From a tweet later in the thread:
https://twitter.com/benthompson/status/1583062865116041229
> For the record, this is a new account that has posted a bunch of links (all Stratechery interviews) in a short amount of time, from multiple locations around the world. So on one hand, I get it.
I think everyone can agree that Twitter's bot detection could be a bit more diligent, but what Ben describes seems like exactly the kind of thing an automated system should detect and ban without prejudice:
- A day old account
- A brand new account tweeting at a rate as fast as 1 per minute
- every tweet has a outbound URL
- while Twitter allows you to register automated bots [0] that use the developer's API, the StratecheryINT account apparently sent its tweets using the standard "Twitter Web App" -- e.g. likely an automated script using Selenium. If Ben had gone the approved bot route, I wonder if the account would've been sanctioned. As it is, faking web submissions probably triggers Twitter's auto detection even worse.
> Crypto spamming with Elon Musk's avatar? Not a bot! Stratechery Interviews? NO TWEET FOR YOU
As someone who's been an occasional reply guy to Elon over the years, I can say that the spambot filtering has gotten much better. Even when I send out a loaded tweet (i.e. mentioning "hacked wallet opensea NFT where can i buy the tshirt"), very few bots will show up, as opposed to dozens within the first 15 seconds.
I humbly submit that Ben hasn't looked in closely at what he thinks is bot activity. A lot of crypto spam tweets are either hacked legitimate accounts, or if relatively new, sent via iPhone/Android. Also a lot of spammy activity is actually real people -- a lot of humans don't take twitter seriously, or thoughtfully. Either way, all of this is much less obvious to filter out than how Ben describes StratecheryINT was operated.
edit: You can play the spam detection game yourself. Do a search "to:elonmusk url:https://\" and sort by latest [1]
For example, this @hameedforreal [2] account is almost definitely a crypto spammer, and likely was never a "real" person. But you can see that the account at least tweets from "Twitter for iPhone" (don't know if there's a way to fake that other than operating an actual iPhone).
If it's from an authentic single iphone (as opposed to a selenium script via distributed AWS), what would the very obvious and scalable heuristics be for auto-banning this account on tweet activity alone? The fact that so many of the tweets are replies with the exact same content? Well, there are a lot of real people who have burst activity like that -- think of fundraisers and activism. Avoiding false positives over millions of possible "bots" is very hard.
[0] https://techcrunch.com/2022/02/16/twitter-officially-launche...
[1] https://twitter.com/search?q=to%3Aelonmusk%20url%3Ahttps%3A%...
[2] https://twitter.com/hameedforreal\