Preprints Pipeline
Research roadmap
What has been published, what is in progress, and what is planned across the six domains of inquiry. Priorities and timing are subject to review as evidence accumulates.
Published
Three preprints released to date. Details, methodology, and PDFs are available on the Publications page.
Beyond Synthetic Benchmarks: Reproducible Evaluation of Engineering-Intelligence Systems for Semiconductor Design
A reproducible, anti-circular, provenance-aware evaluation framework applied to real open-source RTL. Introduces XRecall — a provenance-anchored cross-project engineering-memory architecture.
The AI Demand Shock: A Falsification Test of AI-Era Demand Compression
A falsification framework tested against lifecycle data from 57 consumer products (2016–2026). Finds the AI-compression hypothesis is not supported: compression tracks product type, not era.
Beyond Silicon: A First-Principles Framework for Materials Required by Future Intelligence Systems
A materials-science framework for future intelligence systems (2035–2050). Introduces the XSYDA Data-Movement Principle and a five-tier Unified Energy Hierarchy. Recommends a silicon-anchored, function-partitioned heterogeneous architecture.
In Progress
Preprints currently under development. Working titles and scope may evolve; publication dates are targets, not commitments.
Provenance in Enterprise Knowledge Work: A Reproducibility Framework for Non-Scientific Domains
Extends the provenance discipline developed for semiconductor evaluation to enterprise decisions, procurement analyses, and supply-chain forecasts — where reproducibility is rarely tested but consequential.
Adaptive Computing Under Real Constraints: An Evaluation on Non-Synthetic Optimization Corpora
A reproducible study of adaptive-computing approaches on constrained enterprise optimization workloads. Draws on the anti-circular evaluation protocol established in Preprint 001.
Planned
Studies in the near-to-medium pipeline, drawn from the six research domains. Sequencing depends on data availability and independent-ground-truth construction.
Signature-Anchored Authenticity for Engineering Artifacts
Provenance and authenticity mechanisms for engineering deliverables, extending the SHA-256 signature model already used across the artifacts framework.
Reusable Computation Kernels with Deterministic Provenance
A study of signature-anchored kernel reuse under deterministic-execution constraints, with reproducibility benchmarks against conventional caching approaches.
Retrieval-Quality Benchmarks for Cross-Project Engineering Memory
Independent evaluation harness for cross-project engineering-memory retrieval, with fine-grained failure taxonomies validated against maintainer histories.