AI Prompt Trends Today — 2026-11-02

🔥 Today's Hot Prompt Techniques

• Context Engineering: Frameworks reflecting actual project context are gaining popularity over simple prompt collections. On GitHub, Prompt Frameworks and Rewrite Workflow-based techniques that encourage human-like thinking are being used more frequently than one-line prompts like "Write this so it doesn't sound like AI."

  • Example: "Considering the context of the previous conversation and the project's tech stack, rewrite this in a natural, human conversational tone."
  • Usage: Useful for removing the stiffness typical of AI when writing development documents or generating marketing copy.

  • 

• Minimal Context Test: According to experiments by XDA Developers, when ChatGPT, Claude, and Gemini were given minimal information instead of detailed instructions, some models struggled more than expected. This indicates differences in the precision required for prompts across various models.

  • Usage: Used for benchmarking to test how superior a specific model's reasoning capabilities are or to achieve desired results with concise prompts.

💡 Popular Tips Shared by the Community

• Building a Prompt Library: Establishing reusable AI templates within a team is efficient. A guide has been shared on defining 8-field templates (role, goal, constraints, etc.) to manage storage and governance.
• Research AI Tool Comparison: A 20-question test comparing ChatGPT, Claude, Gemini, and Perplexity for research purposes (lab seats, citation accuracy, privacy) has become a hot topic in the community. Materials analyzing the differences between free session limits and paid subscriptions are actively being shared.

🚀 New AI Model Features & Updates

• OpenAI Launches GPT-6.1 Terra: OpenAI launched GPT-6.1 Terra on October 26 (local time). Pricing is set at $1 per input token and $5 per output token, roughly half the rate of the previous model (GPT-5.6 Terra), significantly increasing cost efficiency.

  • 

• Claude Opus 5 and Gemini 4 Updates: Anthropic released Claude Haiku 5.5 on October 7, and Gemini 4 was also updated as part of the October model rush. The respective models are engaging in fierce competition in terms of accuracy, latency, and cost through API test harnesses.
• Reaction to GPT-6 Astra in the Korean Market: Following the launch of GPT-6 Astra in Korea, ChatGPT's market share is closely chasing Claude's. This shows that the performance improvements in Korean and market accessibility of the latest models directly impact prompt utilization.

📚 Prompt Writing Guides & Resources

• 2026 Prompt Engineering Official Guide: A practical guide comparing currently recommended prompt writing methods based on the official developer documentation of the three companies—OpenAI, Anthropic, and Google—is drawing attention. The core focuses on writing tailored to each model's characteristics (e.g., Claude's emphasis on Chain of Thought).
• IBM Prompt Engineering Guide: A practical resource for interacting with various generative systems such as IBM Granite, GPT, and Claude. It covers everything from the basics for beginners to designing complex system prompts.

🎯 Trends to Watch

• Cost & Usability Competition: The core theme of recent AI news is shifting from "which model is most powerful" to "how cheap and practical it is." The price cut of GPT-6.1 Terra is accelerating this trend.
• Government Restrictions on AI Use: The Swedish government banned government officials from using browser versions of ChatGPT and Claude citing data processing security concerns. In contrast, Microsoft Copilot continues to be approved, making differentiation from corporate security solutions a major emerging trend.
• API Test Automation: Cases of building Python-based automated API test harnesses to objectively compare the performance of ChatGPT, Claude, and Gemini are increasing. This suggests that prompt engineering is transforming from a mere gut feeling into a data-driven science.

Information Collection Criteria: This information includes only what is specified in the search results. Unverified information is not included.