that's what archive.org already does, but if you want to re-implement it, you would have to crawl all the web, eventually save thumbnails of pages with screenshotone (https://microlaunch.net/p/screenshotone)