Research / Methodology

Simulation Methodology

Scientific guidelines, retrieval algorithms, and checking matrices governing the alternate timeline generation.

1. Knowledge Retrieval Layer (RAG)

Every simulation begins with dynamic context loading. Before any agent generates a single word, our retrieval system executes asynchronous searches across:

Wikipedia Encyclopedia Loader

Fetches comprehensive summaries of key historical events, names, dates, and baseline policy records matching the query divergence point.

arXiv Academic Abstract Loader

Retrieves technical papers regarding carbon capture, monetary models, battery chemistries, and computing architectures to anchor calculations in real science.

2. Parallel Domain Collaboration

Agents process in parallel, but their prompts contain dependencies. The Historian establishes the core pivot year and milestones. In the next step, the Economist, Technology, Society, and Climate agents evaluate the secondary consequences of those milestones.

Agent Flow Sequence
Historian Pivot
Parallel Domain Analyses (Climate, Econ, Tech, Society)
Critic Checking Cycle

3. Critic Validation & Plausibility Scoring

The Critic Agent evaluates internal consistency. It checks whether the climate outcomes are consistent with the technological changes, and whether the economic projections are realistic given the social developments. The confidence score is computed based on these comparisons:

• Extreme Projection Check: If an agent outputs an impact score > 75, Plausibility drops by 20%.
• Domain Divergence Check: If there is a massive delta between two indices, Plausibility drops by 35% and warning flags are raised.