AI Assistant Patterns

The viewer adds a file to the question and needs to see what the model will actually read.

Power users need to name a tool, a file or a person without describing it in a sentence.

One box has to take a word, a page, a file and a keyboard shortcut without growing into the whole screen.

The most common screen in the product is an empty box, and it has to teach without running a tour.

The same question wants a fast answer some days and a researched one others.

Several models sit behind one box, and the difference between them stays invisible until the answer arrives.

An empty box tells a first-time viewer nothing about what the thing can do.

The question was badly worded, and asking again pushes the bad exchange down the transcript instead of replacing it.

Two voices alternate down one column, and the reader has to know at a glance which one is speaking.

The answer is wrong or flat, and the only obvious recourse is retyping the question.

Copying, rating, sharing and rerunning all attach to one message, and none of them can dominate it.

The answer went wrong in its first sentence and the viewer has to sit through the other four hundred words.

A model takes ten seconds to finish a thought, and ten seconds of spinner reads as broken.

The answer is a document or a program, and a chat transcript is the wrong container for either.

Generated code gets read in one place and used in another.

One answer runs to eight screens and buries every other turn in the conversation.

The answer is a comparison, and prose is the worst available shape for one.

A well-sourced paragraph and a fabricated one read identically until the viewer can see where each claim came from.

The answer is stated in the present tense about a world the model last saw a year ago.

The model worked for thirty seconds and the viewer wants to know whether it understood the question.

The answer drew on eight pages, and the viewer needs to judge those pages before trusting it.

The assistant did something in the world, and the viewer only sees the sentence describing it afterwards.

A long conversation quietly stops remembering its own beginning.

The useful exchange was three days ago and the viewer remembers one word from it.

The same correction gets retyped at the start of every conversation.

The assistant knows something about the viewer that the viewer never agreed to in this conversation.

The assistant sits inside a product full of data and the viewer cannot tell how much of it is in scope.

The feature has to be findable inside a product people already know, without retraining them on it.

The conversation has to sit next to the work without covering the work up.

A suggestion has to be readable in place and obviously not yet real.

The viewer wants one paragraph changed and does not want to leave the document to ask for it.

The model rewrote a paragraph and the viewer has to see exactly what changed before agreeing to it.

The stream died halfway and left a half-written paragraph on the screen.

The assistant will not do the thing, and the viewer needs to know whether to rephrase or give up.

The viewer is spending a finite resource they cannot see until it runs out.