Skyscrapers takes the plain Latin square (every row and every column holding each number exactly once) and turns the numbers into building heights viewed from outside the grid. Around the edges sit clue numbers: each one tells you how many buildings are visible looking straight along that row or column from that side, given that a taller building always hides every shorter one standing behind it.
Every board here is generated fresh from a random seed and is checked to have exactly one solution, so no guessing is ever required. The seed lives in the page URL, so you can bookmark or share a specific puzzle. Pick a size above to begin: smaller grids keep the sightlines short, larger ones give the visibility clues far more to say.
The visibility clues are what change character across sizes: on a 4×4 board a clue of 1 or 4 is nearly a giveaway, since only one arrangement of heights produces that extreme, but on a 7×7 board those extreme clues still help while a run-of-the-mill middle count like 3 leaves several tall-short orderings in play until neighbouring clues rule the rest out. Reading a sightline is never just about the number on the edge: it’s about how many height orders that number can possibly hide.

Skyscrapers asks its edge-clue counts to carry what a Sudoku box normally would, and the generator holds them to the same bar: a height only leaves the grid once the solver has checked that no other skyline could produce the same visible-building counts from every side. A sightline number is never approximate here: it is exactly as many buildings as that one arrangement allows, proven rather than eyeballed.
A visibility number isn’t chosen and then matched to a height order. It’s the reverse: the generator counts how many heights form an increasing run when read inward from an edge, and that count becomes the clue. Given-count floors sit low at every size precisely because that readout is so informative on its own; the dig loop only stops once no floor-respecting removal would still leave the grid provably unique.