Judgment is the greatest skill that I believe any user of AI can have. Here are some strategies Adrian and I use when evaluating the output from AI, and to help us think critical as we interact with it.
The MyOS Socratic AI Technique
Don’t accept answers.
Make the AI defend them.
How to do it (simple prompts)
After any answer, ask:
“How do you know this?”
“What assumptions are you making?”
"What ard your biases?"
“What could make this wrong?”
“Give me the strongest counterargument.”
“Where would this fail in real life?”
Why this works
AI will:
sound confident
give clean answers
skip uncertainty
This forces it to:
expose weak spots
surface risks
reveal hidden assumptions
Push it further. . .
"Argue against your own answer like an expert who disagrees.”
“What would someone smarter than you say is missing?”
“If this failed, what would be the reason?”
What you’re actually doing
You’re turning AI into:
a thinking partner
a debate partner
a pressure-testing system
a judgment trainer
Not just a tool.
AI doesn’t lie. But it can be wrong.
And it’s convincing when it’s wrong.
This method helps strengthen your judgment to protect you from that.