ML-Powered Social Discovery for Hugging Face - #1
Why discovery primitives like “Similar Creators,” “Bulk Follow,” and transparent “Why am I seeing this?” justifications are the right starting bets for AI communities.
Modern AI communities don’t just host artifacts—they host people.
If your platform helps builders ship models, datasets, or apps, the biggest unlock isn’t only infra. It’s social discovery: helping users find the right people and work, quickly.
In this piece, I’ll outline three low-risk, high-impact features that lift engagement while keeping the door open for emerging voices:
1) Similar Creators
When I follow /TheBloke, who else should I see next? A Similar Creators carousel answers this with a single scroll: overlap of followers, profile similarity, and content proximity. The UI can pair mini profile cards (followers, reactions) with why-text (“Commonly followed with /TheBloke”). This transparency drives trust and conversions.
2) Bulk Follow (Jumpstart the Flywheel)
Fresh accounts often stall because building a graph is slow.
A Bulk Follow pack—e.g., “Top LLama-2 7B finetuners”—lets a user follow a curated set in one click. It’s the fastest way to populate a feed with signal while still being opt-in. Packs can be topic, task, or region-based (e.g., “Popular in India”).
3) Justified Recommendations (“Why am I seeing this?”)
Great recs explain themselves. Examples:
- Out-of-network affinity: “Liked by /TheBloke”
- Profile views: “You visited their profile”
- ALS/Collaborative filtering: “Popular among people like you”
- Second-degree connections: “32 creators you follow also follow /Undi”
- Location priors: “Trending in New York”
These snippets boost click-through without adding noise.
They also create a defensible feedback loop: the more users engage, the sharper the graph becomes.
Balancing Popularity with Inclusion
Ranking purely by popularity hides new voices.
Counter with diversity constraints (e.g., reserve slots for emerging creators) and topic coverage. Discovery should feel both relevant and expansive.
Why Start Here?
Because these primitives:
- Minimize risk (clear user value, transparent UI)
- Bootstrap cold-starts (bulk follow + similar creators)
- Compound over time (graph quality improves with every action)
This is how small platforms punch above their weight—and how large ones stay fresh.
The document below is a more detailed discussion of how Hugging Face can implement a feed that can draw on many different signals and become the home feed of choice amongst their large and growing user base.
