Experience how prediction can feel like mind-reading.
Prediction Lab runs short interactive experiments — a number you pick, a sentence you begin — and then reveals both the guess and the logic that produced it. No paranormal claims. Just attention, priors, framing, and uncertainty, made visible.
Two short challenges — roughly a minute each.
Each experiment ends with a reveal: what the system predicted, how confident it was, and the cues that drove the guess.
Think of a number. We'll narrow in.
Framing, priors, and a few staged questions can narrow a "random" choice to a surprisingly tight range. See where the odds sit before you pick.
002 / WritingBegin a sentence. We'll finish it.
Comprehension and production both rely on prediction. Pick a seed, and watch a rule-based model rank the most likely continuations.
See the same image two ways.
Expectation shapes perception. A short ambiguity demo, with priors we manipulate in front of you.
Imagine a scene. We'll narrow the space.
Constrained prompts and selection patterns collapse a huge image space into a small shortlist — transparently.
Uncanny, but honest.
This is a science-entertainment and AI-literacy project — not a claim of paranormal ability.
- Transparency with restraint Every guess comes with a confidence band and a brief explanation of the cues that drove it.
- No sensitive inference We will not infer health, political, or protected attributes from how you interact.
- Auditable models Constrained, interpretable logic — Bayesian narrowing and rule-based sequence prediction — not opaque black boxes.
- Consent-first data Interactions live in your browser by default. Research participation is strictly opt-in.