@UK_Daniel_Card I’m managing my home lab systems with a claude install on a pi node. Pretty awesome to be able to do most things from my phone while remote.
Hey Claude. Install XXX on node YYYY for user ZZZ
@UK_Daniel_Card Unplug the system.
Bury it in a 6 foot hole.
Cover with concrete.
Armed security to watch
That’s as close to total security as you can get
@MaryTilesTexas Depends on your use. I got one to basically play whatever I want whenever I want, but it’s not super duper amazing graphics. It’s the convenient 90% solution.
@ZackKorman Depends on the audience.
Skilled engineers who keep their hands on keyboards. Make it funny
Business folks or new clients. Keep it professional
Agreed — and I’d push it harder: it’s not “won’t be true for long,” it’s already not true. That separation was never a law of nature, just an artifact of how the tooling got wired. AI is one more automation layer on top of that — it widens what you can instrument and act on without standing up a fresh silo per domain. Which is exactly why “AI SOC” as a product category leaves me cold too: it’s selling the silo back to you.
Production has been folding detect / investigate / respond into a single motion for years — SOAR playbooks with reasoning agents layered onto an observability stack, Prometheus and Alertmanager landing where the SIEM already does. Multi-agent setups are closing the sysadmin loop in lab builds now: ingest the telemetry, hold the IaC baseline, fire a contained remediation on the routine stuff instead of kicking every detection into its own queue. Messy, obviously — the playbooks rot, the agent does something dumb at 0300 and someone pins it back by morning. But the wall between detection and investigation (and response, while we’re at it) reads like a diagram of last decade’s architecture, not a constraint that survives the next one. You’re right not to care about the category. The category is the part still betting the wall stays up.