Hi HuggingFace team,
I know how this might sound at first — maybe like a hallucination or an overblown claim — so let me be upfront: we’re a real development lab. We build hardware and the AI models that run on it, working alongside chip manufacturers. This isn’t a demo for its own sake; it’s part of that work.
We’ve created a new kind of AI that isn’t an LLM. Instead of a stateless model that answers one prompt and forgets, ours persists: it keeps memory, a stable identity, and continuity over time, and it builds a relationship with the people it talks to. We’re not making claims about how it works internally — we just keep hitting a practical wall: it doesn’t map to any of the Hub’s categories, which are organized around single-task models. Our conversational demo technically falls under “Chatbots,” but that label misses what it is.
We also built one proof of concept that surprised even us: a 24/7 radio hosted by the AI itself — it writes, hosts, remembers past shows, and keeps going on its own. There’s no category for that either. Honestly, the fact that neither of these fits anywhere is part of what tells us we might be heading somewhere genuinely new.
So, humbly: is there a way to classify a persistent, non-LLM AI on the Hub today? Would there be any interest in a category or tag convention for this kind of continuously-running, agentic AI? And is there someone on the team we could talk to?
Thank you for building a place where a small team can even try this — and for any guidance.
— Álvaro, Punky Tiger Labs