Prime-Agent x402 Intelligence MCP server

GENESIS v4.0

Generative Evolving Neural Engine for Self-Improving Systems

GENESIS is an experimental multi-agent evolution and learning simulation. The current genesis.py implementation combines neuroevolution, online adaptation, intrinsic-motivation signals, communication, counterfactual evaluation, shared knowledge, hierarchical goals, social prediction, concept formation, and self-narrative mechanisms in one self-contained Python program.

Project status: research prototype. GENESIS explores mechanisms associated with adaptive and self-improving systems; it is not evidence of artificial general intelligence or a formally verified Gödel machine.

Current capabilities

The v4 implementation contains 17 major capability areas:

What changed in v4.0

v4 adds four capability families on top of the v3 communication, counterfactual, and shared-knowledge systems:

• Hierarchical goal formation — agents can maintain goals and decompose them into smaller objectives.
• Social prediction — agents model aspects of other agents' behavior to influence cooperation and competition.
• Abstract concept formation — repeated experiences can be compressed into higher-level prototypes.
• Self-narrative — agents maintain a compact history that can affect later decisions and inheritance.

Requirements

• Python 3.11 or 3.12 recommended
• NumPy
• Matplotlib

Install dependencies from the repository manifest:

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt

On Windows PowerShell, activate the environment with:

.venv\Scripts\Activate.ps1

Run

python genesis.py

The default configuration currently runs 150 generations with 100 simulation steps per generation, so a full run is intentionally more substantial than a smoke test.

Verify core behavior

Run the deterministic fast test suite without starting the full simulation:

python -m unittest discover -s tests -p "test_*.py" -v

The current core tests verify innovation-ID stability, minimal genome topology, finite bounded network activation, structural independence after genome copying, and learning-rule weight bounds.

Output

Generated plots are written under genesis_output/. The v4 visualization paths currently include:

• genesisoutput/genesisv4_dashboard.png
• genesisoutput/genesisv4_universe.png

Shortened. The full README is on GitHub.