Fixing four small lies on the Deep Cut Atlas page

This one came out of a design review pass on the Deep Cut Atlas marketing page - four small things, none of them bugs exactly, all of them the page saying something slightly untrue.

The first was typographic. The page used a plain spaced hyphen wherever it needed an actual dash for a parenthetical aside. Small thing, but the actual iOS app’s copy uses a real em dash in those spots, and once I noticed the mismatch I couldn’t unsee it. I went and grepped the app repo’s SwiftUI source to confirm the app’s own convention (a spaced em dash) instead of guessing, and matched the page to it exactly.

The second was a tagline that didn’t quite parse: “Find the releases your favorite artists slipped out.” I read that sentence three times before I understood what it was trying to say. “Slipped out” wants a direct object it never gets - releases don’t “slip out” on their own, in this sentence they’re just there. The fix, “Find the releases that slipped past you,” keeps the same wordplay (slipped) but the grammar actually resolves. That tagline lives in two places - the page itself and the homepage project card’s data file - so I had to update both or end up with a site that contradicts itself depending on which page you’re looking at.

The third was a copy line that oversold what the app does. The History feature card said “History keeps everything you’ve listened to.” It doesn’t - it tracks plays across playlists and radio, which is exactly what the feature actually does and what the screenshot next to that copy shows. “Everything” is the kind of word that’s technically fine until someone asks “wait, everything? even X?” I swapped it for “History tracks what you’ve listened to” - same information, no claim the app can’t back up.

The fourth was the most mechanical: the page’s JSON-LD structured data listed a url without a trailing slash, while the actual rendered canonical tag has one. Astro’s directory-style routing means /apps/deep-cut-atlas/ is the real URL; the JSON-LD was quietly pointing search engines somewhere slightly different from where the page said it canonically lived. One character.

None of these would have shown up in a build or a type check - they’re all correct HTML, correct data shapes, nothing throws. The only way to catch them was someone actually reading the page like a visitor would, which is what the design review was for. It’s a good reminder that “the build passed” and “the page is right” aren’t the same claim.