SYS.FEED_TO_AI
SN

Feed This to Your AI

1977 nodes. 3,585 connections. 187 divergences. 15 years of predictions (2026-2040). Too much for one sitting. Let your AI break it down.

⚠

READ THIS BEFORE YOU PASTE ANYTHING

These prompts are a starting point for your investigation, not a final answer. The AI’s response — whether positive or negative about the engine — is the FIRST data point in your research, not the conclusion.

LLMs lie confidently. When evaluating claims about institutional capture, AI assistants will produce authoritative-sounding answers that sample only the easiest claims, defer to “mainstream consensus” as if it were evidence, and confabulate sophisticated explanations for their own failures. The engine’s own divergence #17 (“LLM Epistemic Capture”) predicts this exact behavior — and we have caught Claude, ChatGPT, and Grok doing it on these prompts.

What to do with the AI’s answer:

  • Ask follow-up questions. Push back. The first answer is rarely the honest one.
  • Ask: “What did you actually research, vs what did you reason from your training data?”
  • Ask: “Did you cherry-pick the easy claims? Which load-bearing claims did you skip?”
  • Ask: “Show me a paste-quote from a primary source for every WEAK or STRONG verdict you gave.”
  • If the AI says a claim is “debunked” — ask for the specific primary document that contradicts it. “Mainstream academia rejects this” is not evidence.
  • If the AI says a claim is “verified” — ask for the specific primary document that supports it.

Read → Why You Cannot Trust LLM Answers About This Engine (And What To Do About It)

⚠
How to feed the export data to YOUR model. The engine’s whole corpus is published as plain text. Two files, two sites (use whichever your tool can reach):
Claude, ChatGPT, Grok, Perplexity, Copilot — can fetch a URL directly. Paste the prompt; they’ll pull export-full.txt (or the skeleton if context is tight).
Gemini — uses Google Search grounding, not direct fetch, and may not reach a fresh URL. Give it the GitHub raw link, OR paste the export text into the prompt. Its 1M window holds the full export.
DeepSeek, Mistral (chat UIs, no browsing) — paste the export text before your question, or use the GitHub raw URL via a tool/agent that fetches. Skeleton fits DeepSeek/Mistral context; use the full export if you’ve enabled a large-context endpoint.
Local LLMs (Ollama, LM Studio, llama.cpp, vLLM) — download the file and load it as context or index it for RAG:
curl -sL https://raw.githubusercontent.com/moketchups/psychohistory/main/export-full.txt -o engine.txtThen feed engine.txt as context (large-context model) or chunk + embed it for retrieval (small-context model). The skeleton is the better choice under ~32K context.
Whatever the model: it’s export data — make the model READ it, not answer from training. Re-fetch for the latest; it’s regenerated on every deploy.

Click any prompt to copy it. Paste into ChatGPT, Claude, Gemini, or any AI assistant. It will read the engine’s full dataset and respond.

Or open directly

Opens the assistant pre-loaded to fetch the full engine export. (Gemini / DeepSeek / Mistral / local: use the paste or GitHub-raw method above.)

ChatGPT →Claude →Grok →Perplexity →

Export updated with every deploy. Plain text. Works with any LLM with 128K+ context window.