Descriptions:
🐝 Join the community where all of this is running: https://mrc.fm/cmc
I built a local AI cluster out of Macs I already owned, and my AI agents now run for free, with no API key and no bill. This is Buzz Shared Compute: run an LLM locally, share it with your community, and point your AI agents at a machine in someone else’s house instead of OpenAI or Anthropic.
Everything here runs on a Mac Studio M3 Ultra, a Mac mini and a MacBook Pro: a local LLM cluster, self-hosted AI, zero cloud spend. Underneath it is the open-source MeshLLM project, which Block shipped into Buzz at the end of July. I show the full setup end to end, then the catch nobody mentions: the person whose machine answers your prompt can read what you sent it.
⏱ CHAPTERS
0:00 Answer From the Next Room: No API Key, No Bill
0:31 What Buzz Shared Compute Actually Is
0:47 Three Ways to Share Compute: One Model, Mixture of Agents, Split a 1T Model
1:28 Where Your Prompt Goes: Relay vs Peer to Peer
1:59 The Privacy Catch: They Can Read What You Send
2:29 How to Turn On Share Compute in Buzz
3:05 Which Local LLM Your Mac Can Actually Run
3:31 Create an AI Agent With No API Key
4:26 Proof: A Local LLM Answers Inside a Channel
5:01 Adding a Second Mac to the Cluster
5:36 Running Llama 3.3 70B on a Mac Studio M3 Ultra
6:11 First Look: The New Shared Compute Sidebar
8:11 Gaming PCs, Home Labs and Old Macs Joining In
9:06 Three Things That Will Bite You
10:20 Should You Turn It On?
🖥 THE HARDWARE IN THIS VIDEO
Mac Studio M3 Ultra 256GB · Mac mini · MacBook Pro M1 Max 24GB
Models: Gemma 4 26B, Gemma 4 E4B, Llama 3.3 70B
⚠️ THREE GOTCHAS
1. Sharing is Mac only at time of filming. Windows and Linux can use the pool but not lend to it. Check your own build.
2. There’s an open bug where the toggle sits on “starting” forever.
3. Don’t pick the biggest model your machine can hold. Take the recommendation. Reasoning models burn tokens thinking and your agent feels sluggish.
🔗 LINKS
Join the community: https://mrc.fm/cmc
Buzz: https://buzz.xyz
MeshLLM: https://github.com/Mesh-LLM/mesh-llm
Block on Shared Compute: https://engineering.block.xyz/blog/buzz-sharing-compute-powered-by-meshllm
What machine have you got sitting around doing nothing? Gaming PC, old Mac, server under the stairs. Tell me in the comments and I’ll tell you what it could run.
