LOWJOULE / PUBLICATIONS

Research should be
open to scrutiny.

Working papers, mathematical foundations, experimental protocols and technical notes from the LowJoule research program.

01 / WORKING PAPERS
LJ-WP-001
WORKING PAPER · VERSION 1.0

LowJoule AI:
Rethinking Artificial Intelligence Through Energy Efficiency

An initial statement of the LowJoule research program, introducing energy-aware computation, adaptive reasoning, Useful Intelligence per Joule and the hypothesis that computational resources can be allocated dynamically according to expected marginal utility.

AuthorEmmanuel Guerevitz
Version1.0
Year2026
StatusWorking Paper
IdentifierLJ-WP-001
DOINot assigned
PDF · PREPARING PUBLIC RELEASE
02 / FOUNDATIONAL NOTES

Formal framework.

LJ-TN-001
TECHNICAL NOTE · V0.1

Mathematical Foundations of Energy-Aware Computation

Formal definitions for energy complexity, marginal computational utility, adaptive policies and optimal stopping.

Read mathematical framework →
LJ-TN-002
TECHNICAL NOTE · V0.1

Useful Intelligence per Joule

Proposed measurement framework for evaluating useful AI performance relative to physical energy consumption.

Read UI/J framework →
LJ-RP-001
RESEARCH PROGRAM · V0.1

LowJoule Experimental Roadmap

Initial falsifiable experiments covering task heterogeneity, marginal utility, adaptive stopping and measured energy dominance.

Read research program →
03 / PUBLICATION PRINCIPLES

Versioned. Reproducible. Correctable.

01

Versioned

Research documents receive explicit versions so changes in definitions, assumptions and conclusions remain traceable.

02

Reproducible

Experimental publications should describe workloads, hardware, software, energy measurement and evaluation methodology.

03

Falsifiable

Hypotheses should be connected to measurable predictions and predefined acceptance criteria.

04

Correctable

Negative results, revised assumptions and methodological limitations remain part of the scientific record.

04 / CITATION POLICY

Persistent research identifiers.

LowJoule documents use internal persistent identifiers before any external archival identifier is assigned.

LJ-WPWorking Paper
LJ-TNTechnical Note
LJ-RPResearch Program
LJ-EXPExperiment
PROVISIONAL CITATIONGuerevitz, E. (2026). LowJoule AI: Rethinking Artificial Intelligence Through Energy Efficiency. LowJoule Working Paper LJ-WP-001, Version 1.0.
05 / FUTURE ARCHIVAL RELEASES

From working paper to research record.

Future public releases may include archival repositories, machine-readable citations, source material, datasets, experiment configurations and persistent external identifiers where appropriate.