What is the primary cause of most real-world AI failures?	The interaction (collision) of two or more properties.
Name the four properties of AI discussed in the lesson.	Next Token Prediction, Knowledge, Working Memory, and Steerability.
What failure pattern results from the collision of Next Token Prediction and Knowledge?	Hallucinated specifics.
What failure pattern results from the collision of Working Memory and Steerability?	Long-conversation drift.
What diagnostic move is applied when naming the colliding properties to determine the correct fix?	Discernment.
If AI fails due to a property collision, what is the benefit of naming the properties?	It points directly to the most targeted fix.
Name three targeted fixes that can be applied after diagnosing an AI failure.	Verify specifics, re-supply context, or offload to code execution (or invite pushback).