About Prism

Prism is built on one finding that shows up again and again in learning research: watching an algorithm animate teaches almost nothing, while being asked to predict what happens next — and being told exactly why a wrong prediction was wrong — is the cheapest intervention with the largest measured gain. So every visualization here is a real trace of real code, and playback pauses at genuine decision points to ask before it reveals.

Everything on the canvas is drawn like an engineer's working drawing: graphite construction lines for what is untouched, ink for what is settled, one red annotation for what the algorithm is doing right now, and a highlighter wash under the answer. State is encoded in form first — outline, dash, fill, strike — so a greyscale print, or colour-blind eyes, read the same drawing.

What it is

• Forty-two algorithms as scrubbable, questionable traces — not videos.
• Prediction prompts on by default, with "just watch" one click away.
• "Be the algorithm": perform each operation yourself; the trace referees.
• Essays that answer the interview follow-ups, not just the happy path.
• Local-first: progress lives in your browser. There are no accounts.

Who makes it

Prism is an independent project. The algorithms are implemented from the standard literature and each one is checked against an independently coded reference implementation in CI; when the trace and the reference disagree, the build fails.

Found a mistake? Say so — a correction is worth more than a compliment.