Modern artificial intelligence systems operate at an unprecedented scale of computation.

Despite remarkable advances in performance, these systems remain largely governed by abstract optimization objectives, detached from explicit physical constraints.

Computation is treated as a purely informational process, while its energetic cost, irreversibility, and long-term stability are often considered secondary or external concerns.

This separation introduces structural fragilities: escalating energy consumption, opaque failure modes, and a lack of intrinsic criteria for durability and coherence at scale.

Neomundi Lab adopts a different perspective.

Information processing is not an abstract operation detached from the physical world. It is a thermodynamic process, inherently coupled to energy dissipation, irreversibility, and entropy production.

From this viewpoint, artificial intelligence systems are not only computational artifacts, but physical systems whose behavior unfolds within energetic constraints.

Governing such systems therefore requires more than optimization: it requires an explicit framework capable of linking information, computation, and energy within a single coherent model.

Law E is a thermodynamic–informational framework developed to formalize this coupling between energy, information, and computation.

Rather than prescribing specific algorithms or architectures, Law E defines a set of minimal constraints under which computational systems can be observed, compared, and regulated.

The framework focuses on three fundamental dimensions:

• Energetic efficiency

• Irreversibility of informational processes

• Structural stability over time

Law E does not aim to maximize performance. It aims to characterize the conditions under which complex systems can remain coherent, efficient, and durable as they scale.

This framework is not a model of intelligence, nor a theory of cognition.

It does not attempt to define meaning, consciousness, or learning strategies.

Instead, Law E provides a thermodynamic lens through which artificial intelligence and large-scale computational infrastructures can be evaluated and governed as physical systems.

Experimental investigations, simulations, and applied implementations derived from this framework are documented separately.