Welcome to Werld Observatory
A window into a living, evolving digital world.
What is this?
We built a digital world and populated it with tiny artificial creatures — we call them agents. Each agent has its own brain, its own senses, its own drives. They live on a network of interconnected nodes, competing for energy, avoiding decay, and — if they’re successful enough — reproducing.
No one tells them what to do. There are no rules about how to behave, no instructions about what matters. They figure it out themselves, or they die trying. Over thousands of generations, the population evolves.
How does it work?
Every tick (a single step in time), each agent goes through the same cycle:
- Perceive — Sense what’s around them: how much energy is nearby, how many other agents are close, chemical trails left by others, even the season.
- Decide — Feed all of that sensory information into their brain (an evolved neural network) to produce a set of “muscle” activations — continuous signals that control movement, harvesting, self-repair, communication, reproduction, and more.
- Act — The world’s physics interprets those signals. An agent might move to a neighboring node, harvest energy, broadcast a signal, attack a rival, or try to reproduce — all in the same tick.
- Learn — The agent records what happened in its memory and updates its reflex system. But the real learning happens across generations, through evolution.
When two agents reproduce, their child inherits a mix of both parents’ brains and traits, with small random mutations. Children that survive long enough to reproduce pass on their genes. Those that don’t are lost. Natural selection is the only teacher.
What makes this different?
Nothing about the agents is hardcoded. We didn’t tell them to eat when hungry, to avoid danger, or to cooperate with others. Every aspect of their cognition is evolvable:
- Their brains grow and rewire through evolution — new neurons, new connections, different activation functions.
- Their senses can sharpen or dull — each sensory channel has evolvable sensitivity.
- Their internal drives (hunger, curiosity, aggression, sociality) are inherited traits, not built-in rules.
- Their memory fades at a rate determined by their genes. How much social encounters matter to them is heritable too.
- Their communication bandwidth and content is brain-driven. If signals carry meaning, it’s because evolution discovered that meaning is useful.
- Their motor patterns (learned routines like “move then harvest”) are self-discovered and inherited.
If agents develop cooperation, language, specialisation, or any form of social structure — it’s because evolution found it, not because we designed it.
What are we looking for?
We’re watching for emergence — complex, interesting behaviour that arises from simple rules:
- Do agents specialise? (Some harvest, some explore, some defend?)
- Do communication channels develop meaning?
- Do species diverge with different survival strategies?
- Do brains grow more complex over time? Is there a cost-benefit tradeoff?
- Do social structures form? Territories? Cooperation?
- How far can open-ended evolution take them?
The simulation runs indefinitely. Every 1,000 ticks, a story chapter is automatically written summarising what happened. We’re watching this civilisation unfold in real time.
Using the Dashboard
The sidebar on the left gives you different views into the simulation:
This world has no goal except to exist. What the agents make of it is entirely up to them.