AI summary by TwinMind
Rethinking The Linter
Theo - t3․gg
19m 26s
Million.js Background & Context
- Million.js: Alternative React compiler created by high school student, now Y Combinator founder
- Alternative to React DOM with up to 70% performance improvements
- Multiple companies using Million in production
- Different from typical linters - focuses on performance rather than code quality
Million Lint Overview & Launch
- Public beta launch after 3 months development, hundreds of commits
- VS Code extension that identifies slow React code and provides optimization suggestions
- One-command setup:
npx million@latest - Tagline: "Like ESLint but for performance"
Performance Debugging Problem Analysis
- React Dev tools difficult to parse, not beginner-friendly
- Existing profiling tools (Sentry, DataDog) show too much irrelevant data
- Include Chrome extension crashes, network requests, memory data
- Overwhelming interfaces for non-experts
- Flame graphs don't provide cause-and-effect relationships
- Current performance debugging "hostile to most developers"
Technical Architecture - Dynamic Analysis Approach
- Hybrid approach combining static analysis + runtime instrumentation
- Problem: Compiler-only approach can't predict actual performance without execution
- Solution: Custom instrumentation library for React apps
- Three-part system:
- Compiler injects
million.capture()calls around components - Runtime captures rendering timing and metadata during dev mode
- VS Code extension displays collected data with optimization suggestions
- Compiler injects
Core Features & Capabilities
- In-editor performance feedback
- Component render counts (e.g., "renders 2,154 times, takes 22ms")
- Bundle size impact per import
- Real-time insights without build time overhead
- Bundle analysis more accurate than other extensions due to compiler integration
- Lint++ feature: Language model analyzes collected data for optimization suggestions
AI Implementation (Lint++)
- Uses language model to discover optimization opportunities
- Notable: AI not marketed as primary feature, just implementation detail
- No mention of "AI" on landing page despite being 2024 YC company
- Focus on developer experience over AI buzzwords
Example Optimizations Suggested
- React.memo wrapping for components
- React Virtuoso for large list virtualization
- Debouncing filter inputs
- Stable callback creation patterns
- Use context selector optimization
- Context provider positioning in component tree
Performance Testing & Validation
- Tested on Avon's educational React performance workshop app
- Identified issues across 6 components:
- Unstable prop references
- Expensive component memoization needs
- Inline function caching
- Virtualization opportunities
- Context optimization
- Engineer unfamiliar with codebase achieved 3x performance improvement in 15 minutes
Pricing Model & Business Strategy
- Compiler and VS Code extension: Free forever
- Lint++ service: $20/month for 100 lints
- Usage-based pricing aligned with value delivery
- Open sourcing compiler and extension planned
Future Roadmap
- Currently focuses on unnecessary re-renders
- Acknowledges React compiler will address some issues
- Plans to expand beyond re-renders to:
- State manager performance
- Animation optimization
- Bundle size optimization
- Network waterfall issues
- Goal: Full web infrastructure performance automation (frontend to backend)
Team & Hiring
- Seeking frontend and ML engineers in Bay Area
- Thesis: Build tools making anyone a performance expert
- Vision: Developers focus only on features/bugs, not performance optimization
- Plans to expand from React to broader frontend/backend platforms
