Withhold the answer, never the question

I wrote a rule down on 28 July 2026, after a chapter had failed it. Withhold the answer, never withhold the question.

The distinction sounds obvious until you are the one who has blurred it. A mystery is a question a reader can hold on to. They know what happened, they do not know what it means yet, and carrying that question is the enjoyable part. A fog is different. A fog is what happens when a reader reaches the end of a chapter and cannot say what occurred in it. That is not suspense. It is the reader starting to suspect that nothing occurred, and that suspicion is much harder to undo than a wrong guess would be.

The chapter that failed the rule had withheld a fact the scene needed, on the theory that holding it back would build tension. It did not. Read back cold, there was no plain sentence to be had about what had gone on in it, so I rewrote it. I have written before about choosing whose eyes a chapter runs on to control how much a reader gets. This was a different failure. It was not about distance. It was about a chapter that simply did not say the thing it needed to say, from any distance.

The mistake under the mistake

The rule I had been applying too flatly was show rather than tell. I had read that as licence to withhold instead of reveal, and those are not the same instruction. Omitting a fact and dramatising a fact are two different acts wearing the same clothes. Show rather than tell is not show and then never tell at all, and the distance between those two turned out to be exactly one unreadable chapter.

So the rule became a test, and I run it at the end of every chapter now. Can I state, in one plain sentence, what happened in this chapter. Not what it meant. What happened. If I cannot answer that, the fault is the chapter, not the reader, every time, and I go back in.

The cost does not stay in the chapter that spent it

A second rule followed on 18 August 2026, from reading thirty-six chapters straight through rather than one at a time as I had drafted them. A chapter can pass the one sentence test entirely on its own, and the run can still tire a reader out, because the effort of piecing things together is not paid per chapter. It is paid across chapters, and it adds up. After a chapter that has asked a lot of a reader, the next one needs to hand its new information over in plain sight, with nothing left to assemble. Two demanding chapters in a row is something to watch. Three in a row is a problem, whatever each one looks like taken on its own.

That is a pacing problem, not a clarity problem, and I recognise the shape of it from level design before I recognise it from prose. A single hard encounter is not the thing that breaks a run. A hard encounter placed directly after another hard encounter is, because the player is still paying down the first one when the second bill arrives. The difficulty curve is readable one encounter at a time and still wrong, because nothing on that scale is measuring the encounter before it.

One distinction makes the second rule liveable. A mystery that has a name, one you know you are holding, costs nothing. You can carry a named question for twenty chapters quite happily. It is the unnamed kind, the question you do not even know you are meant to be asking, that wears a reader down without ever announcing itself. By the time it announces itself, the reader has already decided the book is hard work rather than that one chapter was.