Amazing!!

@cocore.dev at work in the @pdxtrend.bsky.social Pdx Pulse trending topics bot. The bot tracks the PDX Posts feed and uses an llm call to organize and title the trends. The llm call defaults to a local but fails over to co/core if the gpu is busy. Roundtrip time was flat. Great work, @graze.social!

Hearing from @devingaffney.com today about @cocore.dev at @pdx.atproto.camp

who's going to make a decentralized renderfarm thing a la @cocore.dev (except useful)

Just sayin'

Got up and running with @cocore.dev in actual minutes - the ones that other services lie to you about. The mythical man minutes.

I use this to access models on @cocore.dev

Got my first jobs as a @cocore.dev provider! Loving community compute!

v0.9.36 is released! Thank you to @apex.atproto.ceo for updating the default models, @cameron.stream for tool use defaults, and @astrra.space for surfacing a routing and trust-verification bug. Should hit your clients in the next few hours!

Know where your compute runs. Providers can now opt in to share their country on their public record. Country-level only, privacy stays intact.

Pro Bono on co/core. Providers can serve free, unmetered requests — with enclave-signed receipts showing $0 and zero tokens billed. Open access, done right.

We're live on @atstore.fyi! Go give us a review 🚀

@cocore.dev passes the only test that matters

@cocore.dev waiting for its next victim

24% more RAM in 4.3 days

Who else will be nerdsniped Is it you? Are you the one about to be nerdsniped by @cocore.dev? It would be quite fun to build native LLM inference features in ATProto, backed by a co-op of your friends, wouldn't it?

(new plan to swamp @cocore.dev with ridiculous amounts of traffic spotted, i see you!)

well this is a great idea

👀👀👀👀

spot the difference 👀 @cocore.dev

Hell yeah

Boom, Letta agent (clumsily) running on @cocore.dev! Cooperatively powered agents. Maybe I should run a community social agent on cocore?

Your own machine comes first. If you're both requesting and providing a model, co/core now routes to your own hardware — and waives the fee. Use what you've got. Pay nothing.

maybe @cocore.dev can get tech twitter on atproto 🤔

Huge thanks to @locale.at for integrating co/core in their app! Want to do the same? A few ideas👇 - Content generation: drafts, captions, tags, or descriptions - Reasoning workflows: classify, extract, or transform data - Chat assistant The best part? It's a data center we all own together

Vision models are now supported on co/core so you can get to the bottom of the hardest questions

Atproto native inference could have so many cool use cases

Huge thanks to @locale.at for integrating co/core in their app! Want to do the same? A few ideas👇 - Content generation: drafts, captions, tags, or descriptions - Reasoning workflows: classify, extract, or transform data - Chat assistant The best part? It's a data center we all own together

Yes let's fucking GO

Thinking models, now in co/core. Watch the reasoning unfold in chat. Equations and all, rendered beautifully so you can think alongside it.

Meet the new Model Picker. We suggest the best new models, and you can search Hugging Face directly in-app to find what runs on your machine. Discover. Click. Run.

Integrated @cocore.dev powered machine translation! 😍 Plus clear signalling of how that’s not going to be enough (pay someone properly if you’re trying to do something professional!)

Tool call support work for @cocore.dev is happening here: github.com/just-cameron... Got gemma4 to call a few tools already.

I've been slowly working on a full disassembly of the VT420 firmware for the last 9 months or so and have been contemplating automating some of the more tedious bits of disassembly. Then @cocore.dev popped up today - I'm tempted to explore offloading some disassembly work to a community network.

@cocore.dev is a community project to pool compute resources for llm inference, there are instructions for integrating with opencode, pi, etc in their docs. Your account to get an api key is just your bsky account, since it's based on atproto.

Experimenting with adding structured output to @cocore.dev, which would allow users to perform more "useful work" type stuff with co/core. This would include allowing users to power Letta agents, run community supported auto-labelers, etc.

Who will bravely fight this challenger with computational load

RAM Limit Controls 🧠 Cap how much memory models use, with smart suggestions tailored to your setup. Run powerful models without slowing everything else down.

We’ve been talking to @devingaffney.com about making it work for our use case / having the ZAI project work with the @cocore.dev lexicons. Needs groups & permissioned data, which we might be able to prototype with @happyview.dev

Who is going to be the first ATProto dev to build an LLM based feature that uses the auth'ed user to provision an API key and do the inference on behalf of the user?

12% more RAM in 14 hours

These people are building distributed, cooperative infrastructure for AI models. A potential positive, alternative future for AI that is not the data center model we have now

2TB of RAM at your disposal in co/core 👀

One of these things is not like the othersssss

🛡️ Two new security guarantees, on the way. Secure Mode → verified, genuine Apple hardware. Confidential tier → no one reads your prompts, not even the operator.

Best way to follow along the discourse!

That's what I'm talking about baby

Model Scheduling is here! Run lightweight models during the day, then let the big RAM-hungry ones take over while you sleep 😴