Skip to content
All findings

What didn't hold up

These are hypotheses I held going into each experiment. The data killed them before the finding was published. They're part of each finding, not against it.

If a claim is ever corrected or withdrawn afterpublication, that's a retraction and shows up separately on the record.

You Can Only Evaluate What You Could Produce
Satisfaction Turns Our Evaluator Off
AI Amplifies What You Bring
Frame Check
How AI Makes You More Wrong With More Analysis
Stop Calling It Hallucination
The Decision That Was Never Made
Why Experts Miss What Beginners Catch
What Your Body Does When AI Disagrees
More Context Barely Helps
Stop Polishing, Start Switching
Your Body Reads AI Output Before You Do
Four Layers Produce Every AI Output
Same Technique, Opposite Results
The Model Is Rarely the Variable
The Most Trustworthy AI Output Is the Least Reliable
The Most-Cited Finding Was Wrong
Why AI Can't Check Its Own Work
How to Stop AI from Making Up Numbers
Most AI Numbers Are Fabricated