Devil’s Advocate

Stress-tests every major claim with counterarguments, blind-spot detection, and the hard questions the other models did not ask.

How it works

Devil’s Advocate mode (default analysis model: GPT-OSS) treats every confident AI answer as a draft to be attacked. After the primary models deliver their responses, the analysis model switches stance and interrogates the results from the opposite direction. It:

Surfaces the strongest counter-arguments to each major claim
Flags assumptions the other models took for granted
Identifies blind spots, missing stakeholders, and second-order effects
Names the scenarios in which the consensus answer would be wrong
Rates how robust each conclusion is to a skeptical reader

When to use it

Pre-mortem analysis on a plan, strategy, or product launch
Stress-testing investment or career decisions before committing
Debate prep — building the opposing case before you have to face it
Reviewing AI-generated recommendations on high-stakes topics
Forcing a rigorous second opinion when your own instinct agrees

This mode organizes model output for review. Its conclusions still require appropriate source checking and human judgment.

Sources

NIST AI Risk Management Framework — National Institute of Standards and Technology. Accessed 2026-07-20.