▶ Watch ↗AI Engineer World's Fair 20241:34:56
Building with Anthropic's Claude - The Prompt Doctor is In
Read the full talk →Key ideas
Scroll to read ↓A live workshop moves from an overly talkative patient agent to JSON extraction, writing examples, persona control, image inputs, translation grading, and grounded summaries.
- What should a conversational agent say?0:35 ↗
- Establish a baseline before cleaning up the prompt5:36 ↗
- Replace vague requirements with observable behavior13:22 ↗
- Reproduce the JSON failure—and notice truncation20:46 ↗
- Control where generation starts and stops26:48 ↗
- Check format in code; reduce oversized inputs36:26 ↗
- Show the writing style you want42:17 ↗
- Put useful preparation before the answer54:16 ↗
- Separate persona routing from staying in character58:43 ↗
- Role placement, counterexamples, and steering1:06:32 ↗
- Improve the evidence in an image1:11:03 ↗
- Calibrate what a grade means1:18:23 ↗
- Analysis before probabilities1:27:14 ↗
- Extract evidence before summarizing1:30:36 ↗