Physical
Embodiment, movement limits, obstacles, resources, and consequences.
PROTOTYPE FOUNDATIONARTIFICIAL LIFE · EMBODIED AGENTS · OPEN QUESTIONS
Intelligence has
an environment.
What might develop when an agent has a body, a world, and a history?
We build virtual habitats to explore how constraints shape behavior. Beginning in Blender. Growing through experiments.
Explore the research ↗RUNNING/HABITAT 001 — THE COMMON GROUND
BROWSER EXPERIMENT3D rendering is unavailable in this browser. The research and identity pages remain available.
FIVE AGENTS. ONE SHARED WORLD.
Different tendencies.
Continuous negotiation.
LOCAL SPACE
T + 0000.0 s
T + 0000.0 s
selected agent movement trails
OBSERVATION ONLY
Watch their paths. Change your perspective.
Looking around · interest 0%Agent 1 · Spring0.00 units / s · Exploring
Motion is reduced according to your device preference. Camera controls remain available.
A rule-based browser study adapted from our Blender prototype. Steering, ramp climbs, jumps, and short authored conversations; not evidence of emotions, learning, or consciousness.
01 / THE RESEARCH QUESTION
Not just what an agent does.
What shapes what it becomes.
ASimulation explores a developmental approach to artificial agents: place them in a world, give them limited capabilities and competing drives, and study how behavior changes through interaction.
Our starting point is an embodied sandbox in Blender. The longer-term ambition is an ecology of agents with different histories, constraints, and ways of responding to their environment.
Consciousness is an open research question—not an achieved property of these systems. Observable behavior and adaptation are what we can investigate today.
02 / CONDITIONS, NOT CHARACTER SCRIPTS
A world of influences.
Cognitive
Attention, familiarity, uncertainty, curiosity, and bounded memory.
EARLY HEURISTIC MODELSSocial
Personal space now. Cooperation, competition, and shared conventions as future experiments.
EXPLORATORYAffective
Modeled internal drives and response weights. Emotional language describes a model, not felt experience.
MODELING DIRECTIONMeaning & values
How might priorities, identity, and meaning-like structures arise? A place for philosophical and spiritual questions.
OPEN QUESTION03 / FROM THE LAB
Built in the open.
One experiment at a time.
Follow the development of the Blender sandbox and the ideas behind it.
Embodied navigation in Blender
Geometric sensing, smooth acceleration and braking, personal space, and experience-dependent response. Current local code; broader generalization remains untested.
From individual agents to ecologies
Persistent histories, geometric memory, and specialist cooperation. A proposed architecture to test, not a completed artificial-life system.
04 / RESEARCH DIRECTION · CONCEPT STAGE
LDRM
Linguistic Dimensional
Reasoning Model
Could a language model develop through structured reasoning and interaction, without pretraining on a large text corpus?
LDRM explores this question as a direction toward a zero-data language model. The proposed investigation is to represent concepts and their relationships across explicit dimensions, and study how those relationships could support meaning, composition, and reasoning.
Within ASimulation, a possible testbed is the agent’s world: objects, actions, spatial relationships, and consequences. Early experiments would ask whether a small system can connect symbols to experience and combine known relationships in unfamiliar situations.
“Zero-data” is a research ambition, not a demonstrated capability. Here it means exploring an alternative to corpus pretraining—not an absence of information: designed rules, representations, and interaction all provide structure or evidence. No functioning LDRM language model is claimed or running in this website.
05 / START A CONVERSATION
Something
in mind?
A research question. An observation.
A possible collaboration.
METAVERSAL ARTS · VIENNA, AUSTRIA
YOUR NEXT EXPLORATION
GOOD WORK STARTS WITH A CONVERSATION.Privacy