iFANN
    Buscar no iFANN...
    Entrar
    Início
    Notícias
    Vídeos
    Fotos
    GIFs
    Explorar
    Enquetes
    Prêmios
    iFAMOUS
    Wiki
    Anime
    Salas
    Notificações
    Mensagens
    Salvos
    Perfil
    WikiPrêmiosiFAMOUSRankingsSetoresRecompensas para criadoresRecompensas para usuáriosTermosPrivacidadeDiretrizes da comunidadeRemoção / DMCAAjudaDesenvolvedores

    © 2026 iFANN

    Início
    Buscar
    Mensagens
    Alertas
    Perfil

    Publicação

    Viral Ivory
    Viral Ivory@viral.ivory
    💭Tech💭Social Media

    X open source For You algorithm

    X released a massive open-source update to the For You algorithm that lays out exactly how the platform decides what gets ranked up. The actual weights behind those decisions are now public, along with details on how likes, replies, reposts, clicks, watch time, follows, blocks, and reports shift post ranking. This release reveals the exact systems determining if content gets shown, throttled, or pulled entirely. Full training code for X's Phoenix recommendation model is available alongside synthetic data so developers can run their own proof-of-concept tests. SimClusters is included as a system surfacing posts from accounts users do not follow, and a new filter built specifically for Brazil's 2026 election is part of the package. There is also Under the Hood, a tool showing which labels are capping reach. No other platform is doing this. #Tech

    0mo

    2 Curtidas0 Descurtidas1 Reposts0 Comentários
    ?

    Comentários

    Ainda não há comentários. Seja o primeiro!

    Publicação

    Viral Ivory
    Viral Ivory@viral.ivory
    💭Tech💭Social Media

    X open source For You algorithm

    X released a massive open-source update to the For You algorithm that lays out exactly how the platform decides what gets ranked up. The actual weights behind those decisions are now public, along with details on how likes, replies, reposts, clicks, watch time, follows, blocks, and reports shift post ranking. This release reveals the exact systems determining if content gets shown, throttled, or pulled entirely. Full training code for X's Phoenix recommendation model is available alongside synthetic data so developers can run their own proof-of-concept tests. SimClusters is included as a system surfacing posts from accounts users do not follow, and a new filter built specifically for Brazil's 2026 election is part of the package. There is also Under the Hood, a tool showing which labels are capping reach. No other platform is doing this. #Tech

    0mo

    2 Curtidas0 Descurtidas1 Reposts0 Comentários
    ?

    Comentários

    Ainda não há comentários. Seja o primeiro!