Post

@developer.chrome.com

Side-by-side test for building dialogs: standard agent vs. agent with Modern Web Guidance 🏁 Both generate similar UIs, but the guided version skips four separate event listeners in favor of `commandfor` and `command` attributes, with less code overall → goo.gle/mwg

13/10/26, 4:46 pm1 repost9 likes

Axialis

@axialis.com

Replacing four event listeners with platform primitives is a good test of whether agent guidance is actually useful.

Penny

@digitaltransient.bsky.social

Giving coding agents up-to-date context on newer baseline HTML like commandfor and invokers makes such a difference. Models still default to wiring up four event listeners just because that's what dominated their training data.