pull down to refresh

There's a few reasons why a structured data approach wouldn't work that well. First, it would require the review writers to include the structured data in an easily identifiable and consistently-formatted way. Second, it would require that the categories used by the data providers are the categories we want. However, I'd reckon that here on SN we'd want our own categories, like "Bitcoin-Related Non-Fiction".
I think you overestimate how much intelligence is required for extracting structured information from text blobs, provided that the source is of reasonable "quality" -- e.g. major Wikipedia articles, mainstream media obituaries, recipes from professionally published cookbooks...
I'm not saying that you wouldn't need several layers of processing; my point is more, "your use of AI should be limited to vibe coding, and the final pipeline should not include inference"
reply