The platform has many useful features, such as layout regions. UFG is also a strong feature. It’s good to see that they’re moving toward AI code generation using MCPs.

Not much. Now that I'm seeing the question, I don't have many positive things to say about it. I guess... I like that the filters view lets me see only things that need approval.

Honestly, the visual diffs are the thing. They catch stuff my assertion tests never will — a button that shifted a few pixels, a broken font, two elements overlapping — and one baseline covers a whole page, so I'm not babysitting a dozen brittle locators to catch the same problems. Day to day, the review flow is what I lean on: something changes, I see the before/after side by side, and I accept the new baseline in one click instead of digging through a trace trying to tell a real bug from an intended change. That's the other thing — when it fails, I'm looking at a picture, not a stack trace, so I usually know what broke in seconds. The Ultrafast Grid checking several browsers and viewports in one run saves me a pile of separate runs too. And it's flagged regressions our functional suite didn't even see.