Genesis
Reality that is not
handed over in advance.
The laboratory's central research program: whether an artificial system can construct an operational understanding of reality without receiving a complete, predefined human ontology.
Genesis
in one place.
Genesis is the central research program of Mycellium Lab.
It investigates whether an artificial system can construct an operational understanding of reality without receiving a complete, predefined human ontology. Rather than beginning with fixed categories and asking a system to reason inside them, Genesis asks whether distinctions, relations, internal structures, and useful forms of representation can emerge through experience.
Genesis is not presented as a completed system or a demonstrated breakthrough. It is a research horizon being converted into testable hypotheses, measurable experiments, and explicit conditions of failure.
The reason
it was started.
Most artificial systems inherit the world already divided for them.
Objects, goals, labels, rewards, tasks, categories, and acceptable answers are normally defined in advance by human designers. The system may become highly capable inside those boundaries while never developing a different way of organizing what it encounters.
Genesis exists to investigate what happens when that inheritance is reduced.
What stands
in the way.
Current artificial intelligence systems can reproduce, combine, and reason across human-produced representations at extraordinary scale. This does not by itself establish that they can construct an operational reality of their own.
The research problem is difficult because a system can appear to discover a new category while merely compressing patterns already encoded in its data; an internal representation can be useful without being understandable in human language; novelty can be confused with noise; and any architecture, interface, reward, or dataset may silently prescribe the ontology the experiment claims to avoid.
How the work
is approached.
Genesis is being developed as a staged experimental program.
The immediate work is to define the first narrow hypothesis, the environment in which it can be tested, what counts as a prescribed human distinction, what observable behavior would indicate an independently useful representation, how recurrence, prediction, causality, generalization, and revision will be measured, and what result would weaken or overturn the hypothesis.
What it can
contribute.
Genesis is the program itself.
Every other project contributes only when it produces something Genesis can use without corrupting the experiment: a verified dynamic, a recoverable state transition, an uncertainty boundary, an episode structure, an environmental relation, a governance mechanism, or evidence that an assumed distinction was wrong.
The state
today.
Genesis is a research program, not a running system. There is no complete artificial system to evaluate yet.
The current work is operationalizing the first hypothesis: fixing the question, the environment, the measurement that decides it, and the conditions under which the result would not hold.
No ontology constructed by the system itself has been demonstrated. The other laboratory projects can contribute experiences and dynamics, but those contributions must not be imported into Genesis as universal truths.
- 01Type — Research program
- 02Status — In design