A reproducible recipe for a fully on-device coding agent on Apple Silicon: an MLX-served LLM exposed through an OpenAI-compatible endpoint, driving OpenCode in the terminal. Nothing leaves the machine — which is the point if you care about privacy, latency, or offline work.
Source / inspiration: Apple WWDC26 Session 232 — Run local agentic AI on the Mac using MLX (). The snippets below are
This guide shows you how to establish a persistent SSH connection from a remote Linux system back to your home server, so you can reach the remote host with plain SSH instead of going through vendor remote-desktop tools, web portals, or layered jump-host workflows.
The remote system usually sits behind a corporate firewall or NAT that won't accept inbound SSH — you can't just ssh remote-box from home, because nothing at the remote network's edge will route the connection to it. But that same firewall almost always lets the remote system make outbound connections, including SSH to a server you control.