About the project
CaseClosed puts the player inside a fictional gallery-theft investigation. A bronze sculpture has disappeared, three witnesses offer conflicting accounts, and a suspicious photograph points toward an apparently obvious suspect.
The player must inspect original records, question witnesses, compare a cropped photograph with its full frame, and reconstruct what a surveillance camera could actually see.
The central puzzle involves two matching display cases. An image shows an empty case, but it becomes useful evidence only after the player identifies which display the camera could see around a partition.
To close the case, the player must select a suspect, establish the time window and attach supporting evidence. Incorrect or incomplete theories receive feedback rather than an automatic success.
Who it is for and why it matters
CaseClosed is for people who enjoy detective stories and escape-room puzzles and want to participate in solving the mystery.
It combines entertainment with opportunities to examine context, challenge assumptions and distinguish suspicion from proof. Its educational potential has not yet been validated through user testing.
What makes it distinctive
A witness can lie without being the thief. A genuine photograph can still mislead through cropping. Players must test those accounts against the room's geometry and independent records before reaching a conclusion.
How it was built and where AI fits
The application was built with AI coding assistance. Browser JavaScript handles the investigation, spatial calculations, notebook progress and verdict rules.
The demonstrated version uses authored witness responses. Optional model-assisted question interpretation is implemented but inactive; no live AI generation is claimed.
Development challenge
Audible autoplay varies by browser. CaseClosed attempts playback on arrival and displays an Enable sound control when permission is required. The opening remains skippable, with working mute and volume controls.
Development-process disclosure
The initial prototype was built before installing the official Devpost Learn Skill Pack. The pack was then installed and used for the subsequent planning and review work. The scope and product direction were approved; the technical specification and verification checklist are in the repository as drafts. Uncompleted formal reviews are not represented as completed. The optional live AI feature remains inactive. All events and evidence in the case are fictional. Following the initial prototype, the automatic-opening and soundtrack revision was implemented and tested after the Skill Pack was installed. The demo shows the working investigation before that final opening revision.
Log in or sign up for Devpost to join the conversation.