Aight guys, listen up. ✨I M P O R T A N T✨. If you host twitter spaces regularly or you frequent spaces which are a genuine value add to the community hmu. I'm gonna start posting the summaries of the spaces within the Web 3 fold for those who are too busy to keep up with them.

having a hard time keeping up with all of my ai conversations on different platforms . anyone else have the same problem ? .the perks of being a power user ig 🤷‍♂️.

RT @ThreatT0Society: implemented the MASt3R SLAM paper for high fidelity 3D reconstruction from monocular camera only. making some changes….

RT @ThreatT0Society: reverse engineering X's internal api : a 🧵. we aim to extract insights from Twitter programmatically without using its….

what are the best deep research options available in the market right now ?. have mostly used grok and perplexity for this, would love to know how o1 and o3 compares in all of this.

The resulting data is in this format : ✨. PS : Thanks for reading it through, lmn if you'd want a breakdown for Instagram's internal api next

Step 4 : Profit ?? 💸. This enables us to extract usable data without having to use their api. mind you that x restricts your ability to view all of someone else's followers, but there's still a lots of data you can fetch with this. also, can perform actions such as posting.

we add a "cursor" to figure out where we were in our list. variables={"userid":"{numeric id}", "count":"20", "cursor":"{num1} | {num2}", "includePromptedContent":"false"}. the {num1} and {num2} can be extracted from the "value" of previous response

Step 3 : chaining requests for data extraction. with this, we've successfully made our first query, now in order to extract data for more than 20 people, we'll chain together requests to fetch our entire list. for this, we make a small change in the parameters.

although not necessary, but these are recommened :.x-client-transaction-id.x-csrf-token.x-xp-forwarded-forwarded-for. this query's return headers contain x-connection-hash, which is then used to calculate these for next calls. (will share this computation in some time 😉).

Step 2 : Identify the request headers for the GET request. It requires a Bearer token for authorisation. d_prefs, guest_id, kat, auth_token, ct0, fwid, twitter_sess, guest_id_marketing, guest_id_ads, personalizationid and cf_bm as cookies. these can be found using editThisCookie

{acc spec code} is likely a string which is associated with your account, don't know for sure why it is there. {numeric id} (for the account you want to extract data of) the id for each account can be found from the response to this query.

our url being : https ://x.com/i/api/graphql/{acc spec code}/following?. parameters are : variables={"userid":"{numeric id}", "count":"20", "includePromptedContent":"false"}. and features are supposed to be other options which aren't relevant for this discussion.

Step 1 : Figuring out the endpoint and required structure. this shows us that our required endpoint looks like this . https ://x.com/i/api/graphql/{acc spec code}/following?variables={"userid":"{numeric id}", "count":"20", "includePromptedContent":"false"}&features={". "}

there are different ways we could've approached this problem, but we chose this method cause its more robust and scalable. also, we've used web automation loads of time for data extraction from twitter and don't need its overhead

reverse engineering X's internal api : a 🧵. we aim to extract insights from Twitter programmatically without using its developer api. Here, we'll fetch following data for our account.

Would anyone want a detailed breakdown of X’s internal api !?.

How would you approach the limited compute problem for drones when we implement multiple processes like SLAM, neural network based map localization and detection + classification ?. We can stagger the usage of these, not all need to run simultaneously at all times.

Working on autonomous mapping and navigation for drones. finished mapping, now comes the nav part.

RT @ThreatT0Society: Tech X, i need your help 🥺. I need access to some deep research models for resource compilation and reasoning. which o….

Implementing NeRF SLAM w autonomous path planning next, wish me luck 🍀.