61–70 of 134 posts
I get most of my daily news from ... Teletext [1]. I'm lucky to be in a country that still has it. I much prefer that format over news sites on the web: Headlines are often laid out in a list one entry below the other, which is easy to read through without having to scroll. Because pages are limited to 40 columns × 24 rows, every article is short and to the point. I don't usually read it on TV though but on a web-sit…
So I build NewsWaffle, which for any website:
https://github.com/acidus99/NewsWaffle
* Automatically builds a list of news stores, separate from the navigational hyperlinks.
* Detects RSS/Atom feeds to provide a more accurate list of news stories.
* Uses Readability to show only article content on article pages.
* Uses meta data like OpenGraph or Twitter cards to provide richer formatting, and to determine page type.
It regularly converts 900 KB home pages or 1.2 MB news articles into into 3KB for links to news stories and 5K of text
It does this by:
* Using semantic tags like , , and to determines which hyperlinks are navigational and which ones are likely links to news articles.
* OpenGraph meta data to determine page type news stories and extra metadata.
* A Aggressive HTML parser that strips out a ton of tags, CSS, JS, etc
* Readability library to extract out the text of news articles
I built this as a service in Gemini, so if you have a gemini browser you can try it. Otherwise, here is a HTTP-to-gemini proxy showing you what a NYT article looks like:
Gemini link: gemini://gemi.dev/cgi-bin/waffle.cgi/
NYT Homepage: https://portal.mozz.us/gemini/gemi.dev/cgi-bin/waffle.cgi/li...
NYT Article: https://portal.mozz.us/gemini/gemi.dev/cgi-bin/waffle.cgi/ar...
We should make an 'awesome-text-only' github repo with all these.
Earlier quoted context omitted.
Comment The summaries for several headlines are completely bogus, which undermines any credibility that this was a reliable site. For example, the Lufthansa Strike blames a technical issue, not the strike that all linked articles discuss. Next, the Khazahstan election summary mentions voters in Turkey and Turkmenistan. There’s also duplicated stories, and other obvious issues any human editor would catch. I suspect t…
Hi, thank you for the feedback. I should have added that we launched it yesterday so it there is still some quirks. Yes, we use language models to do the summaries, they are not perfect yet and the feed ordering also needs to be improved. Since you get the actual news headlines and teaser on the detail, we assume some inaccuracies are ok for now. The primary focus right now is clustering and a high-level overview ove…
Earlier quoted context omitted.
Nice. Did you build this? If so, what's the tech stack you used? I like the interface - compact and fast.
Thanks! Yeah, I’m one of the founders. It’s Nextjs on Supabase. We launched the news clustering just yesterday so I thought it would be nice to get some stealth HN feedback for it. We are planning to launch an API (rss style) for the news next week where you can get temporal clusters and everything. Imagine having full access to all of Twitter trending but across all news sources. We are trying to filter clickbait, S…
It would be nice if your interface sorted the sources by bias and let you side between different summaries. In the upper right corner of Modo you can swap between timeline view (seeing past incarnations of a story as it progresses, or lean view that let’s you see various bias.)