Feed AI your full evidence constellation, not a cold question

Why This Is a Rule

You've spent weeks accumulating evidence around an open question — partial answers, contradictions, dead ends, surprising connections. Then you open ChatGPT and type: "What causes X?" You get a generic answer drawn from the model's training data, which ignores everything you've already learned.

The problem isn't the AI. It's the prompt. A cold query gives the model zero context about what you already know, what you've already ruled out, and where your specific confusion lies. The response is accurate-but-generic because the model has no way to target the actual gap in your understanding.

This rule says: assemble the constellation first. Before querying, gather the linked notes — the original question, each partial answer you've found, the contradictions between them, and the explicit gaps you've identified. Feed this entire network to the AI. Now the model can do what it's actually good at: pattern recognition across a complex evidence base, surfacing connections between pieces you haven't linked yet.

When This Fires

• You've been researching a question for days or weeks and have accumulated multiple notes
• You're stuck on a problem where you have lots of partial information but no coherent synthesis
• You want AI to help with a question where the generic internet answer isn't what you need
• You're about to type a simple question into AI despite having a rich knowledge trail behind it

Common Failure Mode

Asking AI a fresh question each time you return to a persistent problem, treating each session as independent. You get the same generic answer repeatedly because the model doesn't know about your previous rounds of investigation. Your accumulated evidence — the most valuable input you could provide — sits unused in your notes while you type cold queries.

The Protocol

Before querying AI about a persistent question: (1) Gather all linked notes — the question itself, each partial answer, contradictions between answers, and identified gaps. (2) Paste them as structured context with labels: "Question: ... Partial answers: ... Contradictions: ... Gaps: ..." (3) Ask the AI to identify patterns across the evidence, not to answer the question from scratch. The model's pattern recognition operating on your curated evidence produces insights that neither you nor the model could reach alone.