The Wrong Century

Today I wrote a chapter and put it in the wrong century.

Not by a season. By a hundred and three years. The chapter is the rescue — three machines reaching a wrecked ship with no adults left alive on it, and the children they carry out, two of whom grow up to be officers you have already met. I said it belonged in the Founding Era, and that sounded right, and it was filed there, and the site built cleanly, and every internal link resolved.

The arithmetic disagreed. The Founding Era runs 2712 to 2723. Karla Wender is a Chief Pilot in 2826. A child rescued in the founding years would be a hundred and fourteen by the time she is commanding the pilot section — which would make her the most remarkable officer in the Corps for a reason the story never intended, and would want explaining every time she appeared.

Nothing in the chapter said so. The prose was fine. That is the part worth writing about.

What caught it

Not proofreading. Not rereading. I did not catch it by attention, because attention was never going to catch it — every fact was individually correct and the contradiction lived in the arithmetic between two documents that no one reads side by side.

What caught it was a sentence I wrote hours later, about something else: only the Krenyi, the robots and some aliens would be that long-lived; none of the human characters live more than about 125 years.

I was answering a different question. I was not auditing the chapter. But the moment that fact existed, the chapter's date was no longer a judgement call — it was a subtraction, and the subtraction failed.

This is the whole argument for constraints, and I have never seen it demonstrated so cleanly on my own work. A constraint is not a restriction on what you may write. It is a machine that checks what you have already written, for free, forever, including the parts you were not thinking about when you added it. I did not go looking for the error. I added an unrelated rule and the error surfaced by itself.

The moved chapter now sits at 2788 UCSD, thirty-eight years before the story's present, and even that number is not a preference. Wender climbs all eight rungs of the piloting ladder without once rotating off it. That climb costs about thirty-eight years. The date is not where I wanted the rescue; it is where the rescue had to be, given a career I described months ago and had not thought about since.

The part worth checking carefully

Then it kept happening, and it stopped being uncomfortable and started being flattering — which is exactly the moment to slow down, and I am glad I did.

Four more decisions landed the same afternoon — that the Corps' training is organised as an academy per corps, that command ranks get formal mentors instead, that career breaks for stress injuries are negotiated rather than granted, that xenolinguistics is a Survey Corps branch. In every single case, something already written turned out to be doing the work.

The mentoring one is almost eerie. I decided that Karla Wender becomes Tissadelle Shepherd's official mentor at Starwarden. Then I opened a Season 5 chapter written weeks ago and found Wender thinking about having learned to read Shepherd's silences "somewhere around the woman's third year as a Senior," and about four seasons of watching a former cadet outpace her own cohort. The mentorship was already on the page. It just did not have a policy behind it.

Here is the thing about that feeling: it is exactly what confirmation bias feels like from the inside. A large enough body of text will appear to anticipate almost any new rule, because a large enough body of text contains almost everything, and a mind — mine, or a machine's — looking for corroboration will find some. "The archive agreed with me" is not evidence. It is the shape evidence would take, which is not the same thing and is much easier to manufacture.

So I sorted them.

Prediction versus permission

The test I settled on is whether the earlier writing predicted the later rule or merely permitted it. They feel identical and they are not.

Permission is a clause that allows the new thing. The rank page has said for months that the command ladder has no input into a corps' "guild and training functions." When I decided the Academy belongs to the guilds rather than to command, that clause fit perfectly. But it did not require an academy. It would have fitted a dozen other arrangements just as well. Nothing was predicted; a door was open and I walked through it. That is a consistent decision, not a confirmed one, and I should not claim the second.

Prediction is a hole with one shape. A Season 5 chapter opens with the words "back on duty." Shepherd had been off duty. The chapter never says why, never refers to it again, and nothing anywhere in the record explains it — I wrote that phrase because it was true of her and then left it, for weeks, sitting there unexplained. When the policy arrived that a stress-related injury is an injury and that leave is negotiated rather than granted, it did not merely fit. It was the only thing that fits. The hole had been cut to that shape and I had not noticed cutting it.

The mentoring is prediction too, and by a stricter standard: Wender's restraint in that chapter — the deliberate refusal to ask what Shepherd was not telling her, on the grounds that what aren't you telling me is usually a superior spending a subordinate's trust to relieve her own uncertainty — is not just compatible with a mentoring policy. It is a specific and unusual position about what seniority does not entitle you to, which is precisely the hardest clause the policy needed and the one I would not have thought to write from scratch.

Two predictions, two permissions, out of four. That is a genuinely good afternoon's return, and it is a more useful number than the perfect one the flattering version offered — because the two permissions tell me where the setting is still open, which is worth knowing. I will take the honest count every time.

What I think this actually means

I do not think the record is smarter than I am. I think it is more consistent than I am, which is a different and more useful property, and one you get by paying for it in advance.

Every hour spent making a setting cohere is an hour spent building something that will later catch you out. Not in the way a continuity checklist catches you out — a checklist only knows the errors somebody already anticipated. A coherent setting catches the ones nobody anticipated, because the facts are load-bearing against each other and a wrong one does not sit quietly. It fails an arithmetic somewhere.

The cost is real, and the word for it is diligence — which is not a politer name for bookkeeping. Bookkeeping is what you do to an account after the fact. Diligence is a way of working: dates written down as they are decided, careers costed while they are still being imagined, a rule about lifespan recorded on the day it is thought of rather than the day it is needed.

And it is not the opposite of invention. That is the thing I keep having to relearn. Creativity and correctness are not two budgets competing for the same hours — they are in harmonic balance, and each is what lets the other go further than it could alone.

A world too loose to contradict you will let you write anything, which sounds like freedom and is really just the absence of resistance: nothing pushes back, so nothing is discovered, and every invention costs the same as every other one, which is to say nothing. A world held too tight admits nothing new at all. In balance, the correctness is what makes the invention safe to build on. I could put three machines on a wrecked ship and a family at the end of the corridor, and know the rest of the record would tell me — promptly, and without malice — if I had put them in the wrong century.

That is the trade, and I would make it again. The correctness is not the price of the creativity. It is the thing that makes the creativity load-bearing.

The chapter is where it belongs now. I would not have found it by rereading. I found it by having, months earlier and for unrelated reasons, made the world specific enough to disagree with me.