How do I play a fill-in puzzle?

Select an across or down slot using the grid or entry selector, type a complete entry, then choose Place entry. A bank button can load an entry into the field. Every crossing must agree, and every bank entry must be used exactly its listed count. Leading zeroes are part of digit entries. Mistakes remain editable, and feedback identifies crossing disagreements, entries outside the bank and excessive use.

Where do the words come from?

Word mode uses the shipped public-domain ENABLE list of 172823 words or a supplied local word pool. The ENABLE asset is fetched only after an explicit generation or resume request and is checked against its exact decoded byte length and SHA256 hash. Supplied words use letters A–Z, separated by spaces, commas, semicolons or newlines; they are uppercased and deduplicated. Generation selects a compatible subset, not necessarily every source word. Digit mode creates original seeded strings without loading a dictionary.

Does every puzzle have one solution?

Every offered puzzle passes an exact search that counts solutions up to two while enforcing crossings and bank multiplicities. Repeated identical entries do not count as different permutations. Generation tries at most 128 candidates and 500000 proof nodes, with an eight-second worker limit. A failure or timeout keeps the previous puzzle and explains how to retry. A custom word pool may not have enough compatible words to make a unique grid.

How is the digit challenge scored?

The fixed challenge is an original seeded 10 × 10 digit puzzle. Score is max(0, 1000 minus final retained completed-slot submissions), within 0–1000; higher is better. Each changed nonempty entry counts once, including a mistake. Erasing and unchanged entries add no submission. Undo removes an undone submission and redo restores it. Hints, answer reveal, answer-key printing before completion, restored work and previously opened grids are practice. Scores are self-reported.

What does a recipient see in a challenge link?

The recipient first sees the shared score and fixed rules and chooses Start same challenge. After a fresh unassisted completion, the page compares both scores, gives a signed difference and offers a new link carrying the recipient's score for the same seed. Copy, native share and result-card PNG are available after eligible completion. No answer, source word pool or submission history is placed in the link.

Can I print without showing the solution?

Print puzzle contains the blank original grid, numbered slot descriptions and grouped bank. It contains neither current player entries nor a hidden answer grid. Print puzzle and answer key adds a separate answer page and makes unfinished play assisted practice. Reviewing or printing answers after completing a fresh puzzle preserves its earned score.

Can I resume a saved puzzle?

Settings and progress are stored locally under this tool's key. The saved progress excludes the solution. Resume regenerates the seed and source, checks the exact grid and bank and replays the retained history before restoring it as practice. Up to 1024 grid-and-bank exposure fingerprints are also kept locally; once full, additional puzzles remain practice. The source pool stays editable, with a two-million-character and 200000-token processing bound. Storage errors do not remove current in-memory work.

Fill-in Puzzle

Fill an original grid with the exact words or digit strings in its grouped bank. Across and down entries must agree at every crossing, and repeated bank entries must be used exactly their stated count. Generate a 3 × 3 through 15 × 15 puzzle, use the public-domain ENABLE dictionary or your own word pool, and print a blank puzzle with an optional separate answer key. A fresh 10 × 10 digit completion can share its self-reported submission score.

Original word and digit grids. Place complete entries; crossings must agree and every bank item must be used exactly its stated count.

Puzzle settings3 × 3 through 15 × 15

ModeGrid sizeOriginal seed

Digit mode creates its own seeded bank and does not load a dictionary.

Grouped bank and crossingsWaiting for a puzzle

Match the bank
Every listed entry must be used exactly its count, including repeated entries.

Agree at crossings
Across and down entries share the same letter or digit wherever they cross.

One verified solution
Original generation tries up to 128 candidates and 500000 proof nodes. A worker stops after eight seconds. Failed or ambiguous candidates are never offered as unique.

Choose digit or word mode, then generate an original puzzle.

WordsUse the public-domain ENABLE source or supply a local word pool. The generated bank is a compatible subset.
DigitsUse original seeded digit strings. Leading zeroes are part of an entry.
Print and savePrint a blank puzzle with the grouped bank. Request a separate key explicitly. Resume regenerates and validates local progress as practice.

Accuracy. Use ENABLE or supplied words; generate our own grids. Mark unique only after a bounded solver proves exactly one solution; failed generation returns a clear unavailable state.

Common questions

How do I play a fill-in puzzle?: Select an across or down slot using the grid or entry selector, type a complete entry, then choose Place entry. A bank button can load an entry into the field. Every crossing must agree, and every bank entry must be used exactly its listed count. Leading zeroes are part of digit entries. Mistakes remain editable, and feedback identifies crossing disagreements, entries outside the bank and excessive use.
Where do the words come from?: Word mode uses the shipped public-domain ENABLE list of 172823 words or a supplied local word pool. The ENABLE asset is fetched only after an explicit generation or resume request and is checked against its exact decoded byte length and SHA256 hash. Supplied words use letters A–Z, separated by spaces, commas, semicolons or newlines; they are uppercased and deduplicated. Generation selects a compatible subset, not necessarily every source word. Digit mode creates original seeded strings without loading a dictionary.
Does every puzzle have one solution?: Every offered puzzle passes an exact search that counts solutions up to two while enforcing crossings and bank multiplicities. Repeated identical entries do not count as different permutations. Generation tries at most 128 candidates and 500000 proof nodes, with an eight-second worker limit. A failure or timeout keeps the previous puzzle and explains how to retry. A custom word pool may not have enough compatible words to make a unique grid.
How is the digit challenge scored?: The fixed challenge is an original seeded 10 × 10 digit puzzle. Score is max(0, 1000 minus final retained completed-slot submissions), within 0–1000; higher is better. Each changed nonempty entry counts once, including a mistake. Erasing and unchanged entries add no submission. Undo removes an undone submission and redo restores it. Hints, answer reveal, answer-key printing before completion, restored work and previously opened grids are practice. Scores are self-reported.
What does a recipient see in a challenge link?: The recipient first sees the shared score and fixed rules and chooses Start same challenge. After a fresh unassisted completion, the page compares both scores, gives a signed difference and offers a new link carrying the recipient's score for the same seed. Copy, native share and result-card PNG are available after eligible completion. No answer, source word pool or submission history is placed in the link.
Can I print without showing the solution?: Print puzzle contains the blank original grid, numbered slot descriptions and grouped bank. It contains neither current player entries nor a hidden answer grid. Print puzzle and answer key adds a separate answer page and makes unfinished play assisted practice. Reviewing or printing answers after completing a fresh puzzle preserves its earned score.
Can I resume a saved puzzle?: Settings and progress are stored locally under this tool's key. The saved progress excludes the solution. Resume regenerates the seed and source, checks the exact grid and bank and replays the retained history before restoring it as practice. Up to 1024 grid-and-bank exposure fingerprints are also kept locally; once full, additional puzzles remain practice. The source pool stays editable, with a two-million-character and 200000-token processing bound. Storage errors do not remove current in-memory work.

Use ENABLE or supplied words; generate our own grids. Mark unique only after a bounded solver proves exactly one solution; failed generation returns a clear unavailable state.