Conversation

This Tweet was deleted by the Tweet author. Learn more

So, very interesting that this account is restricted, is a friend of Lea and was created on 13th July 2020 at 22:49:17UTC, while Lea was created 13th July 2020 at 22:23:03 UTC

Replying to

What a coincidence! Makes you wonder how many accounts he/she/it/they created between 22:23:03 and 22:49:17.

Replying to

It *might* be possible to find all the accounts created in a period, I think, maybe using the "snowflake" (sic) algo -

@jasonbaumgartne

would probably know - but beyond my capabilities, I'm afraid.

Replying to

and

I’ve not done any deep diving into Twitter data. I should read up on the subject.

Replying to

and

The twitter api documentation isn't bad, and Jason's github & gists are a mine of information!

Replying to

and

Also, I'm assuming you already have a Twitter dev account that you could use for this? I need to check the rate limits for user lookups but scanning an hour or two of time shouldn't take too terribly long.

Replying to

and

I do indeed (I wrote the account statistics reply-bot

@stattobot

) & I'm pretty well versed in the ins and outs of the free API , plus I also use Python, so no probs on those fronts ;) I assume the way to proceed is (broadly) similar to firehose reconstruction?

Replying to

and

Yep! Exactly. Basically we would need to create a Python method that reduces the space of user ids by taking advantage of the Snowflake algo. That involves breaking up an id into its epoch time component, server component and sequence id components.

9:44 AM · Jul 29, 2020Twitter Web App

Replying to

Basically the workflow would be to first do an exhaustive search for a small segment of ids to figure out which server ids are in play for the time range we would be working with. Once we get the server ids, we can then construct which ids we would need to scan.