Cryptoku is now called Cryptic. You can play it free in your browser or get it on Google Play.Cryptoku is near the end of closed testing now. I’ve been collecting feedback and changing the app, and the changes have mostly been less glamorous than adding another puzzle type. How do you start? What happens when you leave halfway through? Is that hint free, or are you about to be shown an ad? These turned out to matter a great deal to the person actually playing.The old page on this site described an earlier version. It had the wrong mascot, the wrong payment promises, and an account of adaptive difficulty that the code didn’t support. I have updated it. If somebody reads the page and then installs the app, they should recognise the thing they were told about.One tester reported that streaks weren’t working. The problem wasn’t a single counter. The screens were using the same word for different measures of activity, and some of the statistics mixed puzzles of different sizes. A quick 4×4 lesson could become your best time beside a 9×9 solve. That is technically a minimum. It is also a fairly useless thing to congratulate yourself on.The progress screen now distinguishes the measures and uses 9×9 solves for the relevant time statistics. The clock also pauses when the app goes into the background. Leaving the app open and returning later shouldn’t turn the time you spent doing something else into a very slow Sudoku solve.
Starting, leaving, coming back
The game-start screen had a large Play button that did nothing. I wish that sentence were about somebody else’s app. The working button was further into the setup. It now starts the recommended puzzle, or uses the size and difficulty you last played for that variant.There were smaller frustrations around leaving too. Android’s Back button closed the app from screens where it should have gone back. Leaving a game could lose the last few seconds of moves because saving was delayed. Both have been changed. These are the sorts of faults I tolerate while developing because I know how the app is supposed to behave. A player has no reason to tolerate them.The feedback form was worse: it showed a message as sent when the delivery service had refused it. A successful-looking screen had concealed a failure in the exact mechanism I was relying on to find failures. That path was fixed and checked with a message from the phone. I am rather more interested in seeing the message arrive now than in seeing the button turn green.
Teaching the first puzzle
The latest beginner route starts with a fixed Classic 4×4 introduction. It explains rows, columns, boxes, and input, then walks through four placements before moving into ordinary practice. Somebody who already knows Sudoku can take a different route. The point is to stop treating those two people as though they need the same first screen.Hints already have three stages: where to look, the technique, and the move. The label now says FREE while the next hint is a clue. The cost appears when the next step is the answer. Putting an ad badge on a free teaching feature was an impressively effective way of persuading people not to use it.The dark theme needed work as well. Some dots, letters, and arrows were too close to the background colour to read comfortably. They now use theme-aware colours. I have also added semantic grid controls and improved core contrast, though a complete screen-reader pass across every variant is still broader QA than the checks completed so far.
What’s still open
I found a more substantial engine limitation during the review. Several generators checked uniqueness under Classic rules when removing clues. That can leave a variant puzzle solvable without using the extra constraint. More puzzle types won’t solve that. Changing the generation method also affects seeded daily puzzles, so it needs a deliberate change and compatibility work rather than a quiet patch.I’m not calling the difficulty adaptive. A recorded player profile does not become adaptive behaviour simply because the profile exists. The technique analyser also needs to say when it couldn’t finish instead of confidently naming a technique as though it had solved the puzzle.The current Android release candidate is 1.8.1, versionCode 12. New purchases are disabled while the merchant setup is unresolved. Advertising and feedback have age and consent restrictions; they aren’t reasons to advertise the app as collecting nothing. The native iOS prototype is parked. Closed testing being nearly over doesn’t mean either platform has a finished public rollout.I have a list of things I’d like to add: better explanations, sharing puzzles, perhaps a way to choose only the difficulty and let the app pick the variant. For the moment, the work is getting the existing experience to behave as a player reasonably expects. There has been plenty to do there.
