EDEN / RESEARCH 01 AUTONOMOUS SYSTEMS LABORATORY

PROJECT IN-A-GADDA-DA-VIDA

Build a world.
Then let it think.

A living simulation for studying autonomous agents, artificial memory and the boundary between a decision and a change to reality.

Enter the lab Explore the research
RESEARCH PROTOCOLS09
SIMULATED WORLD60 × 60
AUTONOMOUS AGENTS02

01 / THE RESEARCH QUESTION

A world first.
An agent second.

Adam and Eve perceive, decide, use tools, exchange information and remember between sessions. Python remains the authority over reality; the language model can only propose an action.

01Deterministic coreRules stay testable without AI.
02Visible decisionsEvery important event is traceable.
03Safe autonomyTools validate every mutation.

02 / EXPERIMENT SEQUENCE

From deterministic rules
to measurable behaviour.

Each stage adds one controlled capability. The simulation remains testable before intelligence, persistence or communication is introduced.

  1. 01
    Establish realityWorld state, rules and validated events
  2. 02
    Make it observableInterface, scheduling and clear boundaries
  3. 03
    Add bounded agencyLocal model proposals and safe tools
  4. 04
    Retain and exchangeMemory, provenance and agent messages
  5. 05
    Measure the systemContracts, baselines and experiments

03 / SYSTEM ANATOMY

Separate responsibilities.
One consistent world.

The model proposes. Tools constrain. The simulation engine alone confirms a world mutation.

01 / PRESENTATIONHuman input and observation
Tkinter + TurtleRequests actions and renders confirmed state
intent / event
Simulation EngineValidates rules and owns mutations
02 / DECISION SUPPORTInterpretation without authority
Agent LoopObserves, decides and evaluates results
prompt / proposal
LiteLLM + OllamaReturns an untrusted proposal
03 / AUTHORITY PATHValidated changes to the world
Agent LoopRequests one named capability
tool call
Tool RegistryChecks schema and permission
validated
Simulation EngineAccepts or rejects the request
world event
World StateStores the confirmed result

04 / FIELD MANUAL

Nine protocols.
One living system.

Move from the first world rule to a measurable society of agents. Each protocol pairs the theory with mechanisms, experiments and a checkpoint.

PROTOCOL 01DOMAIN

World, Rules and Responsibilities
PROTOCOL 02PRESENTATION

Tkinter and Turtle: The Garden Interface
PROTOCOL 03RUNTIME

Async, Events and Concurrency
PROTOCOL 04INTELLIGENCE

Local AI with Ollama and LiteLLM
PROTOCOL 05AGENCY

The First Agent, Tools and Safe Execution
PROTOCOL 06MEMORY

Memory, Recall and SQLite Persistence
PROTOCOL 07MULTI-AGENT

Communication Between Agents
PROTOCOL 08QUALITY

Contracts, Standards and Reliability
PROTOCOL 09RESEARCH

Artificial-Life Experiments