I have been thinking about what I could have done better with my novella “All Our Eyes”.

So what worked?

The story had a 3-levels of narrative. I wrote them from the inside out and this worked well. The second level of the narrative had a lot of cross-referencing and writing that in was a lot of fun. The outer layer was meta, and that too was easy/fun.

When I had to change something major, like delete chapters (I deleted at least two) and reorder chapters (e.g. act 2), it caused a ton of pain with cross-referencing. A more solid plan would have allowed these major structural changes to have occurred earlier, perhaps.

The story also had major plot points and a progression for the character from enthusiastic student to obsessive investigator. I divided the story into three acts, and carefully laid out the plot points and progression in behaviour and how it manifested. This generally worked okay as well.

Another thing, I sat with a draft of the book for a long time. Perhaps 6 weeks. This is much longer than any of the 40+ books I’ve written in the past. I normally finish, edit, ship, typically because I’m on a deadline and the market is pulling the book out of me. Here, I wanted to sit with the story and iterate the plot until it was more interesting. And this worked. I continued to add little devices to the story all the way up until the end, and it made the story incrementally better. I was reaching diminishing returns by the end though, time to ship.

I write in markdown and in the early stages I used a static site generator to navigate and read drafts. This worked. It allowed me to have many “meta” directories full of “meta” content about the story, read and navigate it easily and iterate things. Perhaps it was too much stuff and it slowed me down. At the time, it felt like it helped. I could also easily re-read chapters and reorder them easily. I would probably do this again. It felt more messy than using a “book structure” from the beginning, allowing me to really hack on structure without fear.

The first act is slow, the second fast, and the final faster. Whenever I start reading the third act I have to finish reading it. It pulls me through. I’m happy with this, although I know the slowness of the first act will be too boring for most. Oh well.

Many documents are referenced throughout the story. In some cases I wrote the entire document, then had the main character reference parts of it. This worked well. Next time, I would do this for every document.

For example, every newspaper article, redacted document, researcher paper, book chapter. I would write them all and then have the main character reference them. It results in a much fuller experience when reading. You can feel the full document sitting just beyond reach. It matters.

I had a story plan, plans change, that’s fine.

What I should have done better is really hammer on the plan. Iterate it for way longer than I did.

I should have fed it into an LLM and as lots of questions like:

  • “what else?”
  • “what could I do instead?”
  • “how could I add more dread? or fear? or pace?”

Specifically the plan for each “chapter”. I should have hammered on the plans for each chapter for days before writing. The writing is fun, the plan is boring and the plan makes the writing feel stale, but it’s so much harder to re-write or re-plot a chapter after it’s written. You have to throw out so much and it sucks.

So next time, go fractal. High level plan, act level plan, chapter level plan, scene level plan, and hammer on each until they are exciting, awesome, compelling, amazing. Until they ALL pull you along from beginning to end. Iterate on each plan with LLMs for days.

Another failing is not having a crystal clear idea of what other people/organization said and did. Even if they are off screen and we only see the effects or second hand reporting. If it’s not clear to me, it’s not clear to the reader and a confused reader will stop reading. I know I do.

I HAVE TO KNOW. It is so obvious when I’m just bluffing my way through a scene or chapter because I have not thought deeply enough about what happened or what was said off screen. I hate that. I need to know, then have the chapter bounce off the effects.

The risk, of course, is that I will want to use this stuff. To move off screen to on screen. Slowing everything down.

I learned that LLM copy editing is good, but requires iteration. Proofreading is fine, but not super reliable.

Nothing beats a slow manual re-reading again and again.

I also learned that it is totally fine to make a small one-off book for myself and my son. No one to impress or make happy here except us. More of this.

Writing for pure fun. Not a business. Not for a market. Hobby. Joy.

And it was a lot of fun. More!

I’m going to explore more “early Australia” novellas. Also shorter, aim for 100 pages. And try and make the story faster, more compelling from the jump.