Three papers are live on Zenodo, each with a DOI, and every result in them was re-run on the dates recorded inside. They read in order: the foundation, the core, then the application.
Independent execution pipelines of one language agree bit-for-bit on chaotic workloads
Two separate ways of running the same Sanguis program produce byte-identical output, on workloads where a difference of one bit would explode. This is the floor the rest of the work stands on.
A pair-primary associative substrate with exact in-band episodic memory
The substrate itself: memory whose unit is a relation, reading back what it has lived exactly, with every condition stated in the paper.
A pair-primary associative substrate with route-unique composition
Composition over the substrate: each route seeded by its own pair, so composed output is pair-specific and verifiably free of leaks.
Published under CC BY-NC-ND 4.0.
This statement appears in all three papers, in the same words:
"The system and efforts producing the results of this document are the author's design, built and run over the past eight months. AI assistants contributed throughout: initial drafts of the prose, insight, suggested sources, and supervised code execution to detect and patch flaws during the experiments. Every cited source was independently checked before use, and all programs were verified running in a closed environment without AI assistance. Every result reproduces from the scripts and dates recorded in the paper."
The scripts behind the results are not included in the preprints. They release with the project.