Prompt engineering
(In Large language model) Prompt engineering. The art of conversation for communicating well with an LLM. Schools teaching how to talk to machines have appeared, as have companies seeking people who are good at it.
Two categories
It is useful to divide prompt engineering into two categories.
- Category 1: Conversational skills human speakers learn for machine listeners because machines differ from humans.
- Category 2: Conversational skills needed to communicate the speaker’s intent efficiently, whether the listener is a machine or a human.
At present, Category 1 is the bottleneck, and being good at it often matters, but I suspect this will change soon. In fact, according to one study (2025-08-01, Prompting Science Report 3: I’ll pay you or I’ll kill you -but will you care?), the phenomenon where offering tips or making threats improved output quality has now largely disappeared.
Alternatives to the chat-based LLM UI may also reduce the importance of Category 2 (see: Bottleneck of the AI-human interaction loop), but it cannot disappear altogether, nor would that be desirable.
Materials
- Introduction to prompt design by Anthropic
- Prompt engineering by OpenAI
- Prompt engineering guide for GPT 4.1 by OpenAI
- Prompt engineering guide
- Learn AI Prompting with Google Prompting Essentials
Articles
- 2024-06-06 - The Prompt Report: A Systematic Survey of Prompting Techniques
- AutoPrompt
- Cognitive prompting by Changjoon Kim, Geunbong Kwak
Examples
See also
- PromptWizard by Microsoft research
- Prompt suggestion
- Context engineering: An alternative term proposed on the grounds that “prompt engineering” is too narrow in meaning to be appropriate.
- LLM Prompting tips