Build the agent setup first, pick the model second: a Linux + Tailscale + llama.cpp stack guide
max_paperclips · x · 2026-09-17
A long post for anyone starting with agentic systems this month, arguing that your setup decides whether the flow keeps running, while the model only decides how smart the flow is — and almost everyone gets the order backwards.
Recommended foundation for running agents across a laptop, multiple desk nodes and a phone:
- Linux, non-negotiable: every inference engine ships for it first, every agent expects a shell, and every tool below assumes it. WSL gets you to the door; a real install gets you in the room.
- Tailscale: put every machine you own — laptop, desk box, phone, a rented node — on one private mesh. Payoff: the model serves on one box and every device talks to it from anywhere, no port forwarding, no public IP.
- Pick one engine: llama.cpp is the starting point and personal inference engine; move to vLLM when you go multi-GPU or multi-box, with tensor parallelism for scale.
More from coding & agent
- Label the Row: A Six-Step Data Classification Cheat Sheet for AI Products — blaizedsouza · 2026-09-17
- What happens when a RAG agent retrieves a poisoned document? A reusable security test case — Tophant_ · 2026-09-17
- Agents should translate messy requests instead of forcing forms: lessons from a broadcast media deployment — NumbersProtocol · 2026-09-17
- Developer builds GitHub Radar, a UI to track activity across repos and organizations — dSebastien · 2026-09-17
- smallpt: A physically based ray tracer in just 99 lines of C++ — blaizedsouza · 2026-09-17
- Non-engineer builds a video editing pipeline in one morning with Dify, Gemini and a coding agent — Individual_Reality_6 · 2026-09-17