co/core is a place where people share the compute they already own to run AI for each other, instead of renting from a handful of giant providers. It's an experiment in inference we build, share, and own together — and your existing code works as-is, because we speak the same standard API everything else does. The models are open ones that run on the hardware people already have.
I love this app; the concept of us all helping each other build and benefit from the “new” technology that is LLM AIs, not just those who can run gigantic datacentres. If you’ve got a MacBook or Mac Mini, you can lend a hand to help community projects work sustainably (like alt-text for microblog posts!) And if you build apps on atproto, this is how you can provide useful AI features that don’t bankrupt you, or line someone else’s pocket.
This is a ton of fun, for a highly specific type of person… if you like playing with ai models, don’t mind using things that are experimental and might break, and are not going to rely on it for ‘important’ stuff (right now) you might be that person. The promise of locally running models freely available to anyone with a bit of compute to share is lovely. Having an easy way for atproto apps to spin up some ai that will use the users tokens, because it’s all sharing the identity you already have, is also just practically very interesting. Try it out, go hook up that Mac mini.
