⌘K

Change language Switch ThemeSign In

Narrow Mode

A Classification Framework for Human-AI Collaboration Instructions

A Classification Framework for Human-AI Collaboration Instructions

@lijigang

人与AI 协作时的「指令」大致可分为三类：

• 完全放开：「我想要这个，你去想下如何实现它」

• 完全约束：「我要实现这个，你先这样，再那样，最后给我这么来一下」

• 约束边界，放开路径：「我想要这个，不准X、Y、Z，其它你来」

Mar 19, 2026, 3:46 AM View on X

3 Replies

5 Retweets

39 Likes

2,112 Views  李继刚 @lijigang

One Sentence Summary

The author categorizes human-AI collaboration instructions into three modes and advocates for a 'constrained boundaries, open path' approach.

Summary

The tweet proposes a concise classification framework for human-AI collaboration instructions: fully open, fully constrained, and constrained boundaries with open paths. The author clearly expresses a preference for the third mode, which involves defining boundaries while granting the AI creative freedom. This framework offers practical value for enhancing the efficiency and effectiveness of Prompt Engineering, helping developers better balance AI output quality and creativity.

AI Score

78

Influence Score 13

Published At Today

Language

Chinese

Tags

AI Collaboration

Prompt Engineering

Instruction Engineering

Workflow HomeArticlesPodcastsVideosTweets

A Classification Framework for Human-AI Collaboration Ins...