- Статус
- Offline
- Регистрация
- 25 Окт 2021
- Сообщения
- 24
- Лайки
- 434
После него ИИ перестаёт подлизываться, ищет только актуальную инфу и разговаривает по-взрослому, а не как с ребёнком на созвоне.
Here's the Quick Checklist (Before Writing the Prompt) – use it every time to level up your game:
**Define who the AI should act as: Assign a clear role (e.g., "You are a world-class physicist" or "Act as a sarcastic comedian"). This sets the persona and expertise level right from the start.
**State exactly what the AI must do: Be ultra-specific about the task (e.g., "Analyze this data and predict trends" instead of "Tell me about this"). No ambiguity – spell out the action verbs.
**Provide relevant background & constraints: Give context, limits, or examples (e.g., "Use only data from 2020-2026, avoid jargon, keep response under 500 words"). This prevents off-track responses and tailors the output.
**Specify how the AI should think: Guide the reasoning process (e.g., "Think step-by-step, consider pros/cons, then conclude"). Chain-of-thought prompting on steroids!
**Define what the final output must look like: Describe the format (e.g., "Output as a bulleted list with bold headings, include sources"). Makes it easy to scan and use.
**Lock in rules, limits, and stop conditions: Add safeguards (e.g., "Do not generate illegal content, stop if uncertain, always cite facts"). Keeps things ethical and focused.
Забираем в закладки и не теряем.
Here's the Quick Checklist (Before Writing the Prompt) – use it every time to level up your game:
**Define who the AI should act as: Assign a clear role (e.g., "You are a world-class physicist" or "Act as a sarcastic comedian"). This sets the persona and expertise level right from the start.
**State exactly what the AI must do: Be ultra-specific about the task (e.g., "Analyze this data and predict trends" instead of "Tell me about this"). No ambiguity – spell out the action verbs.
**Provide relevant background & constraints: Give context, limits, or examples (e.g., "Use only data from 2020-2026, avoid jargon, keep response under 500 words"). This prevents off-track responses and tailors the output.
**Specify how the AI should think: Guide the reasoning process (e.g., "Think step-by-step, consider pros/cons, then conclude"). Chain-of-thought prompting on steroids!
**Define what the final output must look like: Describe the format (e.g., "Output as a bulleted list with bold headings, include sources"). Makes it easy to scan and use.
**Lock in rules, limits, and stop conditions: Add safeguards (e.g., "Do not generate illegal content, stop if uncertain, always cite facts"). Keeps things ethical and focused.
Забираем в закладки и не теряем.