An interactive applied-AI demo. Pick a category, type a clue, and the model returns a Jeopardy-format question. The point isn’t to win Jeopardy — it’s to show, on a single page, what conversational generation looks like when you constrain the output format.

What it does

You give it a category (history, science, “1980s sitcoms”) and a clue. It returns a question phrased the way a Jeopardy contestant would: “What is…”, “Who is…”. The format constraint is the demo. Everything else — model choice, prompt scaffolding, latency tuning — is plumbing.

Why it exists

Most “applied AI” demos try to do too much and end up demonstrating nothing in particular. AI Jeopardy is the opposite: one input, one constrained output, one page. It’s a working proof that conversational generation can be steered into a narrow shape — which is the part that actually matters for production use.

What it isn’t

It isn’t a product. It isn’t a game you’d play for fun. It’s a small, demonstrable artifact that answers the question “can you actually build with this stuff?” — yes, here, in one page.