ABOUT THIS GAME
Suspect is a free-text interrogation detective game. You question an AI-roleplayed guilty suspect — a small-town diner owner — and must catch her revealing impossible knowledge, contradictions, and slips under pressure. The engine owns all truth and adjudication; the LLM only performs the suspect. Present evidence, track who-knows-what, flag contradictions, and build a convincing accusation. Fully playable in English or Danish — pick your language on the title screen.
KEY FEATURES
Free-Text Interrogation
Ask the suspect anything in your own words. No dialogue trees — the AI roleplays the suspect and answers naturally.
Ask the suspect anything in your own words. No dialogue trees — the AI roleplays the suspect and answers naturally.
Impossible-Knowledge Detection
The engine tracks exactly what you've revealed. When the suspect references something you never told her, it's flagged deterministically.
The engine tracks exactly what you've revealed. When the suspect references something you never told her, it's flagged deterministically.
Engine-Owned Truth
All case facts, evidence, and adjudication live in the engine. The LLM never decides truth, contradictions, or scoring.
All case facts, evidence, and adjudication live in the engine. The LLM never decides truth, contradictions, or scoring.
Living Workstation
A draggable, resizable canvas workstation with an animated suspect, case file, evidence, timeline, witnesses, and contradictions windows.
A draggable, resizable canvas workstation with an animated suspect, case file, evidence, timeline, witnesses, and contradictions windows.
Evidence Presentation
Present evidence to raise pressure and update the information ledger. Watch the suspect's demeanor shift.
Present evidence to raise pressure and update the information ledger. Watch the suspect's demeanor shift.
Deterministic Accusation
Build a who/what/when/how/why accusation and get a rubric-scored verdict: CONVINCING, WEAK, WRONG, or MISSED OPPORTUNITY.
Build a who/what/when/how/why accusation and get a rubric-scored verdict: CONVINCING, WEAK, WRONG, or MISSED OPPORTUNITY.
Server-Side Saves
Your full epistemic state saves per player token. Resume across machines.
Your full epistemic state saves per player token. Resume across machines.
English & Danish
Full EN/DA localization: UI strings, case prose, offline fallback dialogue, and AI suspect replies all follow the language chosen on the title screen.
Full EN/DA localization: UI strings, case prose, offline fallback dialogue, and AI suspect replies all follow the language chosen on the title screen.
GAMEPLAY
Perspective
First-person detective workstation
Focus
Free-text interrogation, evidence presentation, and contradiction detection
Players
1
Native Resolution
Responsive logical resolution (no fixed pixel-art)