# LIVEX: born from a question

Step into the village. The square is warm, a woman sweeps in front of her door, a man climbs up from the well with two buckets, a guard walks along the perimeter wall. You leave for two hours on a quest. When you come back, the woman is still sweeping, the man is still carrying the same two buckets, the guard is at the same corner of the wall. Nothing has moved.

This is not an oversight by the developers. It is an architecture. Each inhabitant follows a behavior tree or a state machine: "if the player approaches, say hello; if night falls, sleep; if the supplies are empty, go to the well." The apparent richness comes from the number of cases planned, all written in advance. The village is not inhabited: it is performed.

What if we built a world where the score does not exist?

## A question, before any feature

> Can we build a virtual environment governed by fundamental laws, in which autonomous entities can produce behaviors and structures that were not explicitly programmed?

This question was asked before the first line of code. It still drives LIVEX's architecture choices today, as well as what it gives up. It comes from a curiosity about science fiction worlds and LitRPGs, those universes that feel truly alive because their inhabitants have needs, perceptions and an autonomy of their own. The goal is therefore not to write richer behaviors. It is to build an environment and laws solid enough to let those behaviors **emerge**.

Solid laws, we say. One is missing: verifiable.

## From intuition to experiment

It is easy to evoke emergence in a living room. "Look, a group has formed!" Such a sentence is worth nothing as long as it cannot be contradicted. To answer honestly, we have to move from intuition to a **testable setup**: a world whose rules are readable, whose behaviors were not written, and whose every event can be observed, measured, analyzed and **replayed**.

### Observe, then interpret

LIVEX carefully separates two things that are often confused.

What we observe, first: system-level properties (groupings, exchanges, specializations) that were not written as goals. After a few hundred ticks, for example, entities end up in the same place. That is a fact: it is in the stream, it can be counted, and it reproduces with an identical seed.

What we interpret, next: the hypotheses we formulate to explain that fact. "They gather there because thirst drives them to it, and some have learned that others come there." A hypothesis is only worth something if it can be disproved, by a replay, a counterfactual, or a measurement whose provenance is known.

In this frame, emergence is not magic. It designates systemic properties that are **traceable** back to the local interactions that produced them, and reproducible from one execution to the next with identical seeds. The word loses its prestige, and gains a meaning.

## Neither puppets nor black box

Two big families occupy the field, and LIVEX belongs to neither.

The first is **scripted**. Predefined behaviors, precise, readable, economical. Its limit is structural: a behavior tree encodes planned cases. It can combine many of them, but it invents none. A conduct for which no case was written remains out of its reach. Orchestrated illusion is precious for spectacle, and useless for the question asked here.

The second delegates cognition to **deep learning**. These are models capable of striking output, but probabilistic. Two executions with the same input can diverge, tracing back the causal chain of a decision becomes an open problem, and keeping consistency over thousands of ticks remains fragile. This is not a value judgment: it is a question of compatibility with what research demands.

LIVEX therefore chooses a third way: a **rule-explicit engine** where autonomous entities, with needs, memory, beliefs and capacities for action, interact in a persistent and partially observable environment. One example is enough to convey the principle. An agent is hungry. It spots a neighbor. It weighs eating, talking or resting. Nobody wrote this scene. It results from a utility formula, drifting thresholds, incomplete perception and a temperament. The intended result is not a spectacular demo: it is a **controlled experiment**.

## The price of honesty

Such an approach imposes constraints that cannot be seen from the audience. It requires verifiable determinism, instrumentation that does not modify the observed world, and versions that do not move silently. And it requires accepting an uncomfortable risk: that the experiment returns a negative verdict.

That risk was taken. In October 2026, LIVEX put two candidate phenomena to the test of a preregistered protocol: three arms, five seeds, fifty agents, one thousand ticks per run. The spatial quest toward a known deposit was **confirmed by contrast** on all five seeds. The homeostatic cycle of the social need, on the other hand, did not meet its conditions: verdict **not confirmed**. And the overall scientific verdict remained what the data imposed: "emergence not established."

A project that only wanted to convince would have swept the second result under the rug. LIVEX published it, because it is not a communication failure: it is the result. A method that cannot say "no" says nothing solid when it says "yes."

## The question stays open, and that is how it should be

This is the spirit in which LIVEX is born: not to produce a definitive answer, but to build a tool for asking the question under **controlled, honest and testable conditions**. A question that can be replayed, measured and discussed, rather than a question that is merely told.

What comes next is written slowly, at the pace of campaigns and measurements. If you want to follow the investigation, two doors are open: the [documentation portal](https://donovan-dev-web.github.io/LIVEX/documentation/), where the reference texts live, and the [project journal](https://donovan-dev-web.github.io/LIVEX/en/actualites/), also available as an [RSS feed](https://donovan-dev-web.github.io/LIVEX/actualites/feed.xml). Results appear there as they come, including the ones that are uncomfortable.

---

*This article is an English translation of the original French post, [LIVEX : naître d'une question](https://donovan-dev-web.github.io/LIVEX/actualites/livex-naissance-d-une-question/), published on the LIVEX journal on October 1, 2026. Project website: [LIVEX](https://donovan-dev-web.github.io/LIVEX/en/). Source code: [GitHub](https://github.com/donovan-dev-web/LIVEX).*
