I wrote JustHTML using coding agents
a day ago
- #HTML5
- #Python
- #Coding Agents
- JustHTML is a Python-based HTML5 parser with zero dependencies and a CSS selector query API.
- The project involved overcoming complex HTML5 parsing challenges, including the 'adoption agency algorithm'.
- Using coding agents, the author iteratively improved the parser to pass 100% of the html5lib test suite.
- The process included multiple iterations, refactoring, and even a detour into Rust for performance improvements.
- Key tools used were VS Code with GitHub Copilot in Agent mode and later Gemini 3 Pro for optimizations.
- The final parser is pure Python, optimized for speed and correctness, and outperforms other parsers in test coverage.
- The author emphasizes the importance of clear goals, code review, and version control when working with coding agents.
- JustHTML is now a robust library with comprehensive test coverage and real-world benchmarking.