Jan: i also wonder, if done well, if big brands would pay to analyse the public reactions in a more structured way than just reading them themselves. anyone knows if any products are already offering this for x? LLM comment analysis.
Defender — 18:09 yeah i think there are many commercial tools that do this that are all private/corporate/kind of shady, but ~zero "open access" ones. The wikipedia/open street map of user sentiment does not exist
(and I think if/when it does exist it will destroy all competition, same way open street map ate google maps' lunch, despite google pouring billions into acquiring the map data and refusing to share it)
Graphika is one big one in this industry: https://www.graphika.com/ this is a random newsletter i check from time to time, its audience is people who want to understand social media trends/make their corporate products or w/e go viral: https://www.socialmediatoday.com/
-
All posts you're the author of with all quoted and commented posts. This is already the case.
-
When you want the trace of your comment on someone elses thread, arguably all posts/comments going up and all comments going down should be accessible to you and anyone else, even if most of it isn't your post. However, this should only be open when the starting point is YOUR COMMENTED POST. This means you cannot get all comments on a large post easily because they don't comment on you, just on the main thread.
-
All your followers and following with their status should be known. Arguably you should be allowed to access a summary from them as well, even if they didn't allow this. This should be accessible FOR YOU at https://x.com/[username] but not everyone, just for the people that have a first degree connection. This prevents XYMake being used for massive scraping.
-
Users you recently interacted with are much warmer connections that you are likely also interested in to interact more with. Therefore, these should be included in a "inner circle" context. Even if they did not authorize anything, we can normalize volume to per-user based on whenever they interacted with you in the past. This shows overlap in interest and creates a better summary. We need
/username/with_username/username2
- Link to https://chatgpt.com/?q={searchTerms}&hints=search with useful prompts.