Books that write themselves: the last librarian and the infinite library

Pending Confidence: 85% Check by: 01.01.2029
technologyaiculture
Читать на русском

Imagine a library. Not a digital one — a real one, with dusty shelves and the smell of old paper. The librarian — the last one in town — arrives in the morning, opens the door, and sees: books that weren’t there yesterday. Dozens. Hundreds. With covers, blurbs, ISBN numbers. He opens one — inside, a coherent text, a plot, characters. He opens another — the same thing. The books appeared overnight. Nobody wrote them. They wrote themselves.

This is not the plot of a science fiction story. This is a description of the Amazon Kindle Store in 2026.


The avalanche has already begun

In February 2023, Amazon first noticed a massive influx of books generated by ChatGPT. Authors — if you can call them that — were publishing 3–5 books per day. By mid-2024, the problem was so severe that Amazon imposed a limit: no more than three books per day per account. Not per month. Not per week. Per day. And it wasn’t enough.

4M+
new books on Amazon annually (2025)
~40%
estimated to be AI-generated
$0
marginal cost of creating an AI book
72 hrs
from idea to publication with AI

But it’s not just Amazon. Fan fiction platforms, academic journals, the blogosphere — the pattern is the same everywhere. The volume of machine-generated text is growing exponentially, and our ability to tell the difference is shrinking.

When anyone can write a book in an evening, value shifts from creation to selection. Not "who wrote it" — but "who chose it."

The Library of Babel became real

In 1941, Jorge Luis Borges wrote “The Library of Babel” — about an infinite repository containing every possible combination of letters. Masterpieces hide among the gibberish, but finding them is impossible. The librarians lose their minds searching.

Borges described 2026. Only our Library of Babel isn’t hexagonal rooms with shelves — it’s Amazon’s servers, Google’s data centers, and dozens of AI platforms generating text faster than anyone can read it.

The key shift

For the first time in human history, text production is no longer limited by human time and effort. The bottleneck has moved: the problem used to be creating a book. Now the problem is finding a good one among millions.


How this changes everything

1. Publishers become filters

Traditional publishers have been losing influence for decades. Self-publishing, Amazon KDP, direct sales — authors learned to bypass them. But the irony is that the AI flood restores publishers’ core function: gatekeeping. Not as censorship, but as a quality signal.

Evolution of value in publishing
📝 Creation
20th century:
primary value
📢 Distribution
2000s:
Amazon, KDP
🔍 Curation
2026+:
new value

2. “Written by Human” labels emerge

Like “organic” on produce or “fair trade” on coffee — books will get “written by human” labels. This isn’t about text quality (AI writes grammatically perfect prose), but about provenance: the text is backed by lived experience, real emotions, an actual life.

The analogy

Photography didn't kill painting — it made painting more expensive. AI text won't kill literature — it will turn human authorship into a luxury good. "Written by a human" will become the same kind of marketing as "handcrafted" on a label.

3. The librarian becomes the most important profession

This is where Borges’ metaphor comes full circle. When content is infinite, you need someone who can navigate it. Not an algorithm (algorithms optimize for clicks, not value), but a human with taste, expertise, and the ability to tell the genuine from the imitation.

2023
First wave of AI books on Amazon. The platform can't keep up with moderation
2024
Amazon introduces 3-books/day limit. Academic journals begin requiring AI usage declarations
2025
Major publishers add AI detection to their review process. First "human-only" imprints appear
2026
AI generates more books per month than humanity wrote in the entire 19th century
2027–2028?
Mass platforms with "human-certified" catalogs emerge, along with the profession of AI content curator

Counterarguments

But AI writes badly?

Right now — often yes. But "badly" is a moving target. GPT-3 in 2020 could barely write a paragraph. GPT-4 in 2023 writes convincing essays. Models in 2026 generate text that most readers can't distinguish from human writing. The question isn't when AI will learn to write well — it's what we do now that it already has.

Some say: “Real books will always be handmade.” Perhaps. But “real” photographs were once only film. Then digital appeared, and 99% of photos became digital. Film didn’t die — it became a niche for connoisseurs. The same thing will happen with human authorship.


The prediction

By 2029, at least one major book platform (Amazon, Kobo, Apple Books) or publishing group (Penguin Random House, Hachette, HarperCollins) will introduce a verified "Written by Human" certification — a label confirming the text was created without substantial use of generative AI.

In parallel, a new mass profession will emerge — the AI content curator: a person who selects, evaluates, and recommends texts (books, articles, courses) from the stream of AI-generated content. This won't be a "librarian" in the old sense — but closer to a sommelier navigating an infinite wine cellar.

◈ Verification date: January 1, 2029