Nielsen Norman Group published a working AI glossary built specifically for product and design practitioners. It covers the terms that surface in actual UX work: tokens, context windows, agents, evals, and prompt injection. Definitions are calibrated to how these terms function in products, not how researchers or vendors define them, which are often different.
The value is not just in the conclusions. Each entry links out to deeper NN/G research, and the glossary is organized for both quick lookup and sequential reading. That dual structure makes it useful as a reference you return to, not just a one-time read.
If you are working on AI products and still fuzzy on why context window size affects output quality, or what evals actually measure, this is the place to start before those gaps cost you in a design review.
[READ ORIGINAL →]