Reasoning
Sycophancy Loop
Returning to the model because it keeps making your first thoughts feel correct.
How it feels
You feel sharpened, but mostly because the machine has polished the version of you that was already winning.
At work
A manager keeps asking AI to "stress test" a plan, but accepts only the critiques that preserve the original strategy.
Hidden cost
Reasoning: disagreement becomes theater instead of contact with reality.
You may have this if
Ask the model for the strongest case that your favorite idea should be killed. Do you feel curious or annoyed?
Try this instead
Use falsification mode: ask for disconfirming evidence, then write one change you will make if the critique is true.
Team rule
For high-stakes plans, require a red-team prompt whose output is reviewed by someone who did not author the plan.
Related book concept
Cognitive Ops and steelman/falsification prompting.
Evidence notes
The book's Cognitive Ops mode treats AI as a sparring partner, not a reassurance machine.
Sycophancy Loop describes the user habit created when agreeable outputs become emotionally rewarding.
Search surface
Related entries
Now measure the circuit
Find where this pattern is hitting you first.
The Anti-AI Brain Score maps your weakest circuit, then points you back to the practices that reload it.