View quoted note →

What I find interesting is that we can have an agent that will be able to maintain this (balancing channels, restarting services, applying fixes). Historically it’s been painful to keep stuff like this running (it still is), but the light is at the end of the tunnel where this kind of thing can be kept running for the lay man.

🙏

Ram upgrade from 32GB to 64GB and RTX3060 > RTX3090 (from another donor)

Freedom flex: your home server’s harder than most nations’ IT

Looks more professional than my old mini HP G2 800 (16gb + 2 TB)😂👍

how did you setupt *arr for piracy? do you have like a fully made docker image or something like that? i try it everey once in a while but it so boring and long to set everything up

They’re all in a stack together with their own network and using gluetun with Proton VPN (wireguard). I couldn’t get QBitorrent to play nicely, so used Transmission instead. Happy to share docker-compose file. Will open the repo up but needs tidying up first.

Is there Datum server as well, for sovereign mining?

you're so smart to do the node, should need quite some courage and opsec to play with sats. I'm too dumb even to build a home server using OpenMediaVault and docker containers on a RPi and I'm stuck at trying to tunnel the server thru a VPN IPv6 server.

Nice, I run a similar stack. What's the host OS ?

Wouldn’t keep large funds on hot storage like that. I think the rate AI is coming along you will be able to point it at a repo and say “I want that one!” and it will do the whole security check stuff for you.

Kubuntu 22.04 with KDE Plasma with X11. I did try KDE Plasma Bigscreen but didn’t work out. Supposed to have voice control integration. Maybe it’s buggy - or maybe just me. Will revisit though.

yeah still, quite impressive work

Dude. Sounds awesome

That’s awesome. Currently working on mine. What’s your hardware setup and OS?

Very cool project. I’m not sure I’d want to mix my node with my media and other stuff though. Similar to what Umbrel is doing but you have the sovereignty and peace of mind. 🙏👍

Yeah, it’s a trade iff between having too many PCs all over the place. Am is to have one that I can keep an eye in easier. They’re all on docker containers so pretty separate (like you would with a VM in the cloud). That said, it’s a hot wallet ultimately … so …

What’s a datum server?

OCEAN

DATUM Setup Guide

For created your own block templates

Ah ok. Then it passes them to miners you may have - even solo miners. Nice. “Using Bitcoin Knots is highly recommended. This gives miners fine controls over how they wish to construct their block templates. Other node implementations that support GBT can also be used. This includes Bitcoin Core, but it is severely lacking in template control options. That is unfortunately a centralizing force which partly defeats the purpose of decentralizing block template creation in the first place.”

Docker containers are definitely the right way to go. I get not wanting multiple machines running all over the house.

Hi Ben, how will this be different than what Start9 and Umbrel are doing? Will it be built into an ISO for spinning up a VM on Proxmox or is the theory to build all from source?

This is mainly for my own use really, but could open up the source. Documenting as I go (including hardware) so Claude Code or similar could spin the machine up from documentation, docker compose files, and scripts. Start9 and Umbrel are great, but this is really meant for a bulky piece of hardware for all your home server needs running docker containers - including running reasonably sized AI models locally. Hardware: AMD Ryzen 5 7600, 64GB RAM, RAID array, Nvidia 3090, XBox Kinect camera and microphone, etc.

Sounds awesome. Are you familiar with ZimaOS?

GitHub

GitHub - IceWhaleTech/ZimaOS: A Simplified, Focused and Open NAS OS_

A Simplified, Focused and Open NAS OS_. Contribute to IceWhaleTech/ZimaOS development by creating an account on GitHub.

Cool. What LLM are you running and at what resolution? How’s the performance (tps) and accuracy on the single 3090? Do you think it can replace Claude?

Trying a few different ones at the moment, such as Llama3.2:3b, qwen2.5:3b, mistral:7b, deepseek-r1:7b and deepseek-r1:14b. Right now qwen2.5:3b (Q4_K_M) is doing around 144 TPS (according to Claude Code). Deepseek-r1:14b was too much for it. Here’s what I’m working on:

GitHub

GitHub - knowall-ai/Nod.ie: On-device local AI - built with Electron and Kyutai Unmute

On-device local AI - built with Electron and Kyutai Unmute - knowall-ai/Nod.ie

I.e. an agent that runs on your machine that you can just talk to, all running locally, that ultimately can call MCP Servers concerning the operation of a Bitcoin Lightening Node. Under the hood it’s using a fork of Unmute for the backend services, but I’ve added MCP tool calling capability and “thinking”. However, this is also using a lot of other GPU activities such as Speech-To-Text (TTS) and Text-to-Speech (TTS) and I was even trying to add video processing stuff (agent facial animation) using MuseTalk but it can’t quite all squeeze on a single RYX 3090 (24GB). Looking at options to run a second graphics card (until then putting animated AI on hold):-) I don’t think it will replace Claude Code (I use that a lot to build it) because of the clever way they grab context of a project code base and alike. If you’re referring to Claude Desktop, I’m using Open WebUI and seems pretty comparable (at least to Open AI - you can even get results from your local Ollama instance and compare it to results from OpenAI API which is pretty cool).

Have you tried goose?

No, I haven’t actually, just read about it. So it’s @jack's baby AI :-) Looks like a similar concept with locally running AI, full open source stack. Even uses an Electron App for the UI too. Will check it out thanks! @Block Open Source

It’s more along the lines of Claude code

Ah ok. Will give it a try. Using AI to build AI is pretty surreal especially how you can see it adjusting the way it’s “brain” works.

Being able to talk to Claude Code rather than type and then realise half of what you wrote didn’t get input as it asked a question would be nice. Perhaps Goose does that. Would allow for much more context to be passed to the agent.

Curious. What's the Xbox Kinect camera doing?

Not a lot at the moment, but for family calls etc. Has a microphone array which is pretty good for picking up audio from around the room. Would also like it to be used for Nodie (local AI agent I’m working on) to know who’s she’s talking to and distinguish different voices etc.