Thinking in Hexo

I’m not sure whether it will become the next generation of static site generators — I’m just thinking how it could be improved.

Current Pain Points

• No ESM support: Hexo's plugin system is built on legacy CommonJS (vm.runInThisContext), forcing all plugins to use CJS style
• Slow builds: npm install + tsc takes ~45 seconds, mostly waiting on I/O

Proposed Improvements

1. Switch from npm to bun

• Build time: 45s → 2s (22x faster)
• Already validated on hexo-v7-bun and master branches

2. Refactor from CJS to ESM

• Modernize loadPlugin() to support dynamic import()
• Allow plugins to use native ES modules
• Better tree-shaking and static analysis

3. AI-Native Features

• Agent-friendly APIs for content generation
• Structured output formats for LLM integration
• Hook system for AI-assisted workflows

Roadmap

• Validate bun compatibility
• ESM plugin loader prototype
• AI integration design doc

Quick Start

# package.json
"dependencies": {
  "hexo": "npm:@qzi/hexo@^7.3.0",
  ...
},