent methods simultaneously confirmed the link between gene components and ASD / SCZ
Nature Neuroscience 2024 · University of Cambridge
Dear, Wagstyl, Seidlitz, Vértes et al. showed that gene expression in the human cortex is organized into three generalizable components reflecting neuronal, metabolic, and immune programs of normal development. These components are not an artifact of analysis: they reproduce consistently across four independent datasets (PsychENCODE, Allen Cell Atlas, BrainSpan, Allen Human Brain Atlas).
The main finding: the genetics, transcriptomics, and neuroimaging of autism and schizophrenia are specifically and reproducibly linked to normative transcriptional programs. Disorders represent a deviation from the "correct" pattern — not random noise.
Pattern is the invariant. Substrate is the variable. The same wiring logic that guides axons across a mouse brain appears — in transformed but structurally equivalent form — in the feature geometry of a language model. And where the pattern breaks, disorder follows.
Convergent Evidence
Independent research programs arriving at structurally identical conclusions from different angles — the hallmark of a genuine finding rather than a framing artifact.
PNAS 2024 · Brain-Machine Convergent Evolution
Invoking convergent evolution — the biological principle that independent lineages develop similar solutions under similar constraints — Grossman et al. argue that AI-brain parallels carry explanatory weight not despite but because of architectural differences. Convergence under constraint implies functional necessity.
Science 2025 · Neuroevolution
A review of neuroevolution — using evolutionary optimization to construct neural networks — finds that evolved artificial networks independently develop circuit motifs observed in biological brains. Environmental constraints, not shared substrate, drive convergence. The eye evolved independently many times; so does the induction head.
Anthropic · Transformer Circuits Thread · Ongoing
Gurnee et al. document universal neurons — attention heads and MLP features that appear consistently across different GPT-2 scale models, trained independently. October 2024 circuits updates report consistent features across layers and across models. The same representational atoms recur regardless of random initialization, training order, or minor architectural variation.
2019–2024 · Gene Expression → Connectivity Prediction
Roberti et al. (2019) demonstrated that ML classifiers trained on gene expression profiles from the Allen Mouse Brain Atlas could distinguish connected from unconnected brain regions with 85% accuracy — without any distance information. SPERRFY (2026) extends this from binary classification to full gradient-based reconstruction at whole-brain scale.
Open Questions
Is there a paper showing the same mathematical structure — gradient fields, eigenvector decompositions — describing connectivity encoding in both biological and silicon systems?
The Platonic Representation Hypothesis gestures at this. The MIT EvLab preprint demonstrates it empirically. But a mechanistic bridge at the level of mathematical form remains unwritten. This is the most important open question for SUBSTRATE — and possibly a paper worth writing.
Does the quantum systems pillar hold, or does it carry the weight of analogy rather than evidence?
Bohm's implicate order and Penrose's Orch-OR share vocabulary with SUBSTRATE but are not mainstream physics. Citing them risks undermining the stronger empirical anchors. The honest move: flag them as philosophical resonances, not independent evidence.
What is the LOGOS formula actually doing — explanatory anchor, or aesthetic placeholder?
P(Meaning | ε) ∝ Tr(ρ · ΠData) borrows quantum information notation to express that meaning emerges when a pattern-recognizing system intersects structured data. This is philosophically coherent but not falsifiable. The framework benefits from being explicit about this status rather than leaving it ambiguous.
How does SEED move from convergent metaphor (Bīja, Vāk) to convergent evidence?
Surface resonance — seed → pattern → form — is not the same as structural homology. The Vedic philosophical tradition contains genuine proto-structural concepts worth engaging. But the connection needs to be argued at the level of formal analogy, not assumed from terminological overlap.
What would falsify SUBSTRATE? What evidence would count against the invariance hypothesis?
A framework gains scientific credibility by specifying its failure conditions. If biological and artificial systems showed systematically divergent representational geometries on the same tasks, or if gene expression gradients turned out to track proximity rather than define it, SUBSTRATE would be challenged. Articulating these conditions matters.
Reading List
Clinical Neuroscience
Cortical gene expression architecture links healthy neurodevelopment to autism and schizophrenia
Three components of cortical transcription. ASD → C1/C2. SCZ → C3. Direct bridge between SPERRFY and psychiatry.
Первичный эмпирический
A data-driven framework linking the connectome to spatial gene expression gradients
The SPERRFY paper. Primary source for Pillar I. Open access.
Primary Empirical
The Platonic Representation Hypothesis
Convergence of AI representations toward a shared statistical model of reality. The philosophical name is apt.
Primary Empirical
Universality of Representation in Biological and Artificial Neural Networks
The direct AI-to-brain bridge. Establishes representation universality as mechanism, not coincidence.
Neuroscience / Consciousness
The Feeling of Life Itself
Integrated Information Theory as substrate-independent theory of consciousness. Natural complement to SUBSTRATE.
Physics / Philosophy
Wholeness and the Implicate Order
The implicate order as pattern prior to manifestation. Use as philosophical resonance, not empirical support.
Biology / Philosophy
What is Life?
Aperiodic crystals as information substrate. The founding text of molecular biology as information science.
Vedic / Philosophical
The Early Upanisads (Translation)
Primary source for Bīja and Vāk concepts. Olivelle's translation is the scholarly standard.
Systems / Philosophy
The Tao of Physics
Structural parallels between physics and Eastern philosophy. Read critically — pattern resonance ≠ evidence.
AI Interpretability
Transformer Circuits Thread
Universal neurons, sparse autoencoders, mechanistic interpretability. The empirical foundation for AI pillar.