Ever wondered why Rubber Duck Debugging actually works?
It doesn't work because the duck is smart. It works because it forces us to unpack the implicit knowledge we have about our code, expose hidden assumptions, and make what's obvious in our own minds explicit to something with zero prior context.
Turns out, LLMs are the most sophisticated rubber duck we've ever had.
In this talk, I'll introduce the Cognitive Empathy Framework - a practical model I've been developing, inspired by cognitive psychology, for designing prompts and context for AI coding agents. We'll use real AI engineering examples to explain why techniques like prompt engineering, context engineering, and iterative refinement work so well, and explore the principles that help both humans and AI systems build understanding.
You'll leave with practical, scientifically backed techniques for writing better prompts, designing more effective AI systems, and collaborating more successfully with LLMs.