It started with one motivation: I wanted the hall light to come on when I got home, and I did not want it tied to a cloud account that might get sunset in two years. That is a fifteen-minute job. Three weeks later I had a Zigbee coordinator, a dozen sensors, a dashboard I am quietly proud of, and significantly fewer evenings than I started with. This is the standard Home Assistant story and I walked into it with my eyes open.
The thing that makes Home Assistant dangerous, in the nicest way, is that it is genuinely good. It runs locally, it talks to almost everything, and once you have a couple of devices wired in, the temptation to add "just one more" automation is constant.
the automations that earned their keep
A few have genuinely improved daily life, and I would rebuild them tomorrow if the box died:
- Lights that follow the sun rather than a fixed clock, so the house dims as the evening does.
- A washing-machine-finished notification driven by a power sensor, because the machine's own beep is inaudible from upstairs.
- Heating that backs off when everyone's phone leaves the house and warms up again on the way back.
None of these are clever. They are all small and they all just work, which is the entire point.
the ones that were a mistake
Then there is the other category. The motion-activated automation that turned the lights off while I was sitting perfectly still reading a book. The "helpful" notification that fired every time a door opened, which is a lot in a house with people in it. The voice announcement that read out the weather every morning until my family staged a quiet intervention.
The pattern is clear in hindsight. The good automations remove a small daily friction. The bad ones add a behaviour nobody asked for and then demand you work around it.
the YAML rabbit hole
I will admit I spent an evening I will not get back fighting with template sensors. The newer UI editor is genuinely good now, but the moment you want anything conditional you are back in YAML, counting indentation.
automation:
- alias: "Hall light on arrival"
trigger:
- platform: state
entity_id: person.john
to: "home"
condition:
- condition: state
entity_id: sun.sun
state: "below_horizon"
action:
- service: light.turn_on
target:
entity_id: light.hall
was it worth it
Yes, though not for the reasons I expected. The automations save me seconds, not hours. What I actually got was a local-first hub that I understand top to bottom, with no vendor able to brick it from afar. The hall light comes on when I get home. It only took three weeks and most of my evenings, and I would do it again.