How to read a long AI answer without scrolling forever

Short answer: ask for the structure you want — headings, the conclusion first, code kept separate — and move by heading rather than by scrolling. An outline turns a two-thousand-word answer into a list you can jump through.

You asked one question and got two thousand words with six code blocks. The part you need is in there somewhere.

Why long answers are hard to read

They arrive as one undifferentiated column. A document of that length would have headings you can scan, a table of contents, and a scrollbar that means something. A chat answer has none of those: the scrollbar is measuring the whole conversation, not the answer, so it tells you nothing about where you are inside it.

Worse, the answer is still growing while you read. Any position you find moves.

Techniques that work now

1. Ask for the shape up front. "Answer in at most five bullets, then expand only the one I pick." Cheaper than reading two thousand words and discarding most of them.
2. Let it finish before you read. Reading a streaming answer means re-finding your place every few seconds. The thirty seconds you save by starting early, you lose twice over.
3. Use headings as anchors. If a long answer has headings, Ctrl+F on a heading word jumps you there instantly. If it has none, ask for them next time — models comply readily and it costs nothing.
4. Ask for the conclusion separately. "What's the one-line version?" as a follow-up gives you a short message you can find later, instead of a conclusion buried at paragraph forty.
5. Collapse code you are not reading. Where a platform allows it, this alone can halve the visual length of a technical answer.

The navigation problem underneath

All of the above is working around the same gap: a chat has no map. You cannot see the shape of a conversation or an answer, only the small window currently on screen, and the scrollbar is measuring the wrong thing.

This gets worse as the conversation grows, which is exactly when you most need to find something in it.

How Tvara handles it

Tvara draws a minimap on the right edge of the conversation: one mark per turn, yours and the model's distinguished, the whole chat at a glance. Hover a mark to preview that message, click to land on it.

An outline of the conversation sits beside it, and in-chat search reaches messages the page has already unloaded — so finding the part of a long answer you need stops being a scroll.

The minimap is free on every supported platform. The outline and in-chat search are free on ChatGPT, Perplexity, DeepSeek and Grok, and part of Pro on Claude and Gemini. See what each one does · The rest of the toolkit