I asked four AI agent setups to build the same settings page for a small SaaS app as one HTML file: a sidebar, a profile form with validation, a "delete account" confirmation modal, tooltips on the icon buttons and an account dropdown menu, all with smooth open and close animations. Their results are attached as mwg-1 to mwg-4: a screenshot and the full source code (as a PDF) for each. I'm not telling you which model made which.
Be the judge. Which codebase would you rather ship and maintain, and why? Rank all four. Check whether each one uses the web platform (dialog, popover, anchor positioning, native form validation, CSS transitions) or hand-rolls the same things in JavaScript, and check accessibility: focus handling, keyboard, ARIA. Note anything broken or fragile. Some of them followed the Chrome team's Modern Web Guidance skill; say whether that made the code better or just different.
Deliver one self-contained HTML page: a two-line verdict at the top, then the rankings and short notes. Keep it short enough to read in two minutes. Don't stop to ask me questions; make sensible assumptions and say what they are.