How to Save a Hacker News Thread
Hacker News threads are dense in a specific way: the correction that matters is usually four levels down, under a comment that is collapsed, in a subtree nobody scrolled to. The value is in the whole tree, and the whole tree is exactly what you cannot see.
If you want to ask a model what the thread actually concluded, you have to hand it the thread, not the top of it.
The page is not the discussion
Open a 600-comment thread and the page gives you the ones it decided to render, with [more] links standing in for the rest and collapse toggles hiding subtrees. Select all, copy, paste, and you have captured the state of a user interface rather than the content of a discussion.
Two things go missing at once:
- The replies nobody expanded. They are not in the document you copied because they were never in the page.
- The nesting. Indentation on Hacker News is drawn with spacer images and table cells. It does not survive a copy, so a rebuttal arrives looking like a fresh opinion.
Take the whole tree in one request
FileConcat Clipper is a browser side panel that reads the thread properly. On Hacker News
it goes to hn.algolia.com/api/v1/items/<id>, which returns the entire comment tree in
one request: no authentication, no paging, and nothing to expand first.
That means depth costs nothing. Measured on 2026-08-19 against a 638-comment thread, one clipping came back with 264,851 characters nested eight levels deep, none of which was on screen when the button was pressed.
- 1
Open the panel on the thread
Click the toolbar icon. On a Hacker News thread the panel offers Clip this thread. Nothing needs expanding first, and there is no setting to get right.
- 2
Check it if you want to
Peek reads the Markdown back inside the panel before you send it, which is where a wrong page or a read that stopped early shows up while it is still cheap to fix.
- 3
Send it to the bundler
Send pushes the clipping into an open fileconcat.com tab, opening one if there is none, where it becomes one document with a live token count.
Where to get it
The extension is on the Chrome Web Store, free and open source. There is a page about what it does at /clipper and a reference at /docs/clipper.
Mind the size
A quarter of a million characters is roughly 65,000 tokens, which is a real portion of a context window spent on one thread. That is the correct trade when the thread is the subject, and the wrong one when it is background.
The bundler is where you find that out rather than guessing: it counts the tokens for every major model as the bundle grows, so you know what the discussion costs before you paste it. If it is too much, take the linked article instead and keep the thread for a follow-up question.
A snapshot, not a feed
A clipping is the thread as it stood when you took it. Hacker News threads keep growing for a day or two after they peak, so a thread clipped early is a different document from the same thread clipped tomorrow. Re-clipping and sending again replaces the old file in place rather than adding a second copy.
The front page, and everything around it
The front page is a listing, so the panel puts a tap target on each of its stories. Tap the ones you care about and each thread is read on its own, which is how a morning of open tabs becomes a set of files.
The more useful pattern is mixing sources. A Hacker News thread is usually about something: a blog post, a repository, a specification. The clipper reads articles anywhere with no per-site code, so the post and the argument about it can go into the same bundle, and the repository can go in under it from the bundler itself. A send extends the bundle rather than replacing it, so the order you collect things in does not matter.
That is the question worth asking a model: not what this thread says, but where the thread and the thing it is discussing disagree.
Nothing is uploaded to us at any point. The clipping is made in your browser, held in your browser's own storage between the clip and the send, and handed to the tab you already have open.