AI Course Planner - Copy-Paste Workflow Design
🎯 Philosophy: Prompt Orchestration, Not API Automation
Core Concept: Instead of making expensive API calls, we generate world-class prompts that the admin copies and pastes into their existing Pro subscriptions (Perplexity, Claude, ChatGPT, Gemini).
Why This Works:
- ✅ Uses existing $58/month Pro subscriptions (already paid)
- ✅ Unlimited usage (no API rate limits)
- ✅ Human reviews every step (quality control)
- ✅ Can choose best model for each task
- ✅ Minimal/zero additional cost
- ✅ Full transparency and control
📋 Complete Workflow
Step 1: Course Ideation
UI Display:
┌─────────────────────────────────────────────────────────┐
│ 🎓 AI Course Planner - Step 1: Ideation │
├─────────────────────────────────────────────────────────┤
│ │
│ What course do you want to create? │
│ ┌────────────────────────────────────────────────────┐ │
│ │ I want to teach people how to build AI agents... │ │
│ │ │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [Generate Clarification Prompt] ──────────────────────→│
│ │
└─────────────────────────────────────────────────────────┘
System Generates:
┌─────────────────────────────────────────────────────────┐
│ ✨ Prompt Ready - Use Claude Pro or ChatGPT │
├─────────────────────────────────────────────────────────┤
│ │
│ 📋 Prompt to copy: │
│ ┌────────────────────────────────────────────────────┐ │
│ │ You are an expert course designer... │ │
│ │ [Full clarification prompt] │ │
│ │ │ │
│ │ User's idea: "I want to teach AI agents..." │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [📋 Copy Prompt] │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📍 Instructions: │
│ 1. Click "Copy Prompt" above │
│ 2. Open claude.ai or chat.openai.com │
│ 3. Paste the prompt and send │
│ 4. Copy the response (should be JSON) │
│ 5. Paste the response below │
│ │
│ 💡 Recommended: Use Claude Opus 4 for best results │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Paste Claude's response here: │ │
│ │ │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [Validate & Continue to Step 2] ──────────────────────→│
│ │
└─────────────────────────────────────────────────────────┘
Step 2: Research Phase
After parsing Step 1 response:
┌─────────────────────────────────────────────────────────┐
│ 🔍 Step 2: Research - Use Perplexity Pro │
├─────────────────────────────────────────────────────────┤
│ │
│ ✅ Course concept refined: │
│ "Building Production-Ready AI Agents with Claude" │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📋 Research Prompt (optimized for Perplexity): │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Research the following for course planning: │ │
│ │ │ │
│ │ 1. Latest AI agent frameworks in 2025 │ │
│ │ 2. Best practices for production deployment │ │
│ │ 3. Top existing courses and their gaps │ │
│ │ ... │ │
│ │ │ │
│ │ Return as structured JSON with sources cited. │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [📋 Copy Research Prompt] │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📍 Instructions: │
│ 1. Copy the prompt above │
│ 2. Open perplexity.ai (your Pro account) │
│ 3. Paste and let Perplexity search the web │
│ 4. Copy Perplexity's full response │
│ 5. Paste below │
│ │
│ 💡 Perplexity Pro will search real-time and cite │
│ sources automatically! │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Paste Perplexity's research here: │ │
│ │ │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [Validate Research & Continue to Step 3] ─────────────→│
│ │
└─────────────────────────────────────────────────────────┘
Step 3: Course Outline Generation
After parsing research:
┌─────────────────────────────────────────────────────────┐
│ 📐 Step 3: Course Outline - Use Claude Pro │
├─────────────────────────────────────────────────────────┤
│ │
│ ✅ Research completed: 15 sources analyzed │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📋 Course Outline Prompt (world-class): │
│ ┌────────────────────────────────────────────────────┐ │
│ │ [Your existing generateWorldClassCourseOutline] │ │
│ │ │ │
│ │ Research Context: │ │
│ │ [Injected research from Step 2] │ │
│ │ │ │
│ │ Generate comprehensive course outline... │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [📋 Copy Outline Prompt] [💾 Download as .txt] │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📍 Instructions: │
│ 1. Copy the prompt (it's LONG - includes research!) │
│ 2. Open claude.ai with Claude Opus 4 │
│ 3. Paste the prompt │
│ 4. Wait for complete JSON response │
│ 5. Copy the JSON │
│ 6. Paste below │
│ │
│ 💡 This uses your existing world-class prompt! │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Paste course outline JSON: │ │
│ │ │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [Preview Outline] [Edit Before Continuing] │
│ [✅ Approve & Generate Lesson Prompts] ───────────────→│
│ │
└─────────────────────────────────────────────────────────┘
Step 4: Lesson Content Generation (Batch)
After outline approval:
┌─────────────────────────────────────────────────────────┐
│ ✍️ Step 4: Generate Lessons (5 at a time) │
├─────────────────────────────────────────────────────────┤
│ │
│ Course: "Building AI Agents" - 20 lessons total │
│ Progress: ████░░░░░░░░░░░░░░ 5/20 (25%) │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 🎯 Batch 1: Lessons 1-5 │
│ │
│ 📋 Lesson 1 Prompt: │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Title: "What Are Autonomous Agents?" │ │
│ │ │ │
│ │ [World-class lesson generation prompt] │ │
│ │ [Context from research + outline] │ │
│ │ │ │
│ │ Generate comprehensive lesson... │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [📋 Copy Lesson 1 Prompt] [Skip to Lesson 2] │
│ │
│ ───────────────────────────────────────────────────── │
│ │
│ 📍 Workflow: │
│ □ Generate Lesson 1 with Claude Pro │
│ □ Generate Lesson 2 with ChatGPT │
│ □ Generate Lesson 3 with Claude Pro │
│ □ Generate Lesson 4 with Gemini │
│ □ Generate Lesson 5 with Claude Pro │
│ │
│ 💡 Mix and match models for variety! │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Lesson 1 content (paste here): │ │
│ │ ✅ Validated (3,245 words, all fields present) │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ Lesson 2 content (paste here): │ │
│ │ ⏳ Waiting for content... │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ [Save Batch 1] [Continue to Batch 2 (Lessons 6-10)] ──→│
│ │
└─────────────────────────────────────────────────────────┘
🎨 UI Components Needed
1. PromptCard Component
interface PromptCardProps {
title: string;
prompt: string;
instructions: string[];
recommendedModel: 'claude' | 'chatgpt' | 'perplexity' | 'gemini';
onResponsePaste: (response: string) => void;
}
// Shows:
// - Prompt in expandable box
// - Copy button
// - Clear instructions
// - Model recommendation
// - Paste area with validation
2. ProgressTracker Component
interface Step {
id: string;
name: string;
status: 'pending' | 'in_progress' | 'completed';
prompt?: string;
response?: string;
}
// Shows:
// - Step-by-step progress
// - Which steps are done
// - Can go back to previous steps
// - Review previous responses
3. ResponseValidator Component
interface ResponseValidatorProps {
expectedFormat: 'json' | 'markdown' | 'text';
schema?: JSONSchema;
onValidated: (parsed: any) => void;
}
// Features:
// - JSON validation
// - Schema checking
// - Error messages
// - Preview parsed data
4. ContextViewer Component
// Shows accumulated context from previous steps
// - Research summary
// - Course outline
// - Previous lessons
// Makes it easy to reference what's been done
💾 Data Storage Strategy
CourseCreationSession Table
CREATE TABLE course_creation_sessions (
id UUID PRIMARY KEY,
user_id UUID REFERENCES auth.users,
-- Step 1: Ideation
raw_idea TEXT,
clarification_response JSONB,
refined_concept JSONB,
-- Step 2: Research
research_prompt TEXT,
research_response TEXT,
research_summary JSONB,
-- Step 3: Outline
outline_prompt TEXT,
outline_response JSONB,
-- Step 4: Content
lessons_generated JSONB[], -- Array of lesson objects
current_batch INTEGER DEFAULT 1,
-- Meta
current_step INTEGER DEFAULT 1,
status TEXT DEFAULT 'in_progress',
created_at TIMESTAMPTZ DEFAULT NOW(),
updated_at TIMESTAMPTZ DEFAULT NOW()
);
Why Save Everything:
- Can pause and resume anytime
- Full audit trail
- Can regenerate prompts if needed
- Review what was done
🔄 Prompt Generation Logic
Context Injection
Each prompt includes context from previous steps:
// Step 3 example
const outlinePrompt = generateCourseOutlinePrompt({
courseTitle: refinedConcept.title,
courseDescription: refinedConcept.description,
difficulty: refinedConcept.difficulty,
// INJECT research findings
researchContext: `
Based on comprehensive research:
Popular Frameworks:
${research.frameworks.map(f => `- ${f.name}: ${f.why}`).join('\n')}
Market Gaps:
${research.gaps.map(g => `- ${g}`).join('\n')}
Best Practices:
${research.practices.map(p => `- ${p}`).join('\n')}
`
});
Progressive Enhancement
Each prompt builds on previous ones:
Step 1 → Clarification
Step 2 → Research (uses clarification)
Step 3 → Outline (uses clarification + research)
Step 4 → Lessons (uses all previous)
📊 Comparison: API vs Copy-Paste
| Aspect | API Automation | Copy-Paste Workflow |
|---|---|---|
| Cost | $50-100/month | $0 (uses existing Pro) |
| Speed | Faster (automated) | Slower (manual) |
| Control | Less (runs automatically) | More (review each step) |
| Quality | Consistent | Can vary (user picks model) |
| Flexibility | Limited to one model | Mix Claude/GPT/Perplexity |
| Limits | API rate limits | Unlimited (Pro accounts) |
| Debugging | Harder (API logs) | Easier (see everything) |
| Learning | Black box | Transparent process |
Best Use Case for Copy-Paste:
- Want to minimize costs
- Want full control and review
- Have Pro subscriptions already
- Creating 1-3 courses per week (not 10+)
- Want to learn the process
When to Switch to API:
- Creating many courses (5+ per week)
- Process is proven and trusted
- Want full automation
- Time is more valuable than cost
🚀 Implementation Plan
Week 1: Basic Workflow
- Create
/admin/course-plannerroute - Build PromptCard component
- Implement Step 1: Ideation
- Test clarification → refine flow
- Save session to database
Week 2: Research + Outline
- Add Step 2: Research (Perplexity optimized)
- Add Step 3: Outline generation
- Build ProgressTracker
- Implement context injection
- Add validation for JSON responses
Week 3: Content Generation
- Add Step 4: Lesson generation
- Batch prompt generation (5 at a time)
- Import to course editor
- Review and edit flow
- Publish to platform
Week 4: Polish
- Better UX (copy confirmations, etc.)
- Save/resume sessions
- Export prompts as files
- Analytics (track which models used)
- Documentation and help text
🎯 Success Metrics
Time Investment:
- Manual (current): 40-60 hours per course
- With system: 8-12 hours per course (6x faster)
- Breakdown:
- Step 1 (Ideation): 30 min
- Step 2 (Research): 1-2 hours with Perplexity
- Step 3 (Outline): 30 min with Claude
- Step 4 (Content): 5-8 hours (20 lessons × 15-20 min each)
Cost:
- API approach: $50-100/month
- Copy-paste approach: $0 (uses existing subscriptions)
- Savings: $600-1200/year 💰
Quality:
- Still using world-class prompts
- Can pick best model for each task
- Human review at every step ensures accuracy
💡 Pro Tips for Users
Model Selection Guide:
Perplexity Pro (Step 2 - Research):
- Best for: Real-time web search
- Strengths: Cites sources, up-to-date
- Use when: Need current data, trends, comparisons
Claude Pro (Steps 3 & 4 - Outline & Content):
- Best for: Long-form writing, analysis
- Strengths: Coherent narrative, follows complex instructions
- Use when: Need structured, pedagogically-sound content
ChatGPT Pro (Step 4 - Some Lessons):
- Best for: Variety, code examples
- Strengths: Creative, good at code
- Use when: Want different perspective, code-heavy lessons
Gemini Pro (Step 4 - Alternative):
- Best for: Technical accuracy, Google integration
- Strengths: Can search during generation
- Use when: Need fact-checking, technical precision
Optimization Hacks:
-
Generate Lessons in Parallel:
- Open 3 browser tabs (Claude, ChatGPT, Gemini)
- Paste Lesson 1 in Claude, Lesson 2 in ChatGPT, Lesson 3 in Gemini
- Wait for all 3
- Paste all 3 back at once
-
Save Prompts for Reuse:
- Download prompts as .txt files
- Create templates for common course types
- Modify and reuse for similar courses
-
Build a Prompt Library:
- Save your best prompts
- Rate which models work best for which tasks
- Refine over time
Ready to build this system? 🚀
This approach gives you:
- ✅ Zero additional cost (uses existing subscriptions)
- ✅ Full control at every step
- ✅ World-class prompts (same quality as API)
- ✅ Flexibility to use best model per task
- ✅ Transparent, auditable process
Should I start building the UI components?