What you will learn.
Prompt engineering acquired a bad reputation around 2024 because the field was crowded with people selling “prompt libraries” and Twitter threads. That reputation has not aged well. The engineers who consistently extract better quality from frontier models are doing something specific, replicable, and teachable — and the gap between them and the median engineer is, if anything, widening.
This is a four-week course on prompt engineering as a real craft. We cover the underlying behavior of modern instruction-tuned models, the patterns that work and the patterns that look like they should work but don’t, structured outputs at scale, the relationship between prompt and eval, prompt rot under model upgrades, and a working method for iterating on prompts that does not waste your week. Short, dense, opinionated.
This course pairs naturally with AI-101 (which precedes it) and AI-301 (which builds the evaluation discipline that makes good prompt work possible).