Tony D’Addeo on Farcaster

Content pfp

0 reply

20 recasts

20 reactions

downshift 🌹⏳💀 pfp

downshift 🌹⏳💀

even when i test engagement farming tactics on this app (for science) they tend not to work as i would expect i truly don't understand their "algo" who's got the best insights on how it actually works? (we shouldn't trust Merkle to tell us how it works...for obvious reasons..)

8 replies

0 recast

17 reactions

Pichi pfp

What are your goals? Aka getting on the leaderboard is different from trying to get a trending cast which is different from getting views/impressions.

3 replies

0 recast

5 reactions

downshift 🌹⏳💀 pfp

downshift 🌹⏳💀

my goal is to understand how it works so i can warn/inform others

1 reply

0 recast

1 reaction

Pichi pfp

But which one do you want to optimize for? They require different strategies.

1 reply

0 recast

0 reaction

downshift 🌹⏳💀 pfp

downshift 🌹⏳💀

i'm making a statement that's anti-algorithm in general. fuck manipulating a content strategy to play into the hands of a wealthy central authority who stands only to profit of of you and all of your hard work. fuck that completely

2 replies

0 recast

1 reaction

downshift 🌹⏳💀 pfp

downshift 🌹⏳💀

my conclusion for now (until i have better data) is that Merkle reserves a secret nerf option for individual casts my most incisive and critical casts get almost zero engagement, especially when Merkle itself is in the crosshairs (when the rules of every traditional algo would see those blow up...everyone knows divisive content wins)

1 reply

0 recast

2 reactions

Tony D’Addeo pfp

lol absolutely not people either disagree and / or are less likely to engage with negative content especially if it’s inflammatory or conspiratorial

2 replies

0 recast

5 reactions

downshift 🌹⏳💀 pfp

downshift 🌹⏳💀

thanks for chiming in. i just want a dialog, and they are hard to come by. i might be too aggressive for your algo/our community, idk, i normally don't try to placate it

1 reply

0 recast

1 reaction

Tony D’Addeo pfp

as far as I know we don't yet do content level analysis in the algo it's based off of follow graph + engagement over time

2 replies

0 recast

1 reaction

androidsixteen pfp

@androidsixteen.eth

I think y’all do, but it’s not so sinister I’ve noticed the algo (prob @akshaan’s ML stuff) start to cluster related content as of a couple months ago

2 replies

0 recast

1 reaction

Tony D’Addeo pfp

we do this to generate trending topics and show them in those feeds but that is a different thing entirely / doesn't effect home algo

1 reply

0 recast

1 reaction

androidsixteen pfp

@androidsixteen.eth

I think it does — sometimes two posts with related topics / keywords are displayed in the algo feed next to each other I’ll try and find an example later

1 reply

0 recast

1 reaction

androidsixteen pfp

@androidsixteen.eth

Example @deodad https://farcaster.xyz/priyanka/0x57ad6781

1 reply

0 recast

2 reactions

Tony D’Addeo pfp

we are pattern matching machines and tongue in cheek but multiple posts on eth price on farcaster is not exactly a miracle

2 replies

0 recast

2 reactions

androidsixteen pfp

@androidsixteen.eth

Just check with @akshaan bro I might be wrong but I’ve seen it multiple times

1 reply

0 recast

1 reaction

akshaan pfp

the model doesn’t have any topics / text related features as input. The clustering of similar posts is usually due to some other factor (closeness in time, similar engagement etc.)

1 reply

1 recast

5 reactions

androidsixteen pfp

@androidsixteen.eth

Thanks for clearing it up! Sorry for gaslighting you @deodad

1 reply

0 recast

2 reactions

Tony D’Addeo pfp

no worries i had no doubts like an anthropologist doing field work watching a people’s realize their Gods aren’t real

1 reply

0 recast

6 reactions

androidsixteen pfp

@androidsixteen.eth

1 reply

0 recast

2 reactions