I built a zero-dependency streaming Markdown renderer for AI chat replies, n× faster than the next one I benchmarked · AI Intel