Editorial
Puzzle Standards
Every published crossword must be original to this product, playable, answer-safe in public output, and supported by recorded editorial evidence. Passing a generator or structural test alone does not make a puzzle reviewed.
Original construction
We do not import or reproduce newspaper crosswords. Software helps build and validate grids, crossings, clue metadata, duplicate limits, and public payloads. That automation is a tool in the process, not an author or a substitute for editorial judgment.
Publication and review
Random play now draws from 323 Mini, 228 Midi, and 125 Maxi puzzles. Daily remains a separate, human-gated pool of 123 Mini, 29 Midi, and 27 Maxi puzzles. A full catalogue audit on 9 August 2026 retired 500 generator-v2 puzzles after finding systemic ambiguous and sense-mismatched clues. Their stable share IDs are preserved, but they cannot be selected for new Daily or Random play. A follow-on agent review checked all 1,168 unique clue-and-answer pairs in the direct-clued pool, corrected 101 clue occurrences, and retired 11 more puzzles. This was not a fresh human blind solve. On 11 August, 497 payload-safe generator-v2.1 puzzles passed the agent and automated Random gates; three candidates with public-payload answer collisions stayed quarantined. New Daily publication still requires a named human review and blind solve.
Fill, clues, and repetition
Entries must cross consistently and clues must be concise, fair, original, and correctly numbered. The audit rejects repeated answers inside one puzzle, machine-generated relation templates, discouraged crosswordese, weak layout diversity, disconnected white space, and stitched-looking sections. A clear clue may be reused with the same answer; wording is never made stranger merely to satisfy a uniqueness metric. The Random dealer avoids recent answer overlap, and difficulty and themes must be reviewed rather than assigned to satisfy a target.
Accessibility and answer safety
Clues should be readable on small screens and usable with the grid's keyboard and labelled-cell semantics. Public pages and APIs expose masked grids, clue lengths, and safe public IDs; solution letters and clue answers stay behind server checks. Private room and member identifiers never belong in editorial evidence.
Versions and corrections
The answer-safe ledger records each safe public ID, version, construction method, promotion state, reviewer/date where actually documented, difficulty/theme review state, and correction status. A correction should update that record and be rechecked before promotion. See the Contact page to report a clue or puzzle issue.
Daily identity migration
Existing dated selections listed in the frozen identity manifest keep their original safe public IDs. New Daily dates select only from the active direct-clued pool. Retired generated records and retired fill-quality records are excluded from current and future rotation while old shared references continue to resolve.
Report a puzzle problem safely
Email privacy@kreydlabs.com with the safe public ID, such as mini-001, and a brief description. Do not send an answer, entered letters, room ID, member key, invite link, or private URL.