All OmniSciences publications, validation results, and our commitments to open science and responsible innovation.
The DeWitt supermetric on GL+(4)/SO(3,1) has signature (6,4). The structure group SO(6,4) has maximal compact subgroup SU(4)×SU(2)L×SU(2)R — the Pati-Salam gauge group. Two-step RG running gives sin²θW(MZ) = 0.2312, matching observation exactly. The negative-norm sector is identified with the Higgs bidoublet.
The Clifford algebra Cl(R6)⊗C of the positive-norm sector produces spinor C8 = S+⊕S− = 4⊕4̅ of Spin(6)≅SU(4). Combined with the negative-norm SU(2)L×SU(2)R, this yields exactly one Pati-Salam 16-plet per generation: (4,2,1)⊕(4̅,1,2). Hypercharge assignments follow from SU(3)×U(1)B−L branching.
The Gauss equation for the embedding X4 → Y14 yields the total curvature decomposition: RY = RX + gauge field strength + torsion. Einstein-Hilbert, Yang-Mills, and torsion terms emerge from pure metric bundle geometry with correct relative signs for the effective action.
All 13 triangle anomalies cancel in the Pati-Salam theory derived from the metric bundle. Mixed anomalies (SU(4)3, SU(2)3, gravitational, and U(1)B−L combinations) are verified computationally. Quantum consistency follows from the geometric Clifford algebra spectrum without requiring additional fermion families.
NG = 3 from the Spinc index theorem on K3 with U(1)B−L line bundle. The base manifold is uniquely K3 (Hitchin-Thorpe + Ricci-flat + Rokhlin). The index computes NG = (c12 − σ)/8 = (8+16)/8 = 3, with no free parameters.
The Born rule is the unique probability assignment preserving the Fisher-Rao metric on belief states. Fisher-Rao = Fubini-Study on quantum state space. Interference and complementarity emerge from blanket-mediated observations. Complex amplitudes are shown to be necessary and sufficient for bounded information processing.
The full effective Lagrangian with gauge couplings g4=gL=gR, Higgs bidoublet (1,2,2), and three fermion generations from K3. Key predictions: αPS = 27/(128π²) ≈ 0.0214 (7% from observed); sin²θW(MZ) = 0.2312 (exact match); Cabibbo angle sin(θC) = 1/√20 (0.75% error).
The Free Energy Principle applied to the metric bundle determines the gauge coupling αPS = 27/(128π²) from a variational principle. The soldering mechanism fixes Λbare with correct de Sitter sign. Localisation width from fibre Fisher-Rao geometry implies observable scale Lobs ~ 80 μm.
Independent implementation of multi-channel vibronic PCET rate theory validated against pyPCET with 0.1% agreement on BIP benchmarks. An initial 6.6× gating discrepancy is completely resolved (Vel = 1.75 vs 1.7 cm−1, 1.03×) through identification of three implementation-level errors. Monte Carlo UQ reveals 50–100% rate uncertainties.
Riemannian geometry on SPD covariance manifolds applied to climate forcing detection. All three major forcings (anthropogenic CO2, volcanic aerosol, solar) detected at p<0.01 on real observational data (HadCRUT5, GloSSAC).
Philosophical paper proposing that consciousness is the intrinsic nature of physical structure. Develops Structural Idealism as a position in philosophy of mind, connecting Hoffman's conscious agent framework with Friston's active inference via the formal functor Θ: Con → MB. Argues that the hard problem dissolves under structural idealism.
Technical paper establishing the formal correspondence Θ: Con → MB between Hoffman's conscious agent theory and Friston's Markov blanket formalism. Proves compositional preservation (Θ respects ⊗), conditional independence from P·D·A factorization, inverse mapping characterization, and connections to Tononi's Φ (integrated information).
Multi-scale geometric features from the DeWitt metric on SPD(3) applied to AIRSAR San Francisco quad-pol data. 96.73% overall accuracy with a single-parameter family of metrics. Few-shot regime: +9.3pp over Euclidean baseline at 5-shot.
96.73% overall accuracy on AIRSAR San Francisco benchmark. +9.3pp improvement in few-shot regime over Euclidean baseline.
76.4% classification accuracy vs 69.2% raw (+7.2pp) on Stanford HARDI benchmark. Patent #64/011,831.
+0.78pp improvement over pyRiemann baseline on 109 PhysioNet subjects (p=0.0048, Wilcoxon signed-rank).
0% SPD failures vs 63% Euclidean. Conditioning validated at p<0.0001 (Fisher exact test, 5 synthetic universes). Patent #64/007,419.
96.9% crystal system classification (vs 72.4% Euclidean) on 1,181 DFT-computed materials. 13,513× representation invariance improvement. 2,018 non-physical tensors flagged in Materials Project. Patent #64/004,047.
Reproducible results across all domains. All code and data publicly available.
We believe in honest science. Here's what we don't yet know.
OmniSciences is committed to open science, responsible innovation, and equitable access.
Perpetual, worldwide, royalty-free, irrevocable license for qualified users: academia, nonprofits, and government agencies. Non-commercial research and education always free.
No patent suits against universities, individual researchers, or nonprofits. Ever. If breached, patents automatically transfer to a public interest entity.
For neglected tropical diseases, food security, and environmental remediation: pricing capped at direct cost + 15%. Ethics Advisory Board can waive entirely.
≥10% of patent licensing + API revenue directed to a Public Benefit Fund supporting open science and education.
Open Research License survives acquisition, bankruptcy, or any change in ownership. Commitments cannot be revoked by future management.
All computational results backed by 166+ automated tests. Code, data, and methods publicly available. Every paper's figures can be reproduced from the repository.
@misc{omnisciences2026,
title={OmniSciences: Geometric Intelligence for Science},
author={Austermann, Sloan},
year={2026},
howpublished={\url{https://omnisciences.io}},
note={Open-source toolkit and research platform}
}