Silence the inner critic. Get words on the page. Worry about structure later. For a short story or even a standalone novel, this advice has merit. The momentum of creation can carry you through inconsistencies that revision will later resolve.

For a nine-book series spanning two million words, this advice will kill your project slowly, and in ways you won’t notice until it’s far too late to fix.

I learned this through my own failures and through watching other writers’ series collapse into contradiction around book four. The dawning horror of realising how many details I’d already forgotten about my own world before finishing the first draft. Character eye colours. Magic system constraints I’d established and then violated. The name of a sword that apparently changed between chapters three and eleven. (I fixed it, eventually.)

The problem is scale. Human memory - even the kind that can recall the exact wording of a paper you read six years ago - cannot reliably maintain thousands of interlocking details across years of creative work. You will contradict yourself. The question is whether you catch it before readers do.

So before I wrote a single scene of my current project, I built what I came to call my series vault. An Obsidian folder containing 175+ notes organised across fourteen domains, totalling over 60,000 words of worldbuilding, character development, system design, and process documentation.

Sixty thousand words before word one of the actual novel.

This sounds like procrastination dressed in productivity clothing. It might be. The jury’s still out on whether I’m being systematic or just avoiding the part where I have to commit to actual prose and discover what I don’t know. But the vault exists, and it’s already saved me from myself more times than I’d like to admit.

The vault isn’t bureaucracy. It’s an immune system.

You write a detail in chapter three that contradicts something you established in chapter one. Caught immediately: trivial fix. Caught during revision: annoying but manageable. Caught after publication: you’re writing errata and hoping readers forgive you. (They don’t. They keep spreadsheets. Some of them keep better spreadsheets than you do, which is frankly terrifying.)

Now multiply this across nine books and a decade of reader experience. The vault catches contradictions before they metastasise. When I write a scene where a character references their childhood, I can check the vault to confirm what I’ve already established - or note that I haven’t established anything yet, and this scene is now setting canon. When I introduce a new constraint, I can verify it doesn’t violate constraints I’ve already locked in. It usually does. The vault tells me this before I’ve written ten thousand words that depend on the violation.

This is where the obsessive documentation starts paying for itself: the vault doesn’t constrain creative freedom. It enables it.

Without documentation, every creative decision carries invisible weight. You’re not just choosing what happens in this scene - you’re potentially contradicting something you’ve forgotten, closing off possibilities you’ll need later. The uncertainty is paralysing if you’re paying attention, and much worse if you’re not.

With documentation, you know what’s fixed and what’s flexible. You can make bold choices because you understand their implications. You can plant seeds for future payoffs because you have somewhere to record what you’ve planted and why. External memory supplementing the internal memory that has already proven itself unreliable.

The vault transforms “I think I mentioned somewhere that...” into “The vault confirms that in chapter four, I established...” That’s the difference between hoping you’re consistent and knowing.

At this point, the obvious objection is that 175 notes across fourteen domains is overkill for most projects. And it is. That’s the scale a nine-book series demands. A duology might need twenty notes. A single complex novel might need only a handful of documents covering the essentials—character facts, timeline, system rules if you have systems. The scale changes; the principle holds. Start with what you actually need. The vault will tell you when it needs to grow.

The other objection is harder to dismiss: this still sounds like procrastination, just with better justification. Fair. But the vault I started with isn’t the vault I have now. Drafting reveals gaps - things I should have documented but didn’t. Revision reveals contradictions that the vault itself contained. The documentation and the writing evolve together. This isn’t about building a pristine reference before you’re allowed to create. It’s about maintaining external memory that grows alongside the work.

Building a vault takes time. For my project, it took months before drafting began, and it continues to evolve with every chapter. Whether the initial investment was necessary preparation or elaborate avoidance behaviour remains genuinely unclear to me. But it’s paid for itself in contradictions prevented, revision time saved, and the confidence that lets me write quickly because I’m not second-guessing every detail against a memory I no longer trust.

If you’re planning an ambitious project - or even a moderately complex one - consider building your vault early. The voice will come. The vault ensures it has something consistent to say.

In subsequent essays, I’ll explore what actually goes in the vault: what domains to include, how to maintain living documents that don’t quietly become fiction themselves, and how to extract relevant information when you need it. The principles matter more than the specifics, but the specifics help make the principles concrete.

For now, the core insight: documentation isn’t the enemy of creativity. It’s the foundation that makes sustained creativity possible without losing your mind.

Build the vault. Then write. Then build more vault. Then write more.

Maybe you’re planning a multi-book series, and the continuity alone feels like it could crush you. Maybe you’re trying to sustain a distinctive narrative voice across hundreds of thousands of words without it flattening into something generic. Maybe you’re curious about AI as a tool, but wary that it might strip away everything that makes your writing yours.

This isn’t a site about writing faster, “optimising” creativity, or outsourcing craft.

It’s about precision tools for ambitious writers who refuse to compromise on quality. Or, to put it more honestly: writers who’ve already given up on sleep, functional social calendars, and having bookshelves that don’t sag ominously in the middle, and have decided that quality is the one hill left worth defending. Possibly to the death. The caffeine intake suggests this may not be metaphorical.

What I’m interested in exploring here

This space exists because certain problems only show up once you move past single-book thinking, and most writing advice stops exactly where those problems begin.

The topics tend to cluster around:

Voice and perspective integrity
How first-person narrators quietly drift toward omniscience. How you end up unconsciously channelling your influences instead of synthesising them. Why your prose suddenly sounds like Douglas Adams wrote it after a very long day, and whether that’s homage or just theft with extra steps.

Mystery and information management
When to answer a reader’s question vs when to complicate it. How to track what each character knows against what the reader suspects. How to resist the overwhelming urge to reveal everything in Chapter 2 just because you’re so bloody pleased with your own cleverness.

Systematic quality control
How to audit for invisible repetition before your beta readers get quietly bored. What’s worth tracking in a continuity document vs what you can safely reconstruct on demand. At what point your continuity tracker becomes longer than your actual manuscript, and whether you should be concerned about that. (Spoiler: probably.)

Creative–technical balance
When systematisation helps vs when it starts strangling spontaneity. How to preserve human imperfection in prose without it reading as careless. How to explain to normal people that yes, you did just spend four hours building a spreadsheet to track metaphor frequency across eighteen chapters, and no, you genuinely don’t see the irony.

AI shows up in some of this work because it’s useful for certain kinds of problems - the tedious, structural, pattern-matching ones. In other areas, it’s entirely absent. The constant is that judgment, authorship, and creative responsibility stay human.

A note on AI collaboration

I want to be clear about something up front:

AI is not writing my fiction. I am.

What it can do - when handled carefully and kept on a very short leash - is take on tasks that would otherwise consume time I’d rather spend on voice, character, and emotional truth. Things like:

• Scanning for close-proximity repetition (turns out I use “precisely” precisely too often)

• Flagging POV violations before they metastasise into structural nightmares

• Cross-checking continuity details across hundreds of thousands of words (Yes, Enkidu’s sword had a name in Chapter 3. No, it shouldn’t have a different name in Chapter 11. Yes, I did that anyway. We fixed it.)

Think of it less as a ghostwriter and more as an extremely literal research assistant who never gets tired, never judges your poor life choices, and occasionally needs reminding that “make this better” is not, technically, actionable feedback.

Used badly, AI flattens prose into algorithmic beige and erodes your judgment. Used well, it works a bit like a structural editor who’s fast, cheap, and operates under constraints you define - not some generic set of “best practices” scraped from the internet’s collective mediocrity.

If that makes you uncomfortable, fair enough. If you’re curious what careful, constrained collaboration actually looks like - with all the failures, course-corrections, and places where it absolutely doesn’t work - this might be useful.

How this space works

This isn’t a feed, a diary, or a productivity system pretending to be profound.

Some posts are focused essays on a single craft problem. Others are messier - exploratory notes, half-formed frameworks, post-mortems on techniques that sounded brilliant at 2 AM and turned out to be structurally incoherent by morning. When AI is relevant, I’ll be explicit about what it’s doing and what it categorically isn’t. When it’s not relevant, I won’t drag it in by force just to stay on-brand.

Subscribe if you want. Bookmark the page and check back when a topic sounds useful. Either works.

Long-form fiction is hard. Managing complexity without losing your voice is harder. AI collaboration that doesn’t flatten everything into corporate-speak mediocrity is harder still.

There’s very little value in pretending otherwise.

This is one attempt to document what actually helps - and what absolutely doesn’t.