@nickbearded i like looking at the llm leaderboard but lately ai wise i have done this - got exo/exolabs clustering app going, tried a model with llamafile and am investigating localai - they have p2p, federation so groups of people can focus interence and training on that sectors siloed and specialized data...

localai has had p2p ai for like 10 mos - being able to run it on a couple boxes cpu only and offload embeddings securely would be nice

you probably should run a gpu or two or three in your cluster but it is not totally necessary - you can process embeddings/tokens for local data inclusion into vector db and do like 1tb in 6 days vs the job taking over a month cpu only

I like the idea of setting up federated group to match up with portals for anything but for specific biz sectors and verticals could be v helpful cause then clients could have p2p ai data lake with relevant topical biz data - it is basically what i alluded to in 90 pt plan but now much more concrete

How Much Can You Get Done in a Few Days?

In a weekend or 3–4 focused days, with a few machines you can:
✅ Spin up LocalAI on 2–4 nodes
✅ Join or form a federated network
✅ Deploy several quantized LLMs, image generators, or audio models
✅ Run small-batch inference jobs across the network
✅ Offload some heavy jobs (like summarization, embeddings) to swarm partners
✅ Start offering services like local search, chatbot assistants, or automated data pipelines for your biz

With good orchestration, you can match or exceed what a $500–1,000/month cloud bill would buy. >>>this is doubtful initially but could be a productivity boost and help sales when people see there is a lot of industry specific info

#unused cycles