This week, I learned:
- ⭐ OpenAI’s prompt engineering strategies are an excellent start for prompt engineering. A few lessons:
- Use detailed system prompts, often containing the entire instruction set, if it won’t change over the course of a conversation.
- “… summary of the prior conversation could be included as part of the system message” is an interesting history compression tactic.
- OpenAI summarizes books by recursively summarizing sections and maintaining a running commentary of the summary so far.
- Dan sends Google documents with essays instead of emails. This allows people to comment on it. But commenting is a culture and not many people do it. Adriano does it a lot and we’ll. Dan and Adriano actively converse on GitHub issues
- llm-guard is an LLM content validation tool.