Project overview
An AI-driven browser mystery. As an archive investigator, use open conversations and evidence to unravel a disappearance. LLM-generated narrative supports different paths through the investigation.
- Zhihu Hackathon participation
- LLM narrative / Open conversations
- Browser-based / No installation
- Branching endings / Evidence-led investigation
Project details & context
- Platform
- Web browser
- Genre
- AI mystery / Interactive deduction
From a question to an experience
Give open-ended conversational deduction a working set of rules and multiple outcomes.
As an archive investigator, the player examines a disappearance through evidence, character conversations, and conflicting information.
Design rules for deduction
Four mechanisms structure information manipulation: contamination comparison, echo evidence chains, AI-judge debates, and faction persuasion.
Deliver a playable system
The implementation includes eight characters, ten ending tiers, solo and 2–5-player modes, and a script-production workspace.
My contribution
Solo development · Gameplay / Systems / AI architecture
Sep 2026
- Designed contamination comparison, echo evidence chains, AI-judge debates, and faction persuasion.
- Implemented ten ending tiers, eight characters, solo and 2–5-player modes, and a script-production workspace.
Delivery & outcome
Completed 48-hour development; received a Zhihu Hackathon Campus Rising Stars participation certificate.
Certificates & records
Continue exploring
Video showcase · Original in Chinese

