The Every UUID Website Explained

A new “every UUID” website turns the UUID space into a browsable, searchable list by generating UUIDs in a randomized-looking but fully consistent order—without relying on an impossible-to-render trillion-pixel DOM. The core idea is to treat each UUID as the output of a reversible mapping from an integer index, so the site can both (1) walk through every valid UUID exactly once and (2) jump directly to a specific UUID (or a matching pattern) during search.

UUIDs are 128-bit identifiers, but only a subset of bit patterns are valid UUID versions and variants. The project focuses on UUID v4-style formatting, where 4 bits encode the version and 2–3 bits encode the variant. With those reserved bits fixed, the remaining entropy is treated as 122 bits. The site then needs an ordering strategy: it can’t just list UUIDs sequentially, and it can’t shuffle them naively because search would break. The solution is to generate UUIDs via a reversible “scrambling” function so that an index can be transformed into a UUID in a way that can be undone.

On the engineering side, the biggest constraint isn’t math—it’s browser rendering. Rendering a list with one row per UUID would require an astronomically tall page (on the order of trillions of pixels). Browsers also cap maximum scroll positions, so the site avoids “real scrolling.” Instead, it fixes the page height to the viewport, tracks a virtual scroll position as a large integer, and renders only the UUID rows near the current position plus a buffer. That keeps the UI responsive while still letting users conceptually traverse the entire UUID space.

For ordering, a linear congruential generator (LCG) was tried first, but it produced patterns—especially in higher bits—making the output look insufficiently random. The approach pivoted to a Feistel network cipher, chosen because it’s reversible by design. The index’s 122 bits are split and processed through multiple rounds using XOR and a round function, producing scrambled bits that still allow the original index to be recovered. After scrambling, the bits must be carefully “bit-smooshed” into the UUID v4 layout: the version nibble must be 4, and the variant nibble must be one of 8, 9, A, or B. The transcript emphasizes that this conversion is bug-prone, but once done it yields UUIDs that look random while remaining consistent for any given position.

Search is the next hard problem. With UUIDs no longer in a simple order, the site can’t rely on “scroll until you find it.” Instead, it unscrambles a candidate UUID to recover its index, then jumps to the nearest matching position. Substring search is more complicated: the site generates valid UUIDs that contain the user’s search string by analyzing where characters could legally appear in the UUID’s hex groups, accounting for fixed dashes and the constraints imposed by version/variant bits. The final method doesn’t enumerate every possible match (which would be effectively unbounded), but it generates a practical set by filling the remaining digits around promising patterns and selecting matches closest to the current position.

The result is a functional “all UUIDs in one place” browser with favorites and copy tools, plus a search experience that’s good enough to cycle through matching candidates. The project ends with a wish list—more social features like trending or friend-favorited UUIDs—and a lingering open question about whether deeper cryptanalysis could make substring search more complete in a shuffled UUID ordering.