Building Fantasy Novel Studio

What it is

Fantasy Novel Studio is my answer to: can I use AI to help write a full-length, professionally polished fantasy or sci-fi novel? Instead of one chat thread, it’s a studio built from Markdown, YAML, and Cursor agents. Working drafts stay out of the Novel Bible until I approve them. I stay Creative Director. This isn’t a public app yet. The system still had to be real enough to write Book 1 of Layers of Reality.

How I built it

Coming from a non-software background, I grew the studio through experimentation in Cursor: folders, rules, checklists, and the idea that AI doesn’t get silent authority over what becomes canon in the novel. Promotion into the Novel Bible only happens when I approve it.

I also learned that I didn’t need every agent for every task. It’s better to use specialized agents for specific jobs: a lighter set when I’m drafting, a stronger set when I’m planning, a pickier set when I’m checking continuity, and a fuller set only when I truly want everything. Matching the agents to the work kept the studio useful instead of overwhelming.

Book 1 was the stress test: outline expands, deepen passes, continuity checks, beta, and cleanup. Continuity quietly broke between chapters, the kind of slip a normal chat conversation is bad at catching. Making those breaks visible was the whole point.

What I learned

What stuck with me wasn’t writing theory. It was how to build with AI for long-form creative work.

I learned how to organize a project in Markdown and YAML so agents follow a process instead of inventing one mid-chat. I learned that specialized agents for specific tasks beat using every agent for every job. I learned how to put an approval gate between “AI suggestion” and “official canon,” because chat tools are happy to rewrite anything in reach. I learned that you can pressure-test a system by writing a real book with it before it graduates to a public app. I learned how continuity can break between chapters, and how to make those slips visible instead of hoping the next chat remembers.

The biggest lesson: if a project has to stay consistent over time, don’t let AI rewrite the official version on its own. Put up checkpoints where it has to check in with you for approval before anything becomes canon.