3 ms·
RSS sucks because it depends on every content source offering the API. The successor to RSS is where each feed is an arbitrary URL that your client hits to gen
by hombre_fatal 1y ago
RSS sucks because it depends on every content source offering the API.
The successor to RSS is where each feed is an arbitrary URL that your client hits to generate a feed. With an LLM this is trivial compared to before.
- xnx 1y agoAgree! Just wrote basically this comment before I saw yours. Disappointingly even pay tools like Feedly have extremely poor RSS creators. All Feedly's features for the past few years seem to focus on some obscure niche of IT security threat awareness.
- imiric 1y agoRSS is not an API, though. It's another content type like HTML, typically XML, but in a specific format that makes it easier for programs to consume. Generating it is trivial in most web frameworks and CMSs, and sites that don't offer it either never bothered to set it up, or actively choose not to. This is hardly a fault of RSS itself. What you suggest as a successor is a workaround for such sites, not something that should be the norm.
- hombre_fatal 1y agoSo RSS isn't an API because it returns XML instead of JSON or protobufs? That doesn't sound right. It's an endpoint for machine consumption and that's what most people mean when they refer to an API. But that's a distraction from the point. It's trivial to implement an RSS endpoint, but what's not trivial is requiring everyone to implement it. Needing someone to implement an API just to have machine-friendly pull access to their latest content is only a trade-off that made sense last decade. It's not a good solution if you had the power of generating feeds yourself.
- HankStallone 1y agoAre you talking about something that can go out to any web site, grab the HTML, and turn it into a feed that's consistent with any other feed and not a big hairy mess of stuff that wasn't actually part of the content? That doesn't sound trivial at all to me, so maybe I'm misunderstanding.
- hombre_fatal 1y agoIt's a trivial task for an LLM to take the HTML at https://techcrunch.com/latest/ https://techcrunch.com/latest/ and extract a perfect list of only the article URLs. And if you can do this, then you can build the tool that I describe. From there you can extract each article content with an LLM or use something like readability.js or just download the whole page for later consumption. I have a prototype of it. You add a feed as { feedUrl, feedPrompt, articlePrompt }. `feedPrompt` lets you append rules for extracting the article url list from the `feedUrl` like "ignore video articles". `articlePrompt` lets you append rules for how to extract article content for a given website, like "translate to english".