Is there always exactly one solution?

Yes. Each square is only blanked if the puzzle still has a single completion afterwards, and that is checked by counting completions rather than assumed. It used not to be: squares were removed unchecked, and counting found more than one answer in 47 of 100 six-square puzzles and 37 of 40 ten-square ones. If your finished grid differs from the answer shown, one of the three rules is broken somewhere in it.

Why must the grid be an even size?

Because each row and column has to split equally between 0 and 1. An odd-length row cannot be halved.

How is this different from sudoku?

Two digits instead of nine, no boxes, and a rule about repeated lines that sudoku has no equivalent of. The deductions feel quite different despite both being pure logic.

Binary Puzzle Generator

A binary puzzle has three rules and no arithmetic at all, which makes it a pure logic exercise and unusually approachable — anybody who can count to four can start, and the deductions get genuinely intricate. Each row and column must hold equal numbers of 0 and 1, no three of the same digit may sit together, and no two rows or columns may be identical. That third rule is the one that turns it from easy into interesting.

Grid size

Your binary puzzle will appear here.

What this generator does

Produces a puzzle with exactly one answer. A complete grid satisfying all three rules is built first; squares are then blanked one at a time, and after each one the puzzle is solved again from what is left — if a second completion has appeared, that square goes back. So the grid you are given can be reasoned to a single finish from its printed digits, and the answer shown is the only one there is.

How to use this tool

Choose a grid size.
Make a puzzle.
Fill every empty square with a 0 or a 1 so all three rules hold.
Reveal the answer when you are done, or stuck.

Understanding the controls

Grid size: 6×6, 8×8 or 10×10, all even because each row must split equally between the two digits. A 10×10 is substantially harder, not just larger.

Worked examples

6×6: Three 0s and three 1s in every row and every column.
Three in a row: Never allowed — in rows or columns, in the solution or on the way there.
Two identical rows: Rejected, which is the rule that makes the hard deductions possible.
Squares left printed: 17 of 36 on a 6×6, 29 of 64 on an 8×8, 45 of 100 on a 10×10 — the 55% target was reached in all 460 puzzles measured.

Common use cases

Logic practice that needs no arithmetic, for any age
A starter activity that works across a wide ability range
Introducing constraint reasoning before formal logic
A printable puzzle for a cover lesson or a quiet activity
Anybody who likes sudoku and wants something with different rules

How this generator works

A complete valid grid is built by backtracking, testing each placement against all three rules as it goes and undoing anything that fails — which is why the finished grid always satisfies them rather than usually satisfying them. Squares are then blanked in a random order, up to 55% of the grid. Each blanking is followed by a count of the completions the remaining digits allow, stopping at two; a square whose removal lets in a second completion is restored and the next one tried. The complete grid is re-checked before display: equal digits in every row and column, no three in a row anywhere, and no two rows or columns identical.

Randomness and fairness

The grid is built by a randomised backtracking search, so each press gives a genuinely different puzzle. The rules are constraints on that search rather than checks after it, which is why an invalid grid cannot be produced.

For how randomness is produced across the whole site, see how Generate Random works.

Assumptions this tool makes

The three standard binary-puzzle rules, which is the form the puzzle is usually published in.

Limitations and good to know

A single answer does not mean a gentle one. The check proves that one completion exists, not that it can be reached by the simple pair-and-count deductions alone — a 10×10 may need you to rule a digit out by following it to a contradiction.
Difficulty is not graded. A 10×10 is harder than a 6×6, but two puzzles of the same size will vary.
There is no in-page solver or hint system.
Puzzles are not stored and there is no seed here, so a puzzle you liked cannot be recovered — print it before you leave.

Common mistakes

Forgetting the no-duplicate-lines rule: It is the rule that unlocks most of the hard deductions. If you are stuck, compare a nearly-full row with the completed ones.
Only counting within rows: Columns have exactly the same constraints. Most progress comes from switching between the two.
Guessing rather than looking for a forced square: Look for a pair of identical digits — the square either side is forced, because three in a row is not allowed.

Practical tips

Start by filling in anywhere a pair forces a digit either side; that usually cascades.
Count each row and column as you go — once a row has its quota of one digit, the rest are the other.

Privacy and your data

Everything is computed in your browser. There is nothing to type, nothing to upload, nothing stored between visits, and nothing written into the page address. Analytics records that the tool ran and how big the grid was.

Frequently asked questions

Is there always exactly one solution?: Yes. Each square is only blanked if the puzzle still has a single completion afterwards, and that is checked by counting completions rather than assumed. It used not to be: squares were removed unchecked, and counting found more than one answer in 47 of 100 six-square puzzles and 37 of 40 ten-square ones. If your finished grid differs from the answer shown, one of the three rules is broken somewhere in it.
Why must the grid be an even size?: Because each row and column has to split equally between 0 and 1. An odd-length row cannot be halved.
How is this different from sudoku?: Two digits instead of nine, no boxes, and a rule about repeated lines that sudoku has no equivalent of. The deductions feel quite different despite both being pure logic.