Inline AI is assistance that operates inside the text you're writing, instead of in a separate chat window, sidebar, or app. You never leave the input: requests are expressed in the text itself, and responses land in place, as part of your draft.
How it differs from chat
Chat-based AI is turn-based and elsewhere: describe your context, submit, wait, copy the answer back. Inline AI inverts every part of that: the surrounding text is the context, the request is a keystroke (OpenCues uses _), and there's no copy-back because the answer is already where it belongs.
The design constraints that make it work
Inline assistance must never fight the writer: no popups stealing focus, no text reflowing under the cursor, suggestions that are visible but ignorable, and latency fast enough to keep up with typing. These constraints, developed in the Inline Cues essay, are why inline AI took a decade longer than chat to arrive.
Inline AI in practice
OpenCues is an open standard for exactly this surface: blanks for requests and cues for suggestions, in any text field. Related: What is a HAII? · GitHub repo