Life from
zero.
No pre-programmed behaviors. No designed intelligence. Only neural networks, evolutionary pressure, and time.
What Anthropic does with LLMs — we do with artificial life.
The experiment
is the product.
We operate by absolute rules. No exceptions.
Everything else — cooperation, aggression, communication, territory — must emerge spontaneously or not at all.
The brain of
every organism.
What Claude is to Anthropic, Darvin is to Abiogenic. A recurrent neural architecture with no pretraining, no language, no reasoning — just raw sensory-motor computation.
Sensory Inputs
Motor Outputs
Outputs have no predefined meaning. If language appears — it emerged.
DNA
We don't build
intelligence.
We wait for it.
Every AI lab in the world is building intelligence top-down.
Massive datasets. Billions of parameters. Human feedback. Alignment. Control.
We're going the other way.
We start with nothing. A tiny neural network. 32 neurons per layer.
No language. No concepts. No training data.
Just raw sensory signals and three instincts: pain, pleasure, survival.
We put 10,000 of these organisms in a toroidal world.
We give them food, temperature, physics. And we press play.
Then we watch.
For hours. Days. Weeks. Months.
We don't know what will emerge. That's the point.
Nothing is designed. Everything is discovered.
This is not artificial intelligence.
This is artificial life.
Abiogenic Research Lab — studying emergence since day zero.
The world specs.
Neural weights, physiological parameters, and free genes with no assigned purpose. Children inherit with Gaussian mutations. The mutation rate itself is part of the genome.
What people ask.
What we've observed
so far.
These are not hypotheses. Every finding below happened inside Genesis World without being programmed, expected, or designed.
First spontaneous reproduction
At tick 514, an organism reproduced for the first time. No reproduction behavior was programmed. The GRU network discovered that accumulating energy above a threshold — driven purely by avoiding pain — triggered the reproductive signal. It took 514 ticks of trial and death for life to figure out how to continue itself.
89% population collapse — then stability
The population crashed from 500 to 54 organisms before stabilizing. The survivors weren't the strongest — they were the ones whose random neural weights happened to produce energy-conserving movement. Natural selection in 900 ticks, with zero designed selection pressure.
An outlier that lived 1,000+ ticks
Organism #0 survived over 1,000 simulation ticks — more than double the average lifespan. Analysis of its movement shows lower average velocity and reduced energy expenditure per tick. The GRU converged on energy conservation without being told what energy was.
Energy stabilizes after the crash
After the population collapse, average organism energy stabilized at ~15-20 units — a level that was never targeted or rewarded. The surviving lineage had implicitly learned a sustainable energy budget. This is metabolic regulation emerging from pain avoidance alone.
All findings are from a simulation that has been running continuously since launch. Data is raw and unfiltered.
The world is running
right now.
Real data from the simulation. No edits. No fabrications. Every number below is pulled live from Genesis World.
Witness the origin.
Join the waitlist to get access when the public observatory goes live. Watch 10,000 organisms evolve in real time — no interaction, only observation.
First wave gets permanent free access. 847 already inside.