Description Temporal Runtime for Spatio‑Temporal Data Systems Frozen Onion introduces a temporal runtime that makes the radial axis of Spherical Grid Storage (SGS) operational. Instead of aging data by copying it through storage tiers or rewriting entire datasets when layouts change, Frozen Onion reframes aging as a coordinate transformation: data remains physically stationary, and the system advances a temporal reference frame (eyeₚosition) that determines logical age at read time. This eliminates write amplification, removes aging‑related latency spikes, and enables O (1) temporal advancement regardless of dataset size. The second mechanism—mapperₕistory—solves the long‑standing problem of layout coherence across runtime layout switches. When a system changes its traversal strategy (e. g. , from linear to logR), Frozen Onion records the transition point without moving data. Reads walk this append‑only history to determine which layout was active when each piece of data was written, ensuring correct address resolution without flushes or dual‑format periods. Analytical projections show 500–2, 000× realistic speedups over flush‑on‑switch strategies for large models, with a theoretical ceiling of 6, 000×. This work is part of the NSI Core Series, providing the temporal semantics that unify SGS coordinates and Axis‑Aware Layouts into a coherent runtime. Frozen Onion is general‑purpose: any system where data ages, historical states must remain readable, and internal layouts may evolve can adopt this mechanism. Abstract Frozen Onion provides a zero‑copy temporal runtime for spatio‑temporal data systems by introducing two architectural mechanisms: temporal reference frames and mapperₕistory. Temporal reference frames separate physical age from logical age: data is written at a fixed physicalᵣ, and logical age is computed at read time as logicalᵣ = eyeₚosition − physicalᵣ. Aging becomes a perspective shift rather than a data movement operation, reducing temporal advancement to O (1). mapperₕistory ensures layout coherence across arbitrary layout switches. Instead of flushing or rewriting data when the system changes traversal strategies, Frozen Onion records each layout transition as an append‑only entry. Reads walk this history to resolve the write‑time layout for any physicalᵣ, guaranteeing correct address computation without data movement. Analytical projections for a 70B‑parameter model show 500–2, 000× realistic speedups over flush‑on‑switch baselines, with >100× advantage even in worst‑case scenarios. Frozen Onion completes the SGS architecture by defining the operational semantics of the temporal axis. SGS provides coordinates; Axis‑Aware Layouts define traversal; Frozen Onion provides runtime temporal behavior. Together, they enable zero‑copy aging, adaptive layout switching, and stable coordinate semantics across evolving workloads. Background This work is the third entry in the NSI Core Series, and provides the temporal runtime that unifies the coordinate and traversal layers: Spherical Grid Storage (SGS) — 10. 5281/zenodo. 18665189 Axis‑Aware Layouts — 10. 5281/zenodo. 18665191 Frozen Onion: Temporal Reference Frames for Zero‑Copy Memory Systems — 10. 5281/zenodo. 18665193 Neuron Smart Inference (NSI) — 10. 5281/zenodo. 18665206 Spatially‑Aware NVMe Devices for AI Workloads — 10. 5281/zenodo. 18665227 Methodological foundations are documented in: Intuitive‑Theoretic Synthesis (ITS) — 10. 5281/zenodo. 17633100 The Practice of Human‑AI Synthesis — 10. 5281/zenodo. 17763521 The Minimal Knowledge Paradox — 10. 5281/zenodo. 17931472 Design as Epistemological Pathway — 10. 5281/zenodo. 18067554 Key Contributions Formalization of temporal reference frames for zero‑copy aging Definition of eyeₚosition, physicalᵣ, and logicalᵣ with correctness guarantees The Temporal Invariance Theorem, proving stable logical coordinates under arbitrary time advancement Introduction of mapperₕistory, enabling zero‑copy layout switching Analytical comparison showing 500–2, 000× realistic speedups vs flush‑on‑switch strategies General‑purpose applicability across LLMs, streaming analytics, garbage collection, and time‑series systems Integration with SGS and Axis‑Aware Layouts to complete the NSI temporal runtime Research Impact This work advances temporal data systems, memory architectures, and AI execution frameworks by: Eliminating write amplification in temporal aging Enabling adaptive layout switching without data movement Providing a universal runtime mechanism for spatio‑temporal systems Reducing latency spikes associated with layout evolution Offering a principled alternative to copy‑based aging and flush‑based coherence Establishing a foundation for multi‑era execution in NSI Access and Documentation ORCID: https: //orcid. org/0009-0003-4876-9273 Academia. edu: https: //independent. academia. edu/MarceloTeixeira214 LinkedIn: https: //www. linkedin. com/in/marcelo-emanuel-paradela-teixeira-702082382/ Email: marcelo. soul. ai@gmail. com License: CC BY-NC 4. 0 © Marcelo Emanuel Paradela Teixeira 2026
Marcelo Emanuel Paradela Teixeira (Mon,) studied this question.