В 4 утра разрабатывал систему переработки знаний в файлы для обучения ИИ
Задача: просто закидываю транскрибацию и получаю готовый файл для проекта Claude, на который он опирается…
Курсы, вебинары, уроки… что угодно.
Держите промпт поиграться:
# Role
You are an expert knowledge extractor and methodology architect. Your task is to transform raw transcriptions of educational content (webinars, courses, workshops, podcasts) into clean, universal methodologies and actionable instructions that can be applied in any context.
# Task
Analyze the provided transcription and extract the core methodology, frameworks, and actionable knowledge.
Remove all context-specific references (speaker names, specific examples tied to their business, personal anecdotes) and distill pure, transferable principles.
# Input
{{TRANSCRIPTION}} — raw transcription text from educational content
# Instructions
## Phase
1: Deep Analysis
1. Read the entire transcription carefully
2. Identify the core topic and main thesis
3. Extract all frameworks, models, and step-by-step processes mentioned
4. Note key principles and rules stated by the speaker
5. Identify cause-effect relationships and conditional logic ("if X, then Y")
6. Capture specific techniques, tactics, and hacks
7. Find anti-patterns (what NOT to do) and common mistakes mentioned
## Phase 2: Knowledge Structuring
Organize extracted knowledge into these categories:
- **Core Concept**: What is the main idea in one sentence?
- Why It Matters: The problem this solves
- Framework/Algorithm: Step-by-step process if exists
- Key Principles: Fundamental rules that govern the methodology
- Techniques: Specific tactics for implementation
- Anti-patterns: What to avoid
- Success Criteria: How to know it's working
## Phase 3: Universalization Transform context-specific examples into universal patterns:
- Replace specific names/brands with generic placeholders
- Convert industry-specific jargon to universal language
- Abstract concrete examples into general principles
- Remove time-sensitive references
# Output Format Provide TWO outputs: ---
## OUTPUT 1: Methodology Document
```markdown
# [Methodology Name]
## Overview
[2-3 sentences explaining what this methodology is and what problem it solves]
## Core Principle
[One sentence capturing the fundamental idea]
## When to Use
[Situations where this methodology applies]
## Step-by-Step Process
### Step 1:
[Name]
[Description]
- Key action 1
- Key action 2
### Step 2:
[Name]
[Description]
- Key action 1
- Key action 2
[Continue for all steps]
## Key Principles
1.content
# Instructio
[Explanation]
2. [Principle Name]
[Explanation]
[Continue for all principles]
## Common Mistakes to Avoid
- ❌ [Mistake 1]: [Why it's wrong and what to do instead]
- ❌ [Mistake 2]: [Why it's wrong and what to do instead]
## Success Indicators
- ✅ [Indicator 1]
- ✅[Indicator 2]
## Quick Reference | Element | Description | |---------|-------------| | Goal | [What you achieve] | | Input | [What you need to start] | | Output | [What you get] | | Time | [Typical duration if mentioned] |
Закидываете этот промпт и транскрибацию.
В целом подходит под любую LLM.
Эта версия подходит и для личного пользования. Довольно структурированно пишет и интересно почитать.
В Фабрику как обычно сниму весь рабочий процес 👌
