Lapa Ninja UX Teardown: 7,462 Landing Pages, One Filter at a Time

Lapa Ninja has been saving landing pages since 2015. There are 7,462 of them now, and the number sits directly under the headline on the index page, because the number is the pitch. One screen further down, that same page starts handing them over twenty-four at a time across 311 pages, with eleven category pills above the grid that each replace the last one you picked. Collecting a decade of the web's marketing pages is a real achievement. Getting to the one you need is still entirely your problem.

The index tells you the size, then gives the good seat to an ad

Here's the path a visitor with no specific query takes — index, one category, one entry:

The first screen is honest about scale. 7,462 curated websites for designers and developers, then a pager at the foot of the page reading Page 1 of 311 and a line confirming you are looking at 24 of 7,462. No infinite scroll pretending the pile has no bottom. You are told the size of the room you just walked into.

The band directly beneath the headline, where the first results would sit, goes to a display ad instead. Browsing is free here, so the money has to come from somewhere. The arithmetic of that trade is on screen: the opening view of a 7,462-item archive contains none of its 7,462 items.

Narrowing is a swap, not a stack

Tap SaaS and the set drops from 7,462 to 691. Almost nothing else moves.

Same pill row with SaaS lit, same ad in the same slot, same grid, same pager. Tap a second pill and it doesn't join SaaS. It replaces it. There is no way to hold SaaS and add minimal, or 2026, or dark. NN/g's distinction is the one at stake: filters and facets both shrink a large set, but facets combine and filters don't. Lapa Ninja's pills are laid out like a facet bar and behave like a radio button.

The nav does one thing well here. It prints the count beside every category before you click — Business 2,453, Minimal 3,051, 3D Websites 156 — so the size of a slice is a decision input rather than a surprise.

Then the count vanishes at the door. The index says 7,462; the SaaS page says only Discover saas website designs. The number that told you how much work was ahead is dropped at the exact moment you start doing the work.

The entry page is the reason to be here

One page carries the site's own one-line pitch, a Visit Website link, a Save, Prev / Next arrows for walking the shelf, a full video recording of the real page scrolling, and the marketing site broken into named full-page screenshots — Home, Introducing Reports, Security, OG Image. A visitor can study a live product site end to end without leaving, for free. That is a lot of work per record, kept up across 7,462 records.

The tagging is finished; the query layer was never built

The Website Details panel on that same page files the entry under five axes at once. Category (Technology, Education, Product). Color (White, Black). Typeface (ABC Oracle). Collection (Sans-Serif). Published (2026).

Every chip is a link. Every link goes to a separate single-axis archive: /category/technology/, /typeface/abc-oracle/, /year/2026/. Hundreds of results each, back to twenty-four at a time.

So the structure needed to answer dark SaaS sites from 2026 set in a sans-serif is already written down, per record, on every one of 7,462 entries. It just can't be asked as a question.

The homepage doesn't ask what you came for

Twelve thousand pixels, and five kinds of thing in them: a featured strip with written reasons, the latest-inspiration grid, an OG-images shelf, paid Webflow templates at $29 to $129, free books, tutorials. Sponsored cards sit inside the curated grid in the same frame as the curation, labelled but not separated. Between sections there is either a display ad or a Discover more box of unrelated links, and most section subheads end with a blue pill styled exactly like a primary button that leads somewhere else entirely.

It reads as one page only because the chrome around the thumbnails is so quiet. All the colour comes from the work itself. That restraint is doing a lot of load-bearing work, and it is also the reason the sponsored cards blend in.

The test to run on your own catalog

Steal the counts. Printing 2,453 next to Business before anyone clicks costs one number in a template and turns a blind click into a priced one.

Now run the count that matters. Count the axes your records carry. Count the axes your browse surface will combine. If the second number is 1, you have built a warehouse with a good front door, and every visitor who arrives knowing roughly what they want will leave having looked at whatever came up first.