Back to Guidesxeditai
Intermediate20-25 minutes

Parallel Mode & Hybrid Authoring

Compare models side-by-side, copy sections between views, and create composite documents from multiple AI perspectives.

Overview

Parallel Mode allows you to generate with 2 different models simultaneously and compare their outputs side-by-side. This enables hybrid authoring—where you cherry-pick the best sections from each model to create a superior composite document.

What You'll Learn

Parallel Execution - Run 2 models simultaneously
Multi-View Display - Compare outputs side-by-side
Cross-View Copy - Move sections between models
Hybrid Documents - Merge best parts
Model Comparison - Identify strengths
Best Practices - When to use parallel mode

đź’ˇ Why Parallel Mode?

Different models excel at different tasks. Claude might write better introductions, while GPT4 might create superior technical content. Parallel Mode lets you leverage each model's strengths by comparing outputs and cherry-picking the best sections.

Real-World Use Case: A product manager generates a PRD using both Claude Sonnet (for structured thinking) and GPT-4 Turbo (for detailed specifications). They copy Claude's "Problem Statement" and GPT-4's "Technical Requirements" into a single hybrid document that's superior to either individual output.

Prerequisites

  • Completed Guide 1: The Editable Story (understand basic editing)
  • At least 2 active models configured in your account
  • Understanding of model capabilities (e.g., Claude vs GPT)

Part 1: Configure Parallel Mode

Step 1: Open Mode Overlay

  1. Navigate to /authoring workspace
  2. Click the "Mode Overlay" button (Lab/Settings icon in toolbar)
  3. Mode Overlay panel slides in from the left (400px wide)
  4. Current mode badge shows "Single" (default)
SCREENSHOT-04-01

Authoring workspace with Mode Overlay open on left side. Panel shows mode selector at top with three options: 'Single' (currently selected with cyan background), 'Parallel' (gray), and 'Strategy' (gray). Below mode selector, Lab settings section is visible but grayed out. Workspace on right shows empty prompt area.

Step 2: Select Parallel Mode

  1. In the mode selector section, click "Parallel" button
  2. The Parallel button highlights with cyan background
  3. Lab settings section updates to show parallel-specific options
  4. Notice the requirement: "Parallel mode requires exactly 2 models"

Validation Requirement

The "Apply" button will be disabled unless exactly 2 models are selected. If you select 1 or 3+ models, an error message appears: "Parallel mode requires exactly 2 models"

SCREENSHOT-04-02

Mode Overlay with 'Parallel' selected (cyan background). Below, Lab settings shows: Template dropdown with 'Product Requirements Document' selected, Available Models list showing checkboxes for models with SKU levels. Two models are checked: '2 - Claude Sonnet 4.5' and '4 - GPT-4 Turbo'. Red validation message at bottom reads: 'Parallel mode requires exactly 2 models'. Apply button is enabled (cyan).

Step 3: Select Two Models

  1. In the "Available Models" list, check "Claude Sonnet 4.5"
  2. Check "GPT-4 Turbo" (or any second model of your choice)
  3. The selected models highlight with cyan checkboxes
  4. Validation message disappears
  5. Select Template: "Product Requirements Document" (or your preferred template)
  6. Click "Apply" button to save configuration
  7. Mode Overlay closes, workspace shows "Parallel" mode badge in toolbar

Pro Tip: Choose Complementary Models

Select models with different strengths:

  • Claude + GPT: Structured thinking vs detailed specifications
  • Gemini + Claude: Deep reasoning vs fast creative output
  • GLM + Claude: Multimodal analysis vs long-context synthesis

Part 2: Generate Parallel Outputs

Step 4: Enter Prompt and Generate

  1. In the prompt textarea, enter:
    Create a Product Requirements Document for a mobile app that helps users track their daily water intake with gamification features
  2. Click the "Generate" button
  3. Both models start generating simultaneously
  4. Progress indicator shows: "Running 2 models in parallel..."
  5. Wait for completion (typically 30-60 seconds depending on models)
SCREENSHOT-04-03

Workspace during parallel generation. Prompt area shows text: 'Create a Product Requirements Document for a mobile app...'. Below prompt, two progress spinners side-by-side: Left shows 'Claude Sonnet 4.5' with cyan spinner, Right shows 'GPT-4 Turbo' with cyan spinner. Status text reads 'Running 2 models in parallel...' with completion percentage '45%'.

Step 5: View Multi-View Display

  1. When generation completes, workspace splits into two columns
  2. Left column: Claude output (labeled at top)
  3. Right column: Openai output (labeled at top)
  4. Each column shows a complete PRD with sections (Problem Statement, Features, etc.)
  5. Both columns are independently scrollable
  6. Notice: Section structures may differ between models
SCREENSHOT-04-04

Workspace showing multi-view display. Left column (50% width) has header 'Claude Sonnet 4.5' with cyan accent, shows 5 sections: 'Problem Statement', 'Target Audience', 'Core Features', 'Success Metrics', 'Timeline'. Right column (50% width) has header 'GPT-4 Turbo' with blue accent, shows 6 sections: 'Executive Summary', 'User Stories', 'Feature Specifications', 'Technical Requirements', 'Milestones', 'Appendix'. Vertical divider line (1px gray) between columns. Both columns independently scrollable.

Part 3: Compare and Copy Sections

Step 6: Compare Section Quality

  1. Read the "Problem Statement" section in Claude's output (left)
  2. Read the "Executive Summary" section in Openai's output (right)
  3. Notice differences in:
    • Tone: Claude may be more concise, Openai more detailed
    • Structure: Different section organizations
    • Content Focus: Claude emphasizes user needs, Openai covers technical specs
  4. Decide which sections you prefer from each model

Comparison Criteria

Clarity: Which section is easier to understand?
Completeness: Which covers more relevant details?
Structure: Which has better organization?
Actionability: Which provides clearer next steps?

Step 7: Copy Section from Right to Left

  1. In the right column (Openai), find "Feature Specifications" section
  2. Click the copy icon (đź“„) in the section header
  3. GPT-4's "Feature Specifications" section appears in Claude's view
  4. Notice: The pasted section retains all formatting and content
SCREENSHOT-04-05

Multi-view display showing copy operation. Right column (GPT-4) shows 'Feature Specifications' section with copy icon (đź“„) highlighted in cyan, cursor hovering. Toast notification in bottom-right corner shows 'Section copied to clipboard' with green checkmark. Left column (Claude) shows '+' button glowing between 'Core Features' and 'Success Metrics' sections. Dropdown menu from '+' button shows options: 'New Section', 'Paste Section' (highlighted in cyan), 'Import from File'.

Step 8: Build Hybrid Document

  1. Continue copying the best sections from each view to your primary view (e.g., left column)
  2. Example hybrid composition:
    • From Claude: Problem Statement, Target Audience, Success Metrics
    • From Openai: Feature Specifications, Technical Requirements, Appendix
  3. Drag sections in the left column to reorder as needed
  4. Edit section names if necessary (click section header to rename)
  5. Delete unwanted sections (trash icon in section header)
  6. The left column now contains a hybrid document with the best parts from both models

Result: Best-of-Both-Worlds Document

Your final document combines Claude's concise problem framing with Openai's detailed technical specifications—creating a PRD that's superior to either individual output. This is the power of hybrid authoring.

Part 4: Save and Export

Step 9: Switch to Single View

  1. Once you've created your hybrid document, you can collapse to single view
  2. Click the view selector (top-right of workspace)
  3. Options: "View 1 Only", "View 2 Only", "Split View"
  4. Select "View 1 Only" to show only the left column (your hybrid document)
  5. The workspace expands to full width showing only your selected view

Step 10: Save to Cloud

  1. Click "Save" button in toolbar
  2. Modal opens: "Save Specification"
  3. Enter name: "Water Tracking App PRD - Hybrid"
  4. Select context level: "L1: Project" or appropriate hierarchy
  5. Click "Save"
  6. Your hybrid document is saved to cloud with both model outputs preserved
  7. Note: Both original views remain accessible in version history

Step 11: Export to PDF

  1. Click "Export" button in toolbar
  2. Select "PDF" format
  3. Modal shows export options:
    • Include branding: Enabled
    • Include model attribution: Enabled (shows which sections came from which model)
  4. Click "Export PDF"
  5. PDF downloads with hybrid content and optional model attribution watermarks

Best Practices & Use Cases

When to Use Parallel Mode

High-Stakes Documents

PRDs, RFPs, executive summaries where quality matters more than speed

Model Evaluation

Testing which models perform best for your specific use case

Cross-Verification

Technical specifications, legal documents, or compliance materials that need double-checking

Creative Exploration

Marketing copy, blog posts where you want diverse creative approaches

When NOT to Use Parallel Mode

Draft Content

Quick drafts, meeting notes, or internal documentation where single model output suffices

Cost-Sensitive Scenarios

Parallel mode costs 2x since both models run. Use single mode if budget is tight.

Real-World Examples

Example 1: Product Launch Document

Models: Claude + Openai

Strategy:

  • Used Claude's "Market Analysis" (better synthesis of trends)
  • Used GPT-4's "Go-to-Market Strategy" (more tactical details)
  • Used Claude's "Success Metrics" (clearer KPI definitions)
  • Result: Comprehensive launch doc in 1/3 the time vs sequential editing

Example 2: Technical RFP Response

Models: o1 (reasoning) + Gemini Pro (multimodal)

Strategy:

  • Used o1's "Technical Approach" (deep reasoning about architecture)
  • Used Gemini's "Implementation Timeline" (better project planning)
  • Combined both for "Risk Assessment" (cross-verified for completeness)
  • Result: Winning RFP with 95% compliance score

Troubleshooting

âť“ Can't select more than 2 models?

Parallel mode is limited to exactly 2 models by design. For 3+ models, use Strategy Mode (covered in Guide 5).

âť“ Sections from different views have different structures?

This is expected—different models organize content differently. When copying between views, you can edit section names and content to maintain consistency in your hybrid document.

âť“ Generation takes too long?

Parallel mode runs both models simultaneously, so total time = max(model1_time, model2_time), not the sum. If still slow:

  • Check if one model is particularly slow (e.g., o1 with deep reasoning)
  • Reduce template complexity or prompt length
  • Use faster models for parallel comparison (e.g., Google: Fast and Effective, Google: Fast and Frontier)

Summary

What You've Mastered

Configure parallel mode with exactly 2 models
Generate and compare side-by-side outputs
Copy sections between views
Build hybrid documents combining best parts
Identify when to use parallel vs single mode
Export hybrid documents with attribution

🚀 What's Next?

Guide 5: Strategy Mode & Multi-Agent Workflows takes parallel execution to the next level—using 3 models with specialized roles (Drafter, Critic, Synthesizer) to create even more sophisticated outputs through ensemble synthesis, debate arbitration, and mixture-of-agents strategies.