Information–Energy dynamics and endogenous regulation in Artificial Systems (Loi E)

This repository contains three complementary experimental reports exploring the emergence and regulation of information–energy dynamics in artificial systems, within the framework of Loi E (Information–Energy Law).

The work is entirely empirical and observational. It does not rely on learning, optimization, supervision, external grounding, or goal-driven mechanisms. All reported effects emerge from minimal algorithmic rules and explicit information–energy accounting.

The three reports are:

1. Information-Energy Dynamics Simulator (Loi E) — Experimental Observations
   A systematic exploration of information–energy regimes in a minimal simulator.
   The document reports observations across multiple experimental questions (Q1–Q10), including non-uniform dissipation, regime recurrence, distributed asymmetry, recoverability windows, and irreversibility thresholds.
   Detailed observations are provided in annexes, together with selected audit-level traces.

2. Regulator–Selector (Reg-Sel): Minimal Endogenous Regulation for Large Language Models
   An experimental study of a lightweight, inference-time regulation mechanism applied to large language models.
   Reg-Sel operates without modifying model weights, prompt engineering, or external supervision.
   Results show consistent improvements in response stability and truthfulness, with bounded computational overhead, across multiple architectures.

3. Filtre E — Audit-Level Information-Energy Regulation
   A low-level regulation mechanism based on explicit information–energy thresholds.
   Filtre E governs response generation by constraining cumulative informational cost, producing more stable and bounded outputs under uncertainty.
   The report documents audit traces illustrating regulated versus non-regulated regimes.

Together, these documents establish a coherent experimental foundation for studying information–energy dynamics and endogenous regulation in artificial systems.
Interpretative discussions (e.g. negentropy, biological analogy, governance implications) are intentionally separated from these reports and addressed in subsequent works.

The experiments were conducted by the author.
An experimental user interface is available to the scientific community upon request for independent exploration and replication.