fc-updates

This is rolling out now! 

May see some funkiness for a few hours as we update the counts for all. users but by the end of the day, your follower counts will not include anyone who has been labelled "likely to be spammy"

Why did my follower account decrease?

1. Yesterday, we changed how Warpcast counts followers by excluding users our algorithm flags as spammy.

2. Twitter (at least historically) and Instagram both do this.

3. We've had an uptick in spam recently, so as part of improving our spam models we decided to make this change.

4. View this less as losing followers and more us giving you a more accurate (but not perfectly accurate!) view of your audience. 

5. There are additional improvements we can make here and we will likely do that over the coming weeks and months.

6. Important: this change only affects Warpcast’s display—it doesn’t alter protocol data; those accounts still follow you. And if you follow a spammy account, i.e. you don't find it spammy, you'll see their casts.

7. Third-party apps can access Warpcast’s open-source spam labels.

8. Please see the next cast for other thoughts I've shared.

Why did my follower account decrease?

1. Yesterday, we changed how Warpcast counts followers by excluding users our algorithm flags as spammy.

2. Twitter (at least historically) and Instagram both do this.

3. We've had an uptick in spam recently, so as part of improving our spam models we decided to make this change.

4. View this less as losing followers and more us giving you a more accurate (but not perfectly accurate!) view of your audience. 

5. There are additional improvements we can make here and we will likely do that over the coming weeks and months.

6. Important: this change only affects Warpcast’s display—it doesn’t alter protocol data; those accounts still follow you. And if you follow a spammy account, i.e. you don't find it spammy, you'll see their casts.

7. Third-party apps can access Warpcast’s open-source spam labels.

8. Please see the next cast for other thoughts I've shared.

https://warpcast.com/v/0x5c4029e2
https://github.com/warpcast/labels

A few thoughts

1. Spammy accounts can be (and more often than not are) humans, they are not necessarily "bots", i.e. run by software

2. When people find bots useful, we call them "AI agents" :)

3. Twitter used to regularly purge followers like this from their UI; Instagram still does this today. Meta estimates that 10%+ of their users are bots, read it in their 10-K

4. There are likely mistakes in both the spammy user classification and users that are actually spammy but have yet to be classified.

5. I don't think anyone is/was fooled that follower accounts on any public social network—web2 or web3—are all real people. So better to adjust to what we think is the best known proxy for realistic and keep making progress on other stuff.

https://warpcast.com/dwr.eth/0x2d8bf05b

A few other thoughts

1. We use a machine learning algorithm to classify spammy users.

2. That requires a manual input when new spam "metas" emerge.

3. Best way to model spam (and fraud or any other "gaming" of a system) is like an organism that evolves based on the environment (the social network) and the various inputs into the environment, i.e. antibiotics / spam filters

4. So what works effectively today will not be effective tomorrow (antibiotic resistance!). 

5. So you have to flag the new metas to the model using signals—user reports, manual classification—and then the model can relabel users.

Working on Farcaster and Warpcast. Longer thoughts at https://dwr.email

Additional thoughts

https://warpcast.com/dwr.eth/0xc216a2fd
https://warpcast.com/dwr.eth/0xc01e40b2

Can you solve spam at the protocol-level?

1. Assumption: a credibly neutral decentralized social networking protocol is desirable.

2. In order to achieve "credibly neutral decentralization" for the protocol data it must be 1) replicated across many servers / nodes 2) user-controlled, i.e. the user does not have to rely on any other party to add / remove messages to the protocol 3) sign ups are permissionless.

3. This means that anyone (or AI) can sign up for an account and post whatever they want at the protocol level. Any compromise on this negates the goal of "credibly neutral decentralization" for the protocol data.

4. However, just because the protocol itself is credibly neutral and decentralized *does not mean* clients / apps have to be. In fact, opinionated filtering is desirable from a user experience point of view.

5. So the practical end state: protocol is as neutral as possible, clients / apps get to choose what to show from the protocol, building a client is permissionless.

Not a Dev..... Not A MeMe Coiner Enthusiast... Like The Majority Of You Just Lost Looking For Better Days in Crypto  

How do we know if we are marked as a spammy acccount? XD
Asking for a friend

My count decreased by 30%, but I always knew many of these followers were just bots/spam. The current count is a higher quality metric and I prefer it.

Can we see this purging action more often?
I love seeing the bot/spammy account removed

I know a couple of great builders marked as "likely to be spammy" :/ 
How can they get out of that label? One of them is an artists collective, so they post a lot of their events, drops, etc. Is that spam around here? what is spam here?

I know is counter productive to say what classifies as spam, but how about a hint? Kinda frustrating onboarding true builders just to get them labeled as spam.

Thank you for all the work you lgnds are doing 🙌🏽

Awesome job! There seems to be an uptick in bots/farmers too

I'm sure you'll sort all these things out.

My followers count decreased by 20% but don’t you think we should remove spammy accounts from following too?

And you please add edit option to our cast so we can re-edit and post it back whenever we made an error??  @dwr.eth

Do you have a spammy checker that we can use to check accounts? I remember someone had made a tools you could use here, but accuracy was a bit dodgy