Towards Viable Google Slides in Open Source

https://chatgpt.com/share/69d3a033-ca24-8331-98ea-093fbcf0c01e

Why This Is Now Achievable

1. Real-Time Collaboration Is No Longer a Research Problem

Previously, this was the hardest part:

• Operational Transform (OT) or CRDT design
• Conflict resolution
• Cursor syncing
• Offline/merge behavior

Now:

• Libraries like Yjs and Automerge provide:
  • Shared data structures
  • Real-time sync
  • Offline-first behavior
  • Awareness (who’s editing what)

Impact:

• You no longer build the collaboration engine—you plug it in

2. The Browser Became a Serious Application Runtime

Previously:

• Canvas/WebGL were immature
• Performance was inconsistent
• Input handling (drag, resize, selection) was painful

Now:

• Stable Canvas/WebGL stacks
• High-performance rendering
• Mature UI ecosystems (React, etc.)

You can build:

• Slide canvases
• Object manipulation (resize, rotate)
• Zoom/pan editors

Impact:

• The visual editor layer is now straightforward engineering, not experimental

3. Editor Frameworks Exist (You Don’t Start From Zero)

Previously:

• You had to write:
  • Selection models
  • Cursor logic
  • Document trees

Now:

• Rich-text and structured editors exist (ProseMirror, Slate)
• Canvas editors (Fabric.js, Konva) handle object manipulation

Impact:

• You assemble instead of invent

4. AI Removes the Integration Bottleneck

This is the biggest shift.

Previously:

• Integrating all components took weeks or months
• Debugging edge cases was slow and manual

Now AI can:

• Generate working scaffolds
• Wire Yjs into an editor
• Build drag/resize behavior
• Implement undo/redo stacks
• Handle serialization (JSON ↔ UI)

Impact:

• What used to be hard engineering glue work is now mostly automated

5. Cloud/Backend Complexity Collapsed

Previously:

• You had to design:
  • WebSocket servers
  • Sync layers
  • Persistence models

Now:

• Yjs providers (WebRTC, WebSocket)
• Firebase/Supabase-type backends
• Serverless infrastructure

Impact:

• Backend is no longer a primary blocker

6. Open-Source Examples Exist (Huge Accelerator)

You are no longer guessing architecture.

There are working systems you can study:

• Collaborative editors
• Whiteboards
• Slide tools
• Notion-like systems

Impact:

• You start from patterns, not theory

7. The Problem Scope Can Be Narrowed

This is subtle but critical.

Google Slides is hard because it tries to be:

• Universal
• Fully featured
• Backward compatible
• Enterprise-grade

You do not need that.

For OSE, the likely needs are:

• Structured slides tied to ideas/build systems
• Embedded editing in wiki
• Export capability

Impact:

• This is a much smaller and more tractable problem space

Links

• Local Server Image Upload Size Spec for Live Cloud Editable Documents