It would be interesting to save every single thing you ever see online to a local cache and index it using machine learning so you can search by topic. This is the kind of storage I imagine you could use for something like that.
Also I’m curious how large the collection is in total file size.
grep is nice, but Recoll gives you stemming and such.
Or ever seen. Screen record desktop at 1080p on codec optimized for text, maybe even grayscale with small samples of color data. Then have system index content (text, label graphics). When viewing, system will upscale, colorize, interpolate frames. Can probably throw in neural audio codec to store speech/audio.