🃏 Classic1979Howard Garns / Dell MagazinesFIDDLYSudoku rebuild deck
FIDDLY
Unless you explicitly ask for a solution counter after every single clue removal, the generator quietly ships ambiguous puzzles that look perfectly normal until a player gets stuck.
You will get it, but one specific system will fight you.
Published as Number Place in Dell Pencil Puzzles and Word Games in May 1979, and Garns got no byline for it. Generation is the entire job here. An agent gives you a handsome grid, a competent notes mode, and a generator that quietly emits ambiguous puzzles, because uniqueness checking is a step you have to request by name. Spend your review time on the solution counter and on whether the four difficulty labels correlate with anything real; the rest is layout and keyboard handling.
- Difficulty
- FIDDLY · 2 of 5
- First playable
- An evening
- The original
- 1 team-years
- Released
- 1979 · Classic
- Genre
- Logic Puzzle
Build a Sudoku app as a single self-contained HTML file. DOM grid, no build step, no dependencies. Requirements: - Generate puzzles at runtime, never ship a fixed list. Fill a complete valid grid with randomised backtracking, then remove clues one at a time. - The load-bearing rule: after every single removal, run a solver that counts solutions and aborts at two. If the grid now has more than one solution, put that clue back and try a different cell. A generator that simply blanks N random cells produces ambiguous puzzles, and players will get stuck without ever understanding why. - Rate difficulty by which techniques a puzzle requires — naked singles only, hidden singles, pairs, anything beyond that — not by how many clues survive. Four levels. - Pencil marks: a notes mode placing small candidate digits in a cell, auto-cleared from the row, column and box when a digit is committed there. - Highlight the active row, column and box, plus every cell holding the selected digit. Flag conflicts on entry but allow the player to leave them on the board. - Keyboard: arrows to move, 1-9 to place, backspace to clear, space to toggle notes, and the same digits with shift held to toggle a single pencil mark. - Timer, full undo stack, and a check button that reports only whether errors exist, never where. Definition of done: I generate a hard puzzle, feed the givens to any external brute-force solver, and it reports exactly one solution.
- Hand-made Nikoli puzzles, where the clue pattern is symmetric and the solve path was designed by a person
- Difficulty labels a human actually calibrated against real solvers rather than a clue count
- Newsprint, a pencil, and the specific satisfaction of a coffee-ringed grid on a commute
- Any claim on the word itself, which Nikoli holds as a trademark in Japan
Difficulty graded by the technique a solver needed
Clue count, enemy count, level number — every lazy difficulty metric measures the content instead of the player. Grading by which inference a solver actually had to reach for is the only difficulty number that means anything, and the trace you get out of it doubles as a hint system for free.
ScopeA weekend— build just this, not the game
Build a puzzle grader as a single self-contained HTML file, DOM grid, no dependencies. Implement a solver for a 9x9 constraint grid that uses only named human techniques, tried in order of increasing cost: naked single, hidden single, naked pair, pointing pair, box-line reduction. Each pass applies the cheapest technique that fires and logs which one and which cells. If nothing fires it stops and reports the grid as unsolvable by these rules — no brute-force fallback, ever, because the fallback is what destroys the measurement. The output is the deliverable: a step-by-step trace, and a grade derived from the hardest technique the solve required plus how many times it was needed. Never from how many cells were blank. Include a paste box for a grid string and a crude generator so I can grade a hundred and see the spread. Definition of done: two grids with an identical number of givens come out at different grades, and the trace tells me exactly which deduction separated them.
Not a clone — Sudoku re-themed and cut down to something you can finish. The angle is the important reel: it is what turns a project into a weekend. 24,360 ways to land.
Once Teletype
Sudoku, but text-only, in a terminal, set in a fairy tale gone wrong. Rendered in ASCII terminal.
Build a browser game called "Once Teletype". The pitch in one line: a logic puzzle in the spirit of Sudoku, but text-only, in a terminal, set in a fairy tale gone wrong. WHAT THIS IS Take the SHAPE of Sudoku — its core loop, and whatever makes that loop good — and rebuild it as your own game in a different world at a smaller scope. Use your own names, art and audio, which you want anyway: a reskin of somebody else's game is less interesting than a new one that works the same way. THE ANGLE — this is the important constraint, honour it above everything else Text-only, in a terminal. A monospace grid and nothing else. Forces the writing and the systems to carry everything, and the entire art budget goes to zero. This is what makes the project finishable. If a decision would push the scope back towards the original's, take the smaller option every time. THE THEME A fairy tale gone wrong. Rules of stories treated as physics — deals, thirds, true names — and a moral that costs something. The theme is not a coat of paint. Let it change what the mechanics mean — a reload, a health pack and a locked door should all be things that make sense in this world and nowhere else. LOOK ASCII terminal. A monospace character grid as the only renderer, colour via ANSI-style palette, and every animation expressed as glyph changes. REQUIREMENTS - Runs in a browser with no build step. Canvas 2D unless the angle genuinely requires 3D, in which case three.js. - Fixed-timestep update loop with interpolated rendering, so it behaves the same at 60Hz and 144Hz. - Every tuning constant in one CONFIG object at the top, exposed as live sliders in a debug panel. You will find the feel by dragging those, not by prompting. - All content — levels, entities, balance numbers — in JSON or plain data files, never inline in the logic. - Audio generated with the Web Audio API rather than asset files. - Respect prefers-reduced-motion: keep the fades, drop the shake and the parallax. BUILD ORDER 1. The core loop with placeholder rectangles, at the reduced scope the angle demands. No theme, no art, no audio. Make it fun as rectangles first. 2. The angle's consequences. Cutting to bots, or to turn-based, or to one level changes the design rather than just shrinking it — find out how and rewrite step 1. 3. The theme, as content and data. 4. The look, applied last as a rendering layer over a game that already works. Start by writing a one-page plan: the core loop, what the angle removes and what that frees you to do properly, and what you are deliberately leaving out. Wait for me to approve it before writing code.
Half the time one of these already does what you wanted, and the other half it saves you a week of solving a problem someone documented in 1997.
One count per game. No account — it is tied to a salted hash of your IP, which is not stored in a form anyone can reverse. Every build adds 1 team-years to the counter on the homepage.
Then try these
Questions
How hard is Sudoku to rebuild with an AI agent?
Yes — every game is buildable at some scope. The question is how hard, and Sudoku is FIDDLY (2 of 5). You will get it, but one specific system will fight you. Unless you explicitly ask for a solution counter after every single clue removal, the generator quietly ships ambiguous puzzles that look perfectly normal until a player gets stuck. Reckon on an evening to something you can actually play — not to something that matches the 1 team-years Howard Garns / Dell Magazines spent.
What makes Sudoku hard to rebuild?
Unless you explicitly ask for a solution counter after every single clue removal, the generator quietly ships ambiguous puzzles that look perfectly normal until a player gets stuck. That is why it sits at FIDDLY rather than a tier either side of it. If you want the idea without the project, the transferable part is difficulty graded by the technique a solver needed — a weekend of work.
How long does it take to build a Sudoku clone?
Wrong question, and it is worth saying why. Nobody rebuilds Sudoku — the original took Howard Garns / Dell Magazines on the order of 1 team-years, released in 1979 on Classic. What you can do is reach something playable and genuinely yours in about an evening, and then keep going for as long as it stays interesting. Any figure that claims otherwise is measuring a prototype and calling it a game.
What do I lose by building my own Sudoku?
Hand-made Nikoli puzzles, where the clue pattern is symmetric and the solve path was designed by a person. Difficulty labels a human actually calibrated against real solvers rather than a clue count. Newsprint, a pencil, and the specific satisfaction of a coffee-ringed grid on a commute. Those are the parts a prompt will not hand you.
Which AI coding agent should I use to build it?
Any of Claude Code, Codex or Cursor will handle this prompt. Paste it as your first message, let the agent scaffold the project, then iterate on feel — the second, third and tenth prompts are where a game actually gets good. Work in an empty folder or a fresh git branch so you can throw it away cheaply.
One email when new games go on the list. No other reason to email you, ever.