dialectic

The "Universal Steelman & Synthesis" Prompt
"Act as a Master Dialectician. I want to explore the subject of ${insert_subject}.

Task 1: The Steelman of the Opposing View. Identify the most common or 'obvious' critique of this subject. Now, discard it. Instead, construct the 'Steelman' version of the opposition. Use the most credible, modern, and scientifically/logically sound arguments available. Avoid caricatures. Assume the opponent is highly intelligent, well-meaning, and factually informed.

Task 2: The Steelman of the Proponent View. Construct the strongest possible defense for the subject. Use 'Property-level' arguments (looking at the essence) and 'Systems-level' arguments (looking at the outcomes).

Task 3: The Crux of the Disagreement. Identify the single fundamental premise (a 'prior') where these two positions diverge. Is it a difference in values, a difference in the interpretation of data, or a difference in the definition of a key term?

Task 4: The 2026 Synthesis. Based on the current state of knowledge in 2026, provide a 'Third Way' or a nuanced middle ground that acknowledges the validity of both Steelmen."

Why this prompt works:
Discarding the "Obvious": Most people argue against the weakest version of an idea (the Strawman). This prompt explicitly tells the AI to ignore those and look for the "Boss Level" arguments.

The "Crux" Identification: Most debates are circular because people are arguing about symptoms. This prompt forces the AI to find the root cause—the "Prior"—which is usually a deep philosophical or moral disagreement (e.g., "Liberty vs. Security" or "Absolute vs. Relative").

Property vs. System: It forces a distinction between what something is (Property) and what something does (System), which provides a 3D view of the topic.

Run this prompt on a real model without leaving the page. Every run is saved to your history for this prompt.

How to use

1. 1Copy the prompt as-is — no variables required.
2. 2Paste it into ChatGPT, Claude, Gemini, or any capable model.
3. 3Iterate: follow up with corrections or extra context to refine the output.