Field notes for AI builders and leaders
Choose the problem your team is facing.
These essays use small experiments to make probabilistic behavior, prompt sensitivity, and reasoning blind spots easier to see and discuss.
Output variability · February 9, 2026
Code Repeat, AI Varies
Why familiar software-testing habits are not enough when a model can answer differently each time.
Prompt sensitivity · February 16, 2026
Gave AI the Same Prompt 8 Times. Got 8 Different Worlds
A visual experiment that makes output variation impossible to miss.
Reasoning blind spots · February 23, 2026
AI Spent 4 Minutes Thinking About a Car Wash. It Still Got the Answer Wrong
A simple test of where a long answer can still miss the situation.