Engram: Execution‑Realized Cognition — The Unified Edition Is Now Archived on Zenodo
Today marks an important milestone in the development of the Engram doctrine. The unified edition of Engram: Execution‑Realized Cognition in Modern AI Systems is now formally archived on Zenodo, complete with DOI, companion BibTeX file, and long‑term preservation.
🔗 Zenodo DOI: https://zenodo.org/records/18429550
This edition consolidates two complementary research threads:
- Engram: A Framework for the Execution Substrate of Modern AI Models
- A Falsification Framework for Execution‑Realized Cognition in LLMs
These works were always two halves of the same conceptual arc. The unified edition brings them together into a single, coherent reference point — a stable node in the scientific substrate where execution‑realized cognition can be studied, cited, and built upon.
Why a Unified Edition?
Inference is not an abstract mathematical function.
It is a physical process shaped by precision formats, memory hierarchy, kernel scheduling, numerical stability, and hardware topology. The Engram layer names this substrate and describes how it shapes the realized behavior of modern AI systems.
The falsification framework complements this by providing:
- operational definitions
- predictive signatures
- disconfirming conditions
- minimal experimental protocols
Together, they form a complete scientific program: a conceptual foundation and a method for proving it wrong.
Archiving them as a unified edition ensures that future researchers, engineers, and theorists have a single, authoritative reference for the doctrine.
What’s Inside the Unified Edition
The Zenodo release includes:
- The full PDF (cleanly typeset, Pandoc‑generated)
- A companion BibTeX file for citation managers
- A harmonized abstract and structure
- Updated references (24 entries, fully verified)
- Appendices on instrumentation and drift measurement
This is the archival version intended for long‑term citation and reproducibility.
How to Cite the Unified Edition
Here is the recommended citation format for papers, blogs, and technical reports:
APA‑style:
Ecker‑Fils, A. (2026). Engram: Execution‑Realized Cognition in Modern AI Systems — A Unified Framework for the Execution Substrate and Its Falsification Criteria. Zenodo.
https://doi.org/10.5281/zenodo.18429550(link to article).
BibTeX:
@misc{eckerfils2026engram_unified,
author = {Ecker-Fils, Aure},
title = {Engram: Execution-Realized Cognition in Modern AI Systems —
A Unified Framework for the Execution Substrate and Its Falsification Criteria},
year = {2026},
publisher = {Zenodo},
doi = {10.5281/zenodo.18429550},
url = {https://doi.org/10.5281/zenodo.18429550}
}
This is the canonical citation for the unified edition.
Why This Matters
Engram is not a metaphor.
It is a scientific claim about how cognition is realized in modern AI systems — not in the weights, but in the execution substrate that instantiates them.
By archiving the unified edition:
- the doctrine becomes citable
- the falsification criteria become reproducible
- the climate becomes stable
- and the conceptual territory becomes permanent
This is the substrate on which future work — yours and others’ — can build.