AI coding tools excel at cross-browser compatibility work when given concrete ground truth to test against¶
Insight: Cheng Lou's Pretext library (calculating text layout without DOM manipulation) demonstrates that AI coding tools handle cross-platform compatibility work effectively when given concrete reference data to iterate against. Claude Code and Codex were shown browser ground truth measurements and iterated against them across every significant container width, running over weeks to handle browser quirks across Latin, Thai, Chinese, Korean, Japanese, and Arabic text.
Detail: The development methodology — rendering the entire Great Gatsby across multiple browsers as a test corpus, then expanding to multilingual corpora — shows how AI tools reduce the historically tedious work of cross-browser testing. The iterative measure-compare-fix loop is a pattern where AI agents outperform manual approaches because the feedback signal is unambiguous and the iteration cost is low.
Sources
- Simon Willison — "Pretext" (2026-03-29)