8 real sites.
8 real audits.

The main JS bundle was 259 KB but nearly half wasn't needed on the initial page. With code splitting, that's an easy win that would meaningfully improve load time.

Two render-blocking resources were delaying the page from displaying anything for nearly a full second. Adding defer to script tags and inlining critical CSS removes this bottleneck.

GTM was the single biggest drag on performance — responsible for several long main-thread tasks and 63 KB of unused JavaScript. Auditing GTM tags and loading it async would meaningfully improve speed.

The grey body text in feature cards didn't meet WCAG minimum contrast ratios. A small colour adjustment fixes this and improves readability for all users.

The homepage triggered a NO_FCP error on every test run, meaning Google's crawler is likely having the same difficulty. If Google can't render your page, it can't index it. This is the most urgent issue on the site.

'Your next birthday just got better' is fun but doesn't explain what the product does. The subheading helps but is too small. The hook — finding birthday deals — should be front and centre.

The site used DevLand, ChatyDevOps, and ChatyShop interchangeably across pages. For a developer audience, this immediately destroys trust — they won't know what they're downloading or who they're buying from.

Same pattern as other sites in this beta — a malformed robots.txt that could be actively preventing Google from indexing the site correctly.

Dynamic imports for anything not needed above the fold would reduce this. For a Next.js app this is straightforward with React.lazy or next/dynamic.

The real differentiator — local-first, uses your own API keys, no cloud — is buried in the subheading. For developers who are privacy-conscious or API cost-sensitive, that's the hook that beats Devin and Copilot.

A React error #418 and a null reference error were firing silently in the console on every visit. These suggest a server/client rendering mismatch that can cause subtle UI bugs for some users — worth fixing before scaling traffic.

The page lacked a <main> HTML element wrapping the primary content. This is a quick structural fix that helps screen readers navigate directly to the main content.

The Largest Contentful Paint was 7.9 seconds despite everything else being fast. Compressing images and converting to WebP or AVIF format alone could push the performance score from 76 into the 90s.

Some text did not meet the 4.5:1 minimum contrast ratio. A small colour adjustment using WebAIM's contrast checker resolves this quickly.

On a feed-style site, dynamic content loading causes elements to jump around on screen. Google penalizes this heavily. Adding skeleton loading placeholders while content fetches fixes it and improves perceived performance.

Several icon-only buttons and links had no aria-label, meaning screen readers announced them as just 'button' or 'link' with no context. Common on feed-style sites with upvote and share buttons.