About Invarians

From soil texture to blockchain substrate.

BACKGROUND

I am a geologist based in France, specializing in geotechnical engineering, with a strong affinity for computer science, electro-hydraulic systems, and machine learning.

Geotechnical engineering involves measuring soil properties to dimension structural foundations. Two requirements are central to this field: measurement reliability and data veracity.

With the rise of automated systems and artificial intelligence, a critical question arises in my practice today:

  • Was the data genuinely produced in the field?
  • Or is it the result of construction, smoothing, or artificial generation?

This issue led me to take a close interest in blockchains, not as financial objects, but as technical substrates capable of preserving the continuity and provenance of signals.

ANALOGY

From Drilling to Digital Signal

In geotechnical engineering, engineers use well logs (diagraphies): continuous recordings of physical signals (advancement, rotation speed/torque, pressure, drill head energy) captured without immediate interpretation.

These logs do not describe a decision. They describe a texture: the texture of the soil. They provide a structural context, indispensable prior to any interpretation or decision-making.

A Question of Substrate

By professional reflex, I began searching for the equivalent of this texture within blockchains. Not performance indicators, but structural invariants.

What is the structural substrate of the blockchain upon which developers or AI agents operate?

Existing tools mainly describe traffic, performance, or congestion. They say little about the continuity, density, and deep structure of the on-chain signal over time.

INSTRUMENT

Invarians

I created Invarians as a measuring instrument, comparable to a drilling parameter recorder.

Invarians:

  • does not analyze,
  • does not interpret,
  • does not predict.

Like a well log, it captures public on-chain signals, compresses them, and preserves them in the form of deterministic structures.

These structures, called InKe (Invariant Kernels), condense tens of millions of signals and make the texture of a blockchain readable at a given moment. Cumulatively, Invarians captures billions of signals, encapsulated without interpretation, compressed in their raw state, and made available.

CONCEPT

A Simple Analogy

On a road, Invarians would describe neither vehicle speed, nor traffic, nor congestion.

It would indicate solely if the substrate is:

  • continuous (road passable or cut),
  • fragmented (potholes),
  • stable (expressway),
  • or altered (gravel, degradation).

It is then up to the system designer, for example of an AI agent, to decide on its behavior: should it advance quickly, slow down, or adapt its strategy?

An agent deprived of structural context, of texture, may adopt erratic behaviors or those one might qualify as hallucinatory.

METHODOLOGY

Methodological Principles

InKe rely on three non-negotiable invariants:

Deterministic Production

From the same public on-chain data, the same InKe is produced, without variance.

Absence of Human Interpretation

No signal is weighted, qualified, or manually corrected. There is neither subjective threshold nor a posteriori adjustment.

External Verifiability

Any actor can recalculate, compare, and analyze the relative drift of two InKes over time.

Invarians does not ask for trust. It enables verification.

SCOPE

Scope

Invarians currently captures signals on several major blockchains, including Ethereum, Solana, and Polygon.

InKes are available for any use: AI agent instrumentation, research, structural comparison, or independent exploration.

I hope this instrument will be useful to you.