I thought remembering the facts would be the difficult part. It turned out that remembering the people was harder.
When you invent a character, you tend to know things about them that never make it onto the page.
You know what they find funny. You know which kinds of problems make them uncomfortable. You may have a sense of how they would react to a friend being hurt, a ridiculous proposal or a situation they cannot control.
The reader doesn’t know those things yet.
An AI writing assistant doesn’t necessarily know them either.
At first I could describe a character with a handful of traits and receive something quite convincing. But the longer a story became, the more I noticed how much personality could disappear between scenes.
Someone who was wonderfully peculiar in one chapter might become surprisingly ordinary in the next.
So I began building more detailed character references.
More than a collection of adjectives
A character bible can include the obvious things: background, relationships, abilities and appearance. Those details are useful, particularly when a book has a large cast.
But I became much more interested in recording how a character actually behaves.
How does someone joke when they feel safe? What changes when they’re frightened? What kind of help do they offer? What do they notice that other people overlook?
I can describe Dude’s unusual priorities, Chef’s approach to practical problems or the ways Hex communicates through movement. Yet a description is only the beginning.
What I really want the AI to understand is how those differences become visible when the characters are placed in the same situation.
A cast of distinctive people should produce distinctive disagreements.
They should not all become variations of the same sensible protagonist.
When consistency becomes a problem
The more information I collected, the better some things worked.
And then another problem appeared.
If the AI tried too hard to demonstrate every character trait, the personalities became exaggerated. A humorous person had to make a joke in every scene. Someone known for caution could barely take a risk. A distinctive mannerism began appearing so regularly that it felt less like personality and more like a reminder.
A character bible is supposed to help preserve someone.
It shouldn’t prevent them from changing.
People behave differently depending on who they’re with, what they have experienced and what they stand to lose. Fictional people need the same freedom.
Sometimes the most convincing thing a normally funny character can do is remain silent.
Giving the world room to surprise me
I have encountered a similar problem with worldbuilding.
Rules matter. A fictional ability becomes much less interesting if it can do whatever the story requires. A danger loses its meaning when an entirely new solution can appear without consequences.
But I don’t want every detail of the world explained in advance either.
I enjoy exploring the possibility that the characters misunderstand the place they inhabit. Something can look like a monster and turn out to be a worker. An apparently simple machine may raise uncomfortable questions about intelligence and autonomy. A strange routine may be more meaningful than anyone initially realizes.
The development of the husks, loopers and Residency workers is one example of how an idea can become much richer when I resist settling on the first explanation.
That is the balance I keep trying to find.
I want the world to behave consistently enough that its surprises matter, while leaving room for discoveries that change how both the characters and I understand it.
The character bibles and world notes help me remember what I’ve already established.
The storytelling begins when I ask what those things might mean to the people living with them.