The same decision can cost someone else.

Who becomes the guarantor depends on your opening conversation. What you prepare controls the options at the gate. Your final choice determines whether your promise holds. If you break the silence agreement, the named guarantor bears the extra detention. The story never rolls a hidden loyalty score.

The content is connected.

Sanity holds separate characters, facts, scenes, choices and ending fragments. References connect them. Choices disclose facts and set explicit state values; ending fragments match those values. The runtime accepts a small list of state keys and values, never executable code from content.

Editors can revise these relationships in Studio. A story-review tool checks the published graph and traverses every reachable path. The game refuses incomplete graphs and clearly labels its bundled sample when live content is unavailable.

What is verified right now?

These checks establish consistent execution, not whether the game is emotionally effective or likely to win. Player feedback still matters.

What changed during the build?

The first prototype was a warehouse investigation. Ben challenged its clarity, then challenged a conventional detective-game suggestion. We explored trust and betrayal, rejected a contrived lift-capacity dilemma, and reduced the scope to a short story with replayable consequences.

A local AIE review flagged missing causal rules. Its feedback was advisory; it did not prove originality or improved winning odds. The private engines were kept unchanged and are not part of this website.

Privacy and access

No player account, payment or runtime AI service is required. Choice IDs are saved in this browser’s local storage so a refresh can resume the story. They are not sent to Sanity. Clearing site data removes them. Story content is public and contains no private conversations or credentials.

The authoring interface requires a Sanity account with access to the project. Open authoring Studio →