Nobody Built the Right Tool for Literary Fiction. So I Did.

Writers who are serious about literary fiction share something that's hard to explain to people who aren't: a particular quality of caring. Not just about whether the book gets finished — about whether the scene is alive. Whether the sentence is true. Whether the thing they were reaching for in a chapter actually made it onto the page or stayed somewhere in the space between intention and execution.

This kind of writing takes years. It asks for the kind of attention that becomes its own way of being in the world — reading differently, noticing differently, caring about craft in a way that doesn't switch off. Writers like this read obsessively. They hold other books up to the light. They spend time on a paragraph that most readers will pass through in seconds because they know what that paragraph is carrying.

And when they sit down with their own work and need to know whether it's working — they have almost nowhere to go.

The gap is real and it's large

I was writing my own book when I started looking for what I needed. Not a writing partner. Not someone to generate sentences for me. A reader — someone who could tell me whether the chapter I'd just written was doing what I needed it to do at that point in the arc. Whether the voice was holding. Whether the slow scene I'd written was earning its slowness or just sitting there.

What I found was a gap so wide it surprised me.

There are no tools built for serious literary editorial feedback. The market has produced grammar checkers, readability scorers, AI writing assistants that help you write faster. None of these are the thing. They're solving a different problem for a different kind of writer — one who is writing at volume, who needs surface corrections, who wants to produce more words faster. That writer exists. That's fine. But that isn't who is writing literary fiction with real ambition.

For the human editor route: the real thing, valuable, and genuinely hard to access. A good developmental editor who understands your genre and respects your voice costs several thousand pounds or dollars for a novel-length manuscript. It requires finding the right person among hundreds — the directories exist, but quality and sensibility vary enormously, and you don't fully know what you're getting until you've paid for the sample edit and read what came back. Good editors are booked months out. And most work with complete manuscripts, which means a writer mid-draft — trying to know whether chapter eight is working before chapter nine compounds the problem — has essentially no option at all.

So writers do what they can. They share chapters with friends who read sensitively. They join writing groups. They live with the uncertainty and push forward hoping the problems they sense aren't as bad as they fear.

What happens when you try the AI tools

At some point, every writer who's been in this position opens a chat tool and pastes in a chapter. I did this. Most writers I know have done this.

The feedback comes back confident, comprehensive, and frequently wrong in ways that land somewhere between useless and actively damaging.

A chat tool doesn't know your book. It has no record of the promises your narrative has made, the patterns you've established, the intentions you've confirmed across months of work. It reads the text in front of it, applies everything it knows about writing in general, and produces an answer. The answer is confident regardless of whether the confidence is warranted.

What this looks like in practice: your deliberately slow scene gets flagged for pacing. Your unreliable narrator — carefully constructed, every inconsistency intentional — gets flagged for inconsistency. Your chapter-one ambiguity, doing exactly what it's supposed to do for the genre you're writing in, gets flagged as unclear. The feedback is written in the tone of someone who has read your chapter carefully. It is the output of a system that read your chapter without knowing what your chapter is part of.

The deeper problem isn't that the feedback is wrong. It's that the tool can't tell you when it doesn't know. It has no mechanism for saying "this might be intentional — I'm not in a position to judge." It fills every gap with confidence. For a writer who is still figuring out which of their instincts to trust, who is genuinely uncertain whether the thing they sensed was off is actually off — this confident wrongness is genuinely harmful. It teaches you to distrust the choices that were right.

Literary fiction asks for craft that general tools simply aren't built to see. An unreliable narrator is a technique, not a flaw. A slow scene can be grief rendered in prose, not failed pacing. A line of dialogue that sounds unnatural may be the most psychologically precise moment in the chapter. Seeing any of this requires knowing the book, knowing the genre, knowing what the writer was reaching for. The tools that exist right now don't know any of it. They read without that knowledge and respond as though they have it.

Why this matters enough to build something

I built Giluma because I needed it and I couldn't find it.

What I needed was something that reads the way a specialist editor reads: with knowledge of this specific manuscript built up over time, with enough understanding of the genre to know the difference between a convention and a flaw, and with enough honesty about its own limits to not flag what it doesn't understand as a mistake. Not confident noise. A reading that earns its feedback.

Giluma's layers are specialists. Each one reads for something specific — structure, voice, subtext, narrative energy, reader experience — and each one is trained to hold the boundary of its authority.

The hardest thing to build wasn't capability. It was constraint: making each layer honest about what it doesn't know.

The book accumulates too. Every session builds on what came before. Intentions the writer has confirmed, patterns established, choices made deliberately — these are held, so the reading calibrates to this manuscript over time instead of starting from zero every time.

I'm not going to tell you this replaces a great human editor who is right for your book. It doesn't. But a great human editor who is right for your book is one of the hardest things to access in literary fiction right now. What Giluma can be is the reader who is there when you need to know whether something is working — mid-draft, finished draft, any stage — without the cost and the wait and the mismatch risk that makes real editorial support inaccessible to most writers most of the time.

Writers who care about literary fiction — who read the way people read when books have become necessary to them, who are writing something they hope will matter to someone — have deserved better tools than the ones that existed. The books they're trying to make deserve to be built with the right kind of help.

That's what I'm trying to build. It's not finished. It may never be finished, because this kind of problem doesn't fully resolve. But the gap is real, and it needed someone to start filling it.