OverlayQA
Visual Bug Reporting & Design QA
I5 years as a product designer. The QA loop breaks on every project.
Developer builds a page from a Figma spec. Spacing drifts 8px. Colors shift. A contrast ratio fails WCAG. You screenshot it, write a ticket that says "padding looks off," but you don't give the dev enough info or clarification and then you spend 3 days clarifying what you meant.
Vibe coding made it exponentially worse.
I ran QA on AI-generated apps and found ~160 issues per app. Contrast ratios failing WCAG. Spacing matching no design system. 82% of AI-generated sites sharing the same generic layouts. Cursor, Bolt, Lovable, Figma Make (which I use professionally) don't know your design system exists.
Teams shipping three builds a day still need the same QA pass they were skipping at one build a week.
"The padding looks wrong" is a 3-day conversation. "24px top padding, spec says 32px, here's the selector" is a 2-minute fix. The difference is context at capture. Human or AI code, same bottleneck.
I built OverlayQA to collapse this into one pass . Even without it: if your QA can't match your shipping speed, you ship bugs faster.
Founders building with AI tools: what does your QA process look like?
I've been a designer for 15 years. The same problem kept showing up on every project: designs look perfect in Figma, then something drifts in the build. Spacing is off, colors don't match, an accessibility issue slips through.
The usual fix? Open Figma in one tab, the staging site in another, squint back and forth, then write up a ticket that says "the padding looks wrong on the hero." The developer asks which padding. You screenshot it. They screenshot theirs. Three days of back-and-forth for a 2-minute fix.
So I built OverlayQA.
It's a Chrome extension (in-product capture coming soon) that lets you compare your Figma designs directly against the live page, run AI-powered accessibility audits, check design system token consistency, and export everything to Jira, Linear, Notion, or Slack and more with full context (CSS values, screenshots, viewport, selectors) so there's nothing to guess.
The core loop is: capture what's wrong, get AI analysis, export with full context. No more "can you send me a screenshot of what you mean."
We're live on the Chrome Web Store. Would love feedback from anyone who's felt this pain.
-
1
happy to answer questions or give a walkthrough if anyones interested!
1 Comment
WOO WOOOO EMILIA what did I just notice on your app's homepage hold on hold on let's see, first "Catch the bug your screenshot missed.
Ship the fix today." yeah? that's actually an absolute cinema copy right here but wait a second the headline while close to perfection misses one. Single. Lethal touch. It currently talks about the technical benefits the user will get from using your app which is catching the bug their screenshot missed mhmm? yet it doesnt quite stop them and make them ask themselves "hmm but still what is it for me? a screenshot miss?" when they stop and think instead of feel the urgency of the pain that their screenshot missed that's when they say something like "later" and later means never which makes you the one who has a genuinely valuable tool lose the potential money, instead we focus on the emotional benefit they get from using your tool. Hmmm let me brainstorm with you for a sec, how about we rewrite like this I got something ready in mind do I send it here?