🧠 Vibe Coding with ChatGPT – Local & Pragmatic
I don't use ChatGPT as a hype tool but as a practical assistant in my daily sysadmin work. Everything runs locally – on a dedicated high-performance workstation using self-hosted Ollama and Open-WebUI, with no cloud and no external dependencies.
Why? Because solid prompts and the right vibe in coding bring real efficiency – especially for Linux sysadmins.
Examples from my workflow:
- Iteratively refine shell and Python scripts
- Generate systemd units, Nginx configs, and Postfix setups using GPT
- Context-aware test and deploy snippets instead of StackOverflow hunting
- Write Markdown docs, commit messages, and emails with the right tone
- Document and systematically improve prompt strategies
🖥️ Hardware Setup:
- AMD Ryzen 9 5950X · 16 Cores / 32 Threads
- 64 GB DDR4-3600 RAM
- GeForce RTX 4060 Ti 16 GB
- 1 TB NVMe SSD
- DeepCool air cooling
- Debian 13 · Headless + GNOME Workstation mode
🤖 AI Workflow: I use a multi-stage prompt loop to take advantage of different models:
- Ollama (local) with deepseek-r1:14b via Open-WebUI for fast, privacy-friendly ideation
- Grok for creative impulses and alternative perspectives
- ChatGPT Plus for polishing, structure, and project-ready results
📢 Offer: Local AI for your project Open-WebUI doesn't just provide a chat interface – it also offers API access. Perfect for your own tools, games, or web apps powered by a local LLM backend.
I offer dedicated access to my self-hosted Ollama instance or help set up your own environment.
Just ask: pragmatic, privacy-respecting, and vendor lock-in free.
🎯 Goal: Vibe Coding for admins – efficient, local, in full control. From idea to working setup in one flow.