Skip to main content
  1. AI × Life/

Where StoneStory Came From: Can It Predict the Missing Ending?

·1 min
QQder · The Miniature Boat
Author
QQder · The Miniature Boat
Eight iOS apps — all free, no ads, no tracking. Pick one and try it. Also a running log of how a humanities-background sysadmin builds apps from scratch with AI vibe coding. I also take fixed-scope agent-automation commissions — see Services.
App 來歷 - This article is part of a series.
Part 8: This Article
This is the origin of StoneStory (motivation, audience, difficulties, current state), not a how-to. To download and try it → see the App. The thinking behind it is in the fate and thermodynamics posts.

Released 2026-03-16 · Store name “紅樓夢模擬器” · v1.2.2 · 106 MB · free · offline after setup

StoneStory turns Dream of the Red Chamber into an interactive AI narrative simulation, so literature is no longer only static reading.

Why it was made
#

An AI (LLM) specialises in predicting the next word, which leads naturally to the question of whether it could predict the missing ending of Dream of the Red Chamber.

There was no single trigger. I have always wanted to know the ending, and always wanted to predict the future.

Who it is for
#

Readers who love the novel. Researchers and users interested in what AI can and cannot do.

How it was done, and what it revealed
#

Technically it used AIGC to generate the series art library, and extracted the factors that shape the plot and the ending into a programmable flow.

Along the way I found that having AI play every character has no usability and no playability at all, because the reading value still comes from the author’s arrangement and prose, and dialogue the AI improvises simply does not have that. So the focus moved to reading comprehension of the text, guiding the user along the original rather than letting the AI act freely.

Where it stands
#

The ideal would be many agents playing the characters and interacting in some structured way, with parameters you can tune in real time to affect the world inside. Only both development and inference burn an enormous number of tokens, and with memory prices this high, it is unlikely to be built for at least the next few years.

App 來歷 - This article is part of a series.
Part 8: This Article