Senior Linux engineer seeks guide to build local AI coding assistant

A veteran Unix/Linux engineer—terminally addicted to the console—has decided to embrace AI on his own terms. With only a few years until retirement, he recognizes that AI is reshaping software development and wants to adapt without relying on cloud services from Microsoft or OpenAI. He convinced his CEO (who shares the goal of reducing vendor lock-in) to spend a few thousand euros on a dedicated AI-ready machine. His only prior AI experience is a few hours with Claude Code to generate boilerplate.

His plan is to install Linux on the machine, run an open-source LLM, and deploy coding agents locally. He intends to start with trivial code generation tasks to save typing time, then gradually increase complexity until the hardware or model limitations become clear. This methodical approach is the 'reverse of vibe coding'—he wants to understand exactly what the AI can and cannot do. He's giving himself 3 months to set everything up and evaluate it. If AI proves disappointing or he simply dislikes the new workflow, he has told his employer he'll seek a non-computer position. He's asking the community for a guide tailored to experienced developers who know nothing about AI but want a local, offline setup.