Prompt Engineering in Depth

by bam

The Builder track's deep dive into prompting as engineering, not luck. You finished Building with AI (F2); now you learn to design prompts the way you design code: system vs. developer vs. user roles, structured output and tool/function calling with schemas, few-shot example selection, chain-of-thought and self-consistency, ReAct-style reasoning, decomposition and prompt chaining, controlling determinism (temperature, top-p, seeds), defending against prompt injection, and (the part most people skip) evaluating prompts with real test sets, budgeting tokens, and versioning prompts in production. Assumes you write code and call model APIs. Trust DNA throughout: treat model output as untrusted; evaluate, don't vibe-check. F2 (Building with AI) is the recommended prerequisite; L1 (Prompt Engineering, literacy track) is the plain-language sibling for everyone.