CIVI-GENESIS

CIVI-GENESIS is an AI-powered synthetic society simulator that lets you safely “test drive” policies and business ideas on a virtual city before applying them in the real world. My system generates up to 50,000 diverse virtual citizens with demographics, income levels, jobs, locations and personalities. Each citizen has traits like risk tolerance, openness to change and political view, so different groups react differently to the same decision. When a user describes a new policy or product change in natural language, CIVI-GENESIS simulates how this synthetic society evolves over multiple time steps. I use Google Gemini to deeply simulate a sample of citizens each step: Gemini reasons from their point of view, updates their happiness, support for the policy and income, and writes short diary-style entries that show how their life is affected. These LLM-generated reactions become “teacher data” for a small neural network that I train during the run. The neural network learns to approximate Gemini’s behavior and can then scale the same kind of reactions to tens of thousands of citizens in real time. This hybrid LLM + NN design combines the richness of Gemini with the speed of a compact model. The result is an interactive Streamlit dashboard that feels like a control room for a digital city. Users can explore how average happiness, policy support and income change over time, and break the results down by income level and city zone. They can inspect individual citizens and read their diaries, view inequality metrics, and get Gemini-generated expert summaries from the perspective of an economist, a social activist and a small business owner. CIVI-GENESIS is not a prediction engine, but a powerful thinking tool to reveal blind spots, compare scenarios side by side, and design policies and products in a more transparent and responsible way.