Align docs around LLM-driven scenario generation

docs/social-post.md

@@ -4,9 +4,9 @@
 
 I built LifeChoice Simulator for the Build Small Hackathon.
 
-It is not a chatbot. It is a small causal future simulator: enter a difficult fork in the road, choose one path to live through, and watch eight decisions change money, fulfillment, validation, stress, family pressure, the world state, and the final report.
+It is not a chatbot. It is a small LLM-powered causal future simulator: enter a difficult fork in the road, choose one path to live through, and watch model-generated scenarios interact with deterministic state across money, fulfillment, validation, stress, family pressure, the world state, and the final report.
 
-The first scene is deterministic for instant startup. Later nodes can be enriched by a 7B model, but the simulation engine owns the state, consequences, safety rules, and final report.
+The 7B model drives the adaptive scenario experience. The simulation engine gives it structure: state, consequences, validation, safety rules, and the final report.
 
 Live Space:
 https://huggingface.co/spaces/build-small-hackathon/LifeChoice-Simulator
@@ -32,7 +32,7 @@ You enter a dilemma such as "MTech vs Software Job", choose one path to simulate
 
 Those choices also create durable facts, obligations, and closed options. Earlier decisions return later through cascade moments, and the final report is computed from the causal ledger instead of from a loose conversation summary.
 
-The model does not own the simulation. The deterministic engine owns state transitions, scoring, validation, fallback behavior, and safety boundaries. A 7B model is used only for optional bounded node enrichment.
+The 7B model generates the adaptive scenarios. Deterministic code owns state transitions, scoring, validation, fallback behavior, and safety boundaries so the model output stays grounded in a causal system.
 
 Built with Python, Gradio, Hugging Face Spaces, Qwen/Qwen2.5-7B-Instruct, and Codex as the implementation partner across architecture, UI, validation, docs, and deployment polish.
 
@@ -50,7 +50,7 @@ This is LifeChoice Simulator, a causal future simulator for difficult real-life
 
 I start with a fork in the road: MTech versus Software Job. I choose the future I want to enter, add the pressure that makes the decision real, and pick the voice that follows me through the simulation.
 
-The first scene loads immediately because it is deterministic. Every choice changes five world metrics: money, fulfillment, validation, stress, and family satisfaction.
+The experience is powered by a 7B model that generates bounded future scenarios. Every choice changes five world metrics: money, fulfillment, validation, stress, and family satisfaction.
 
 As I continue, the app remembers what I chose. Earlier decisions create facts, obligations, and closed options that return later as consequences.
 
@@ -58,4 +58,4 @@ The environment also reacts. The world can become stable, thriving, or strugglin
 
 At the end, LifeChoice generates a report from the causal ledger. It does not predict my future or tell me the correct answer. It helps me see what each path costs, what it protects, and what it asks me to carry.
 
-Built for the Build Small Hackathon with Gradio, Hugging Face Spaces, a 7B model for optional bounded enrichment, and Codex as the implementation partner.
+Built for the Build Small Hackathon with Gradio, Hugging Face Spaces, a 7B model for bounded adaptive scenario generation, and Codex as the implementation partner.