My feedback is that a "front page photo" is a very high bar and most of the images fall well below what deserves the space. I would avoid:
- boilerplate page previews for e.g. github
- generic site logos e.g. arxiv, aide
- stock images and ai equivalents e.g. the models in suits stuff
- "decorative" images e.g. prime number
- author photos
- hate autoplaying gifs in this context. Very distracting. They might be great content better "play on demand" for me
Images work better when it's a relevant and a unique visual e.g. the gladiators for the history article or the cockroach but they are still are not really front-page material.
To truly earn their place, an image has to add information above and beyond the text. Identifying when an article is actually about an image e.g. space photo, data visualisation etc. would add some real value to the presentation.
I made this to experiment with embeddings and explore how different ways of displaying information affect your perception.
It gets the top 100 stories, sends their html to GPT-4 to extract the main content (this was not producing good enough results with html parsing) and then gets an embedding using the title and content.
Likes/dislikes are stored in local storage and compared against all stories using cosine similarity to find the most relevant stories.
It costs about $10/day to run. I was thinking of offering additional value for a small subscription. Maybe more pages of the newspaper, full story content/comments, a weekly digest or ePub export or something?
I think some of the highest value from HN comes from the comments, and it's much harder to find the "best" ones, since they might be in threads you might not have otherwise read.
Not sure if it's a "premium feature" so to speak, but would be very cool to extend this to comments generally.
Definitely, comments are usually better than the article. I thought of a 'Letters to the Editors' section that shows top comments (https://news.ycombinator.com/bestcomments) and references the parent story, but it might not be as useful without the context.
Maybe 'See Comments' here could load the comments on the same page? In a newspaper like style.
> Likes/dislikes are stored in local storage and compared against all stories using cosine similarity to find the most relevant stories.
You're referring to using the embeddings for cosine similarity?
I am doing something similar with stocks. Taking several decades worth of 10-Q statements for a majority of stocks and weighted ETF holdings and using an autoencoder to generate embeddings that I run cosine and euclidean algorithms on via Rust WASM.
Nice – I like this a lot. I feel like I'd use this for slow-lane reading and the original HN site when I'm in a rush.
Regarding HTML to GPT-4, I seem to remember commenters here saying they got better results by converting the HTML to Markdown first, then sending to an LLM. Might save a bit of money too.
This doesn't look like a print newspaper. Print newspapers are much denser (in general) and have different headline sizes to emphasize the editor's choice of stories. This looks like a corporate blog home page or something. Some people will like this presentation; I'm pretty happy with HN as it is. But congratulations on shipping!
I kept it running for 5 or 10 years but eventually let it die.
edit: I'm not hating on OP btw. their version has pics, which mine doesn't. just agreeing that I believe the visual hierarchy inherent to newspaper title design is an important benefit of the format.
> the visual hierarchy inherent to newspaper title design is an important benefit of the format
Agreed. This is also why old-school print design product catalogs often had superior presentation compared to today's web UIs for browsing hierarchically organized products. Everything is given the same visual weight and is formatted the same way.
Anyway, improving on what you did with the tooling that's easily available in 2024 but wasn't in 2009 seems like a fun challenge.
I’ve been running https://dailypopulous.com/ for years which is basically this but for general news. It generates a static paper edition every 6 hours from the most popular news links on social media with archives of all previous editions available.
Thanks for the feedback! Print newspaper's have curation, which this lacks. I guess the main thing it takes from newspapers is the image and blurb that help give you a preview of the story.
There is a form of curation on HN and "editorial judgment" on HN and that's in the points a post has. A closer approximation of a newspaper would be possible by looking at the points of a post and maybe comparing that to other posts and then sizing headlines appropriately based on how "important" the HN community sees a given story.
Currently the 131M Buildings story shows the blog author picture and BIO instead of a summary of the actual story. Is this easily fixable or is it a tough problem.
Papers only work because they know exactly what the view portal is and can design the layout relative to that. Unless you have an a3 sized screen this will not work very well online.
Good point. Of course Helen Keller wouldn’t have been able to use this if she were alive today… Or could she have? Have you ever considered that Helen Keller was faking it all along? I mean think about it how can you read and write if you are deaf and blind??
A few years ago, a similar project was posted on HN that I thought was really cool too - E Ink smart screen puts a newspaper on your wall (https://news.ycombinator.com/item?id=22831323).
As the son of two journalists (one was an editor) news-editing and layout is a lot more than just using serif and playing with typography a bit
The closest thing I’ve found to something actually resembling a proper masthead’s layout would actually be this from HN a few months ago:
https://cybernetic.dev/grid
Balancing high information density and readability is the key to a proper layout. You’ve erred to far on minimalist “readability”
I’d suggest looking at older Indesign/Quark Express magazines if you want to see elements of publishing layouts done digitally
There’s going to be a great layout one day that fuses the news-editing with web, but this isn’t it
I'd seen that when first loading the site shortly after the story was posted. Some 12 hours later or so, content is loading as expected. No network adjustments made on my end; I suspect the site operator changed their configuration(s).
This is pretty cool, it’s nice to have a clean interface that puts more focus on individual posts (as articles here) rather than tons of headlines where I feel I skim over posts a lot more (particularly the post about Jupiter only caught my attention on your site, not the front page).
I’d like if there was some support for customising it without liking and disliking so I could push topics I’m interested in first (e.g. those tagged with emacs). It would also be nice to hide the like and dislike buttons in general as it gives more of a social media feel that the newspaper style UI does well to shake.
Interesting, for me it's a bit the opposite. In the standard view I really read every headline and consider what might be through the click. In this version I skim more in mindless scrolling fashion.
Readit News
My feedback is that a "front page photo" is a very high bar and most of the images fall well below what deserves the space. I would avoid:
Images work better when it's a relevant and a unique visual e.g. the gladiators for the history article or the cockroach but they are still are not really front-page material.To truly earn their place, an image has to add information above and beyond the text. Identifying when an article is actually about an image e.g. space photo, data visualisation etc. would add some real value to the presentation.
It gets the top 100 stories, sends their html to GPT-4 to extract the main content (this was not producing good enough results with html parsing) and then gets an embedding using the title and content.
Likes/dislikes are stored in local storage and compared against all stories using cosine similarity to find the most relevant stories.
It costs about $10/day to run. I was thinking of offering additional value for a small subscription. Maybe more pages of the newspaper, full story content/comments, a weekly digest or ePub export or something?
Not sure if it's a "premium feature" so to speak, but would be very cool to extend this to comments generally.
Maybe 'See Comments' here could load the comments on the same page? In a newspaper like style.
It should not cost more than a dollar a day.
Take AWS and azure credits and run it for free for years
Dead Comment
You're referring to using the embeddings for cosine similarity?
I am doing something similar with stocks. Taking several decades worth of 10-Q statements for a majority of stocks and weighted ETF holdings and using an autoencoder to generate embeddings that I run cosine and euclidean algorithms on via Rust WASM.
How well does it work?
Regarding HTML to GPT-4, I seem to remember commenters here saying they got better results by converting the HTML to Markdown first, then sending to an LLM. Might save a bit of money too.
https://github.com/gilesbowkett/hacker_newspaper/blob/master...
I kept it running for 5 or 10 years but eventually let it die.
edit: I'm not hating on OP btw. their version has pics, which mine doesn't. just agreeing that I believe the visual hierarchy inherent to newspaper title design is an important benefit of the format.
Agreed. This is also why old-school print design product catalogs often had superior presentation compared to today's web UIs for browsing hierarchically organized products. Everything is given the same visual weight and is formatted the same way.
Anyway, improving on what you did with the tooling that's easily available in 2024 but wasn't in 2009 seems like a fun challenge.
I've wanted to take a stab at it because I think it would be "neat" but haven't actually found any good reference implementations.
also seems like with almost everyone on mobile it's just not worth it.
A few years ago, a similar project was posted on HN that I thought was really cool too - E Ink smart screen puts a newspaper on your wall (https://news.ycombinator.com/item?id=22831323).
The closest thing I’ve found to something actually resembling a proper masthead’s layout would actually be this from HN a few months ago: https://cybernetic.dev/grid
Balancing high information density and readability is the key to a proper layout. You’ve erred to far on minimalist “readability”
I’d suggest looking at older Indesign/Quark Express magazines if you want to see elements of publishing layouts done digitally
There’s going to be a great layout one day that fuses the news-editing with web, but this isn’t it
I’d like if there was some support for customising it without liking and disliking so I could push topics I’m interested in first (e.g. those tagged with emacs). It would also be nice to hide the like and dislike buttons in general as it gives more of a social media feel that the newspaper style UI does well to shake.