AI summary by TwinMind
AI chat apps are driving me insane
Theo - t3․gg
1h 15m 35s
Claude UI Issues
- Title Generation Problems
- Titles don't generate until stream completes
- Non-deterministic behavior when saving empty chat names
- Sometimes creates titles with chat emoji + truncated input instead of AI summary
- Navigation & Performance
- Full browser refresh on delete operations
- Sticky scroll container behavior causing UI overlap
- Massive DOM node leaks (100K+ nodes from navigation)
- High CPU usage (17-20%) just from mouse movement
- Thousands of re-renders during normal usage
- UI Quirks
- Three help buttons in five-button menu
- Chat controls menu with no actual controls
- Z-index issues with modal close buttons appearing over chat
- Email visibility cannot be hidden from sidebar
- Rename input doesn't auto-focus
- Memory Usage
- 35MB+ memory consumption vs competitors
- Constant DOM node growth during use
ChatGPT UI Problems
- Sidebar Loading Issues
- Layout shifts on page load as features pop in
- Non-deterministic feature appearance (Sora, Dolly buttons)
- 6MB initial content transfer
- Pagination Bugs
- Uses offset-based pagination causing duplicates
- Slow pagination requests (1.2-1.6 seconds)
- "Too many concurrent requests" errors during scrolling
- Duplicate thread display when scrolling
- Navigation Problems
- Threads require loading time on first click
- No prefetching on hover
- Ghost navbar cuts off content with invisible bar
- Inconsistent thread naming and state management
- Performance Issues
- Constant re-renders during message generation
- High CPU usage during streaming
- Memory leaks and unstable thread states
Grok (X.ai) Experience
- Positive Aspects
- Stream resumability after refresh (unique feature)
- Clean command-K style history view
- Good overall performance
- Mobile app quality
- Issues
- "Unable to load last message" errors
- Input lag with large text pastes
- High CPU usage (70-80%) during responses
- Error states causing message failures
T3 Chat Performance Comparison
- Render Optimization
- 3 sidebar renders vs thousands in competitors
- Individual message rendering instead of full thread
- Beta highlighter for improved performance
- Virtualized sidebar for large chat histories
- Memory Efficiency
- 10K DOM nodes vs 100K+ in Claude
- Local storage of entire chat history
- Lower memory usage despite local storage
- User Experience
- Immediate title generation during streaming
- No browser refreshes for any operations
- Double-click title editing
- Instant navigation between threads
Technical Analysis
- React Performance Issues
- React Scan showing excessive re-renders across all apps
- Mouse movement triggering hundreds of renders
- Poor component optimization
- Network Inefficiencies
- Multiple requests for same data
- Poor caching strategies
- Slow API response times
- Design System Problems
- Inconsistent button treatments
- Poor semantic HTML usage
- CSS hack implementations
Business Strategy Discussion
- Company Focus Alignment
- OpenAI/Anthropic focused on model development, not UI
- T3 Chat focused exclusively on user experience
- Different missions create natural competitive advantages
- Market Positioning
- Model providers benefit more from API usage than direct subscriptions
- UI-focused companies can partner rather than compete
- Example: Groq (infrastructure) partnering with T3 Chat (interface)
- Industry Comparison
- Epic Games/Fortnite model - showcase for Unreal Engine
- Anthropic's MCP protocol - ecosystem investment over UI
- Focus on single mission vs trying to do everything
