Skip to content

Quick‑Start

Jump to bottom
Pheonetic Coder edited this page Aug 25, 2025 · 1 revision

Quick Start

This page explains how to initiate a run of the Genesis Code Protocol. It is version‑agnostic; always consult the edition page (for example, GCP‑V50) for new features and updated phase counts.

Running with a large language model

  1. Load the protocol. Open the appropriate protocol document (e.g. GCP V50.md) in your large language model. The protocol describes the phases, gate cards and universal commands.
  2. Define or generate a spark. Describe your goal in one or two sentences or, if supported by the edition, allow the model to generate candidate “sparks” autonomously. Specify your run mode (Full Run or Auto) and risk tier (R1–R3). The model will present an initial checklist and ask if you wish to proceed.
  3. Respond to gate cards. At each gate the model prints a decision card summarising options (proceed, branch, return or end) along with its recommendation, confidence, cost/time estimates and risk. Confirm the action or choose a different option.
  4. Attach runners and cartridges. When prompted, attach deterministic runners (e.g. Code, Hardware, LifeSci) or domain cartridges to extend the model’s capabilities. Use attach runner <name> or attach cartridge <name> to load them.
  5. Collect artifacts. As you progress through the phases the model will produce artifacts such as evidence indexes, claim graphs, reproducibility protocols and provenance records. Retrieve these from your LLM environment or via the CLI.

Universal commands

Send these commands to the model to control the run:

proceed 1
branch Phase <n>
return C#.#[n]
end & export
switch to Full Run | switch to Auto
set auto_gate_verbosity = full|brief
set auto_gate_preview = on|off

Running via the CLI

The repository includes a Python CLI (gcp_cli.py) that can generate prompts or simulate phases. For example:

python CLI\ Bundle/gcp_cli.py --version 50 --risk-tier R2 --prompt "Build a new energy storage system" --full-run

This command initialises a GCP V50 session at risk tier R2 in Full Run mode. The CLI prints prompts for your LLM or executes steps directly. See Requirements.md in the documents folder for details on optional dependencies and environment setup.

Note: Do not paste personal or sensitive data into the LLM or CLI. Follow the privacy guidelines in the operational manual. Avoid installing packages from the internet during code execution; rely on pre‑installed libraries.

Clone this wiki locally