Fill every empty cell in the grid with either 0 or 1. The completed grid must satisfy three rules — no exceptions.
Rule 1 — No triple: No three consecutive 0s or 1s in any row or column. For example, 1 1 1 is forbidden, but 1 1 0 is fine.
Rule 2 — Equal count: Each row and each column must contain exactly the same number of 0s as 1s. In a 6×6 grid, that's three 0s and three 1s per line.
Rule 3 — Unique lines: No two rows can be identical, and no two columns can be identical. Every line must be unique.
Tap an empty cell to place a 0. Tap again to change it to a 1. Tap once more to clear it. Pre-filled cells (darker background) are locked and cannot be changed. When the grid is complete and all three rules pass, you win!
Pick any grid size from 6×6 to 12×12 using the Grid dropdown — Takuzu always uses even sizes so each row and column can be split evenly. Difficulty (Beginner → Cyborg) is independent and controls how many starting clues you get: higher difficulty means fewer clues and deeper deduction. Mix and match: a 6×6 Cyborg is short but brutal; a 12×12 Beginner is long but gentle.
Stuck? Use the Hint button (up to 3 per game) to reveal one correct cell. Hints cost time but save your streak!
Takuzu goes by a few names — Binairo, Binary Puzzle, Takuzu — but the appeal is the same everywhere: it's Sudoku stripped down to two symbols, which sounds easier and absolutely isn't. With only zeros and ones, every cell is a tight logical consequence of its neighbours, and the satisfaction comes from proving each one rather than guessing it. Seven difficulty levels grow the grid and thin out the starting clues, so the deductions get longer and more interlocking the further you climb.
Takuzu looks trivial — just zeros and ones — and then quietly refuses to let you guess. The three rules do all the work: no more than two of the same number in a row, an equal count of each per line, and no two lines identical. What I love is that those constraints feed each other, so a grid that looks half-empty is usually far more determined than it appears. You learn to stop filling cells you "think" are right and start finding the ones the rules force. It's pure binary logic, and once it clicks, every puzzle becomes a chain of small certainties rather than a coin-flip.
Takuzu looks like the simplest puzzle on this site — two symbols, a square grid, three rules — and it gave me one of the most stubborn generation problems I've dealt with. I want to explain why, because understanding the generator's enemy will make you a better solver.
Two of the three rules are local. No three identical symbols in a row you can check the moment a third lands. Equal counts per line you can track as you go. But the third rule — no two rows identical, no two columns identical — is global. It doesn't bite when you place a symbol; it bites at the very end, when a nearly-finished line quietly duplicates one you completed ten moves ago. From a generation standpoint, that's brutal: a board can pass every local check and still collapse into multiple solutions through a late duplicate. My fix was to enforce uniqueness during generation, not just at validation time. The generator refuses candidate lines that mirror an existing one while the board is still being built, instead of discovering the clash after the fact. That single change is what keeps every published board's solution unique, and it's the difference between a puzzle you can trust and one you have to hope about.
Now the solving side, because players make a precisely symmetrical mistake. Everyone learns the no-three rule first — it's the visible one, and pairs practically shout at you to cap them. So players work it exclusively and forget the other two rules exist. The balance rule alone often forces a cell outright: each line holds an equal count of both symbols, so the moment a row contains all of its allotted 1s, every remaining cell in that row is a 0. No cleverness required — just counting. And the uniqueness rule, the one the generator fought so hard, is a solving tool too: if a nearly-complete row would duplicate a finished one under one of its two possible completions, the other completion is forced.
My method, for what it's worth: sweep pairs first and cap the obvious triples. Then count. Any line with its full quota of one symbol fills itself. Only then do I compare nearly-done lines against finished ones for forced uniqueness plays. Three rules, three passes, in that order — cheap deductions before expensive ones.
The small grids teach the rhythm in a few minutes. The larger ones are where the balance-counting really starts carrying the load, and where the puzzle turns from a doodle into something closer to meditation. The daily is the same board for everyone at midnight. Bring all three rules with you; the players who stall are almost always working with one.