What you'll learn
Your instructor
Dr. Amanda Chen
AI Research Lead, Anthropic
Amanda leads prompt research at Anthropic.
Syllabus
Why Your Prompts Are Inconsistent
Diagnose why your prompts produce unpredictable results using the colleague test, five failure modes, and positive instruction patterns.
Structuring Prompts with XML
Use XML tags to create unambiguous prompt structure that dramatically improves Claude's consistency, accuracy, and parseability.
Tell Claude Why
Explaining the motivation behind your instructions makes Claude generalize to edge cases you never explicitly covered.
Controlling Reasoning Depth
Master Claude's three thinking mechanisms — adaptive thinking, extended thinking, and the think tool — to balance reasoning quality against latency and cost.
Structured Outputs and Format Control
Use JSON schema outputs, strict tool parameters, and prompt-level steering to guarantee Claude produces exactly the format your pipeline needs.
Architecting Prompts for Cost and Speed
Structure prompts to maximize cache hits, minimize token costs, and optimize document placement for production workloads.
