The lab works best when it is honest about what is reliable, what is experimental, and what is still being figured out.
Reliable Core, Experimental Edge
DNS, remote access, backups, and Home Assistant need to be understandable because they affect daily life. AI agents, automation runners, desktop VMs, and new service experiments can be more complex because the point is learning.
Backups Are Not One Thing
Redundancy, snapshots, sync, and backups solve different problems. SHR helps with a failed drive. Snapshots help with quick rollback. Sync helps with availability. Backups are for recovery. A mature setup needs to know which layer is solving which problem.
Open Source Is Powerful And Uneven
Docker, Jellyfin, Guacamole, Home Assistant, Forgejo, and related tools make the lab possible. They also come with rough edges: odd client behavior, less-polished UIs, integration quirks, and maintenance chores. The tradeoff is control and learning.
Agents Need Guardrails
AI agents become more useful when they can see Git history, work in isolated environments, and operate with reviewable changes. The hard part is not making them powerful; it is making them bounded, observable, and reversible.
Access Should Be Boring
The most useful remote-access setup is the one that does not require opening broad firewall ports. Private overlay access handles trusted personal devices, while identity-protected browser access covers the cases where installing client software is not realistic.