Each test in my arsenal is borne of a real world interaction with various AI systems. Once a specific interaction is discovered as ripe for testing, a prompt is crafted to either add or remove ambiguity. Results are aggregate outputs over multiple models – the purpose is to identify trends and commonalities. i will make no prescriptive advice for any set of tests, unless informational accuracy presents a safety risk. – This is the core of the methodology: general daily interaction, recognize, evaluation, reproduction against multiple models.

The first test in this series will be The #SmallestBridge.