Content
@
0 reply
20 recasts
20 reactions
Yassine Landa
@yassinelanda.eth
1/3 Excited to announce that @mbd has secured $3M in pre-seed funding to bring AI-powered recommender systems to Farcaster builders! Co-led by @masknetwork, @polymorphicvc w/ participation from @a16zcrypto /CSX, @socialgraphvc, Forward Research and WAGMI. Creating engaging clients—whether trading-focused, art & NFT-focused, or even AI agents—requires more than open access to raw data. To sustainably engage users, applications must deliver personalized, spam-free quality content to their users. Since early 2023 at @mbd, we’ve been building advanced machine-learning recommendation systems pre-trained on Farcaster data and designed to deliver dynamic, relevant content and actions tailored to each user. All are accessible through a dead simple API.
31 replies
24 recasts
106 reactions
Apurv
@apurvkaushal
this seems interesting. how do you think it would differ from products like Daylight (which kinda already gives social graph based recos) ? Also, curious if you've been using an ensemble model of sorts to train?
1 reply
0 recast
2 reactions
Yassine Landa
@yassinelanda.eth
Hey! We use machine learning to predict what the user will do by creating a score in realtime for everything onchain for all users, not rules based or what waitlist you are on or airdrops you are eligible for. Deep learning gets better with more data so we are riding web3 data explosive growth 🚀 You can ML to « rerank » existing airdrops/quests you are eligible for so mbd is complementary with daylight. We use transformers architecture like LLMs but for FC content call it Large Farcaster Models LFC 😄
0 reply
0 recast
1 reaction
