An Obsessive Guide to How the Twitter Algorithm works:

Ok, imagine this: you’re a software engineer tasked with cleaning up Twitter. It’s supposed to be a respectable News Site. The new CEO is even a conservative. You want to be respected in the way that other news platforms are respected. You want to not get sued for having Nazis, fake news propaganda, and *Porn*. So what do you do? Let’s think like a software engineer trying to clean up Twitter. 

One of the inventions of the internet that was popularized by social media platforms like Facebook is the use of network nodes. There are entire web networks of people’s connections to each other designated by their personal interactions. This is visible in the natural world as well through things like viral contamination. Which is probably the best analogy for the transfer of shadowbans. (there are also some funny graphs using node connection networks, such as The Friendship Paradox (most of your friends have more friends than you); or the slut cloud). 

The Friendship Paradox: You are much more likely to know the Red Node because the Red Node is more likely to know more people. You are more likely to know people who know a lot of people than people who are anti-social and don't know a lot of people. 

The majority of Sexual Activity of people with a "high number of sexual partners" is happening within a "slut cloud" of those who are sleeping with those who also sleep with those with a high n of sexual partners. 

Twitter's algorithm employs user clustering, penalizing posts outside one's established cluster. Engagement metrics similarly discourage inter-cluster interactions. This system likely employs variable reporting thresholds as a spam prevention mechanism.

Not everyone will be reported, and not every report can be processed by a human. So, instead, Twitter estimates your likelihood of being bad by your connections to accounts that are directly bad/reported (or by bad behavior). Due to limited human moderation capacity, automated systems estimate infraction probabilities.

Hypothetically, content categories like nudity/gore may have distinct thresholds, potentially serving to insulate mainstream content. This algorithmic approach effectively curates user experiences within predetermined parameters, creating digital echo chambers.

Aka, if someone replies to your tweet, it has a 27x boost, but if someone reports your tweet, it has a negative 369x deboost. 

Basically, deboosting or shadow banning is a less explicit version of explicit shadow banning. It limits engagement and exposure to your post. There are those old shadowban checkers, but most of them don’t work because Twitter banned API integration.

Twitter closed its free API, so no external apps can use it unless they want to pay exorbitant fees or do something illegal, and none of them work anymore.

Accounts get clustered into groups based on who you follow, who follows you, and what posts you interact with. This is why most SWs are default engagement limited, but you’ll see normal accounts get like 7k likes on a photo that has the smallest bit of cleavage.

Random example that popped up on my feed but normie accounts can post more scandalous content cuz they’re not associated with sw.

It makes me sad cuz I wish I could engage more with provider accounts but every time I do my engagement totally plummets but the choice is basically to connect and get shadowbanned or isolate and have higher engagement.

One of the interesting things is if you go to X Pro on a Premium account, they have suggested groups, and you can loosely see the groups they filter your content into because of these suggested groups.

You will get actively punished for trying to post outside of your feed, but posting outside of our contaminated feed is also the only way to boost virality and main page engagement. It’s a catch-22.

However, I'm dubious about how much community account engagement matters for virality on Twitter. I spoke to a friend who consistently gets 5k-20k likes on photos, and she does basically no engagement?

I also noticed this weird trend where the posts I made while I was at Burning Man got several thousand likes, which kind of goes against my intuition that Twitter rewards "acting" like a normal account.

Ways that I initially set up new accounts:

I would schedule four posts daily: during the mornings before 9 AM, during lunch, and after 5 PM to come out for the next week. That way, you can have volume and then go on your feed/try to follow accounts. I’d follow like 20 to 30 accounts a day.

The goal is to try to make your engagement seem as “Human” as possible, or at least what a human thinks the most human human would engage with Twitter like in an ideal world sigh.

I recommend going on Twitter for 15-60 minutes/day, Splitting into groups at least once daily to engage with your feed and follow different people. 

My earlier notes for Engaging With Twitter (useful for lower engagement accounts/when first starting out):

Weird ways I’ve been able to tell I’ve been deboosted:

-Opening twitter for it to say 20+ notifications and then quickly changing to 7/8.

-Going to the most recent (photo)post and seeing if there’s suggested “posts like this”

-Who is my home page being clustered with is who is also seeing my posts: aka if my home page is people with under 50> likes you’re in the engagement pit. 

-the only way to fix this is restrict twitter engagement until the algorithm is giving the notification that there is 20+ likes & a gain of followers without active twitter usage, this means the algorithm is recommending your account/content without you having to engage.