Research Notes: The Feedback Loop We Can't Exit

How I Found the Pattern

Started researching AI model degradation expecting a future threat. Found something worse.

The trail began with scattered reports about models getting worse through training on their own outputs. The technical term: “model collapse.” The metaphor that stuck: “AI inbreeding.” Spent three days down this rabbit hole. The pattern that emerged was already run…