Building with Anthropic Claude: Prompt Workshop
summary
Hands-on prompt workshop where attendees drop real prompts into the Anthropic console and iterate live, proving concrete fixes: use an assistant prefill (an assistant-role message) to lock the model into JSON mode, wrap output in JSON tags + stop sequences to reliably extract parsable JSON, set temperature to 0–0.1 for deterministic results, put source information above instructions, supply positive and negative examples, break large docs into chunks, and run many test cases in the evaluate tab — these tactics cut preambles and hallucinations in demos but require scale testing.
AI-written summary · about
chapters (12)
- 0:00Prompt Workshop intro
- 3:08Workshop format
- 19:35Patient scheduling prompt
- 39:25Spreadsheet data extraction
- 42:20Tweets from articles
- 59:00Multi-persona role-play
- 1:06:42System vs user prompts
- 1:11:12Images and OCR
- 1:17:11Debugging tracebacks
- 1:18:36Translation grading
- 1:29:49Hallucination mitigation
- 1:34:12Closing & Q&A
related talks
48mStop Fighting Your AI Engineering Prompts That Actually Work
37mWorkshop: Prompt Engineering with Watsonx
44mSupercharge Your Productivity Mastering Prompt Engineering
47mThe Art of Prompt Engineering: Best Practices for Effective AI Interactions
29mSupercharge Your Productivity: Mastering Prompt Engineering
01h 08mIn Prompts We Trust
44mThe Science of Effective Prompting
40mPrompt engineering best practices for LLMs on Amazon Bedrock