1 Curious about Twitter thread discovery with API. I have this https://github.com/houshuang/hypothesis-to-bullet () which tries to convert Twitter threads to bullet-lists for Roam or other tools. Right now you give it the last tweet, and it follows the parent references to the top..
2 Sometimes it works really great (like this example https://twitter.com/houshuang/status/1213422234552995840), but many threads that look nice on https://twitter.com are actually not sequential, but threaded (more like this https://treeverse.app/view/JR7YNint ). So how would you capture this using the API...
3 Perhaps using Twitter search, for each parent, do a search to see which children it has? Of course, how to visualize it is a different question, but even getting the same linear view that https://twitter.com has, but locally, permanent and searchable
4 (Another issue someone raised is that currently I just link to the Twitter images, he wanted to store all images locally to really be sure the information survives for 100 years). Anyway this is the thread I wanted to capture that prompted all this: https://twitter.com/RichDecibels/status/1230555517916860416
5 With @treeverse we get an awesome graph of this thread. Diving into the source code, it uses what seems like an undocumented API `2/timeline/conversation/`, https://github.com/paulgb/Treeverse/blob/master/src/common/util.ts. Maybe I'll try forking treeverse to enable bullet-list output. Great work @paulgb!
Well that was quick. See the "Copy to bullet list" button? Works like a charm. Am still going to experiment a bit with output format https://github.com/houshuang/Treeverse @paulgb https://twitter.com/RichDecibels/status/1230555517916860416 @RoamResearch
So I've been using my fork of Treeverse ( https://github.com/houshuang/Treeverse) to scrape Twitter for a while, and there were some things I was unhappy with, so I spent a few hours, and now I'm _really_ happy! I recorded a little video https://www.loom.com/share/ea29fee446b740618cf4348d3b12d517
I made the algorithm to turn trees into indented hierarchy better, switched to using linked display_names, removed a lot of cruft, displaying images, expanding URLs, etc. This is now really robust, and readable, a great way to capture the richness of Twitter! @Conaw @RoamResearch
Some examples here: https://roamresearch.com/#/app/stian-research/page/04-13-2020, featuring @vgr @visakanv @TheOrangeAlt and @Conaw
Came across an awesome thread by @xuenay, love the shape of it. Exactly the kind of stuff I really want to capture to do deep processing and integration later. https://twitter.com/xuenay/status/1248643701347319811