VCN #41: Bare Metal

the vibe

No API bill. Fully private. Your model on your metal. Season opener. Every coding agent you run today phones home to someone else's GPU and bills you per token. Tonight we cut the cord. Point a coding agent at a model running on your own laptop, watch it complete a real task, and walk out with a rig that costs nothing per run and never leaves the building. VCN #41: Bare Metal is the night we run a coding agent against a local open-weights model. Format: The walkthrough. Wire a coding agent (Claude Code / Cline / Aider) to a local OpenAI-compatible endpoint. Stand up the server (Ollama / LM Studio / vLLM), pull an open-weights coder model (Qwen3-Coder / GLM / DeepSeek-Coder / Llama), point the agent at localhost. Local vs hosted, side by side. You have z.ai + Claude Code in hand for the hosted baseline. Run the same task against your local model. See where open-weights holds and where it bends. Hands-on build. Drive a real task end to end on your local model. Laptop can't host it? Run the open-weights model on Nebius Token Factory credits instead and keep going. Demos. Show the room your agent closing a task with zero tokens billed to anyone. By 10pm you have a coding agent completing a real task driven entirely by a local open-weights model. No bill. No data leaving your metal. Builders only. Bring a laptop and a task you'd hand a coding agent. Doors 7pm. Walkthrough 7:30. Frontier Tower Floor 10. Hosted by Vibe Coding Nights: Rayyan Zahid (Immersive Commons), Michalis…