Why should anyone use Microslob? Just started switching from Mac to Linux.

As a European citizen I started switching to local LLMs wherever possible. We have basically understood the massage of the US government and companies. After several unethical decisions by OpenAI I quit my license (as ten of thousand other users did as well). For Coding purposes I still use Opus 4.8. The block has not affected me at all. If I really want to use this model I call my friend in the US and pay him a few bugs to use his license with a VPN. Imho this ban is just a door opener for a black market for tech-savy people.

Thank you so much for sharing this blueprint. I was kind of surprised that one of the first tasks would be the planning of the LAN architecture. That is finally the piece I need to get started but I have some additional questions: - ZimaBoard running bare-metal OPNsense as my edge router. Would you recommend a PCIe Intel 2.5G network card extension to upgrade from 2 to 4 ports right from the start? I'm planning to start with 4 VLANs. - I work as a freelance Data Scientist and AI engineer and would like to run local ML/AI environment and hooking it into N8N automation pipelines. Currently I run everything on a Macbook Pro M2 Max using local tools. Would you recommend using the eGPU passthrough right from the beginning or is this a very steep learning curve?

@David Vincent I have tried it and would recommend to use an independed subnet and completely isolate that thing from your standard network and any of your devices. It's behaviour is creepy and astonishing at the same time. It has installed loads of software packages over night and did not really tell me about it 😅

I use Python in VS Code with extensions to Continue and the CLI of Claude Code for my business projects. I have several agents running on N8N on-premise combined with the frameworks LM Studio and Ollama with open-parameter models like chatgpt-oss or Llama 3 and Quen. I've tried OpenClaw once and shut it down after one night. It's behaviour is creepy and astonishing at the same time. My general recommendation would be not to install this on your local subnet and give it no access to any of your local environments, users or passwords. I have just ordered a eGPU docking (new 125 EUR) with a used Sapphire Nitro+ RX 5700 XT (used 150 EUR) as an addon for my old MacBook Pro 2020 (intel based) and would like to compare the performance to my MacbookPro M2 Max and outsource business agents on this device for my clients. Besides that I also think about upgrading my Lenovo m720q with a NVIDIA RTX A2000 12GB at a later point. I wonder how many tokens I can create with that environment.