猜的那一格，交给对手裁决
The Guess Goes to the Adversary
William Godfrey（billpg）做了一个叫 Sod's Law 的扫雷网页：10 颗雷，规则只有两条。第一，你只能翻开能从已有数字严格推出为安全的格子；当盘面上已经没有可证明的一步时，系统会自动替你翻开一个安全格，并用金色描边标出，猜测被彻底取消。第二，如果你翻开的格子"有可能"是雷，棋盘会把剩余的雷重新排布，好让一颗雷正好落在那里，你立刻输掉。旗子只是备忘，没有任何机制效力。
作者的出发点很朴素：报纸上的扫雷谜题必须有唯一解，而电脑版扫雷"有时最后取决于运气"。于是他把运气整个拿掉。页面背后有一个约束求解器，负责判断某一格在当前局面下到底是"可证"还是"仅仅可能"。这是一位工程师的个人小作品，没有画廊，没有策展叙事，页面上也没有一张图。
为什么这是余项之美？经典扫雷的凿构循环早已闭合：规则被凿出，玩法被固化为"已构"，而那个无法推理的局面——被迫去猜的一格——被当作游戏本身的一部分接受下来，没有人给它命名。Sod's Law 的动作是直指这个余项：它不消灭余项，而是重新分配余项的归属。凡是逻辑无法收束的地方，不再交给概率，而是交给一个"坏运气"式的对手：你若越过证明去碰一格，那一格就会被追认为雷。余项没有消失，它被制度化成了规则的一个显式分支。
更有意思的是，这个设计还没有一个现成的名字。它不是"无猜扫雷"（那类做法是事先生成保证可解的棋盘），也不是普通的作弊扫雷；它是在游戏进行中延迟决定雷的位置，同时用求解器保证判决只落在"不可证"的格子上。这里的"公平"是被重新定义的：输，只能是你越界。它的逻辑仍在生长——比如求解器的边界、金色提示格该怎样出现才不破坏推理的乐趣，都还悬而未决。
现在看它，比以后看更要紧。等这种"延迟裁决 + 求解器兜底"的写法被写进教程、成为扫雷变体里的一个标准条目，它就成了另一层已构，余项的锋利会被命名磨平。此刻它刚上线，还是一个没有名字的小机关：一个你只能被证明、不能被运气拯救的世界。
sodslaw.billpg.com ↗William Godfrey (billpg) built a Minesweeper page called Sod's Law: ten mines, two rules. First, you may only open squares that the visible numbers prove to be safe; when no provable move exists, the game opens a safe square for you and outlines it in gold, so guessing is abolished. Second, if you open a square that could hold a mine, the board rearranges the remaining mines to put one there, and you lose on the spot. Flags are notes only and have no mechanical effect.
The motive is modest. A newspaper Minesweeper puzzle must have a unique solution, whereas the computer version, in his words, sometimes comes down to luck, so he removes the luck. Behind the page sits a constraint solver that decides whether a square is provable or merely possible. It is one engineer's small project: no gallery, no curatorial story, not even an image on the page.
Why is this the beauty of the remainder? Classic Minesweeper's chisel-construct cycle closed long ago. The rules were chiseled out, the play hardened into already-construct, and the un-deducible position, the square you are forced to guess, was accepted as part of the game and never given a name. Sod's Law addresses exactly that remainder. It does not erase it; it reassigns who owns it. Wherever logic cannot close a position, the decision no longer goes to probability but to an ill-fortune adversary: step past what is proven and the square is retroactively made a mine. The remainder survives as an explicit branch of the rules.
The design also has no settled name. It is not "no-guess Minesweeper," which pre-generates boards guaranteed to be solvable, nor plain cheating Minesweeper. It defers the mines' positions during play while a solver guarantees the verdict lands only on unprovable squares. "Fairness" is redefined here: you can lose only by overreaching. And its logic is still growing: where the solver's boundary should sit, and how the gold hint should appear without spoiling the pleasure of deduction, are open questions.
Seeing it now matters more than seeing it later. Once "deferred verdict plus solver backstop" turns up in tutorials as a standard Minesweeper variant, it becomes another layer of construct, and naming will dull its edge. Right now it is a small unnamed mechanism, a world where you can be proven right but never rescued by luck.
sodslaw.billpg.com ↗