- buy the cheapest hardware that gets the job done
- ignore negative reviews that are software related
- upon receipt of devices, ask claude for local-first cfw
- have fun for a few hours acting as a meat proxy
- boom, cheap local-only smart home is done
Use a SBC or nas for home assistant. This process worked for me for switches, a thermostat, an air quality monitor, a thermometer, and a window fan.
Would’ve paid ~2.5x for hardware that had the appropriate local-first capability out of the box. Instead this all fit within an existing $20 subscription to an inference provider. DS 4.1 running locally was on track too just more slowly.
I wish Insteon had gotten more popular, Back in the day it was easy to get development material to write your own software or build your own devices. I really liked how the stuff all had a decent amount of capability on its own, or you could have a computer attached for more advanced automation. I still use a few Insteon things in my house, Still use a few X10 things for that matter. I really wish they had kept up their local support software.
Please please please use your first draft or refine it yourself without LLMs.
I managed to get to:
The cloud is the weak link
- No internet, no problem. ...
---
If you are writing for other machines to ingest it, whatever. But if you are writing for people it for me at least signals that you don't really care about the message that you want to get across, and are fine with a watered down LLM average.
Software stack: if you don't want to spend too much time tinkering with the software, go for something simple like HomeAssistant with nothing else around and you will be just fine. For most people any software is more hassle than they desire, all the extreme customizations are not really needed. If you feel so special, home automation is not that special, not if you want mass adoption and convenience.
Author here. This is ten years of home automation rules, led by one: nothing core should depend on a vendor's server. Setup core is Tasmota, MQTT and Node-RED on a Pi. I've given this talk a few times and started working on a third for the agentic era. What's the most pointless internet dependency you've found in a device?
:( Your notes are probably more fun to read than the generic AI voice. Your actual notes are fun, but seeing the ~extra cruft from an AI writer in every line irritated me enough to quit before the end.
I've bought multiple internet radios over the decades. The demise of them is always its dependency on a remote server which enables setting which stations you want to listen to.
The remote server always goes away after a couple years go by, and the radio is only capable of playing the stations that were set as presets. Then the preset radio stations go offline, or the URLs for them change.
My personal situation imposes even stricter versions of your #1 and #2:
1. No network, no problem. Anything core to the house has to work with Wi-Fi or automation server shut down.
2. Smart functionality must not remove dumb functionality.
As one example: having a physical switch that controls a light and the switch can be operated by automation is good. Having automation control a light and the switch just tells automation what to do is bad.
This does make things hard sometimes. For instance, your sprinkler controller doesn't pass that bar. In fact the only one I've found so far that has a local API for automation plus full "dumb" on-device control (no Wi-Fi, no app) is RainMachine, and they are gone. If only OpenSprinkler hardware had a touch screen for local UI.
I tried to stick to this philosophy, specifically the part about it working even if my local automation server is down, for a long time. But I finally found an automation case where I wanted functionality where I could not simultaneously have the functionality and have the switch continue working when the server was off.
Specifically, I wanted the switch to turn on different lights depending on the time of day. In the evenings in early morning I wanted it to turn on the dimmer floor lamps, and during the day, I wanted it to turn on the brighter Center room light. This was not possible without turning the switch from a light turning on device, to an automation activating device.
I thought really long and hard about if I was worth the trade off, and in the end I decided that my automation server was only likely to be down for significant periods of time if the power was out and the lights weren't working anyways. If I run into a situation where this becomes a problem for me, I'll reconsider my decision. Which is a great thing about the local control. If I make a decision that I don't like I can just change it.
Basically any appliance is completely ridiculous to require an internet connection imo. When are you going to need to use a stove or dishwasher without being home? What value do you actually get from having a smart fridge that you can view analytics about from your phone?
the only useful thing about a smart fridge is whether if it fails.... i hope it would make a noise. One trick i used was a cheap ble temp sensor in the fridge that would report to the BLE adapter of a tasmota device that sends mqtt messages to grafana. then i would get alerts if it was open or something broken. happened a few times!
Alexa is really convenient. i still think my rules still follow as long as core systems are still controllable offline. Which I really try to achieve through use of tablet on the wall (local wifi is still offline access in my head)
I'm not really sure what this is doing on HN tbh, as it reads like 2018 entry-level IoT talking points, but it's 2026 now, but regardless of that.
__
I suspect that for resilience reasons alone, we might see a shift towards connecting less stuff to the public internet.
Not because consumers would complain. No one cares about consumers - including consumers themselves.
But because of the threat to the stability of the system, internet connected stuff poses.
The fridge meltdowns are an example of that, but you can equally stun an enemy economy by just messing with the smart lights in people's homes, as it is the people who work.. or don't because they're tired.
The internet is not the weak link, the internet is a security nightmare. You don't want a Chinese guy shut you off because all your Tuya devices can be controlled by them. You don't even want a random hacker to set your fridge to heat, like the recent France incidents.
Also, QR codes as a good UX? I like what you are smoking, it is very effective in disconnecting from reality.
I think a smart home hub of sorts should be something you can install in any home, like a water heater or an internet router. The hub connects to your wifi and manages all of your smart devices. You register your devices with it using a web interface that is only just like configuring a router
The web interface part would be the really complex piece I think. You would have to have some way for devices to tell it what features they have, so it could dynamically produce the appropriate controls per device
I dunno if this is really feasible to actually build but it's one of the only ways I can think that would allow average non-technical people to do this without having to administrate their own hardware and network settings and such
my next iteration i talk about migration to home assistant. at the time i started this process HA was very difficult to use and not flexible enough for me. that has changed.
The new rules I have come to use are:
- buy the cheapest hardware that gets the job done
- ignore negative reviews that are software related
- upon receipt of devices, ask claude for local-first cfw
- have fun for a few hours acting as a meat proxy
- boom, cheap local-only smart home is done
Use a SBC or nas for home assistant. This process worked for me for switches, a thermostat, an air quality monitor, a thermometer, and a window fan.
Would’ve paid ~2.5x for hardware that had the appropriate local-first capability out of the box. Instead this all fit within an existing $20 subscription to an inference provider. DS 4.1 running locally was on track too just more slowly.
My take currently. Exactly thos. Buy Sonoff things, Raspberry Pi, and do some vibe coding. All local first.
I wish Insteon had gotten more popular, Back in the day it was easy to get development material to write your own software or build your own devices. I really liked how the stuff all had a decent amount of capability on its own, or you could have a computer attached for more advanced automation. I still use a few Insteon things in my house, Still use a few X10 things for that matter. I really wish they had kept up their local support software.
Please please please use your first draft or refine it yourself without LLMs.
I managed to get to:
The cloud is the weak link
- No internet, no problem. ...
---
If you are writing for other machines to ingest it, whatever. But if you are writing for people it for me at least signals that you don't really care about the message that you want to get across, and are fine with a watered down LLM average.
got it, the first draft was a set of rough notes but i can clean up today.
Software stack: if you don't want to spend too much time tinkering with the software, go for something simple like HomeAssistant with nothing else around and you will be just fine. For most people any software is more hassle than they desire, all the extreme customizations are not really needed. If you feel so special, home automation is not that special, not if you want mass adoption and convenience.
Author here. This is ten years of home automation rules, led by one: nothing core should depend on a vendor's server. Setup core is Tasmota, MQTT and Node-RED on a Pi. I've given this talk a few times and started working on a third for the agentic era. What's the most pointless internet dependency you've found in a device?
:( Your notes are probably more fun to read than the generic AI voice. Your actual notes are fun, but seeing the ~extra cruft from an AI writer in every line irritated me enough to quit before the end.
I've bought multiple internet radios over the decades. The demise of them is always its dependency on a remote server which enables setting which stations you want to listen to.
The remote server always goes away after a couple years go by, and the radio is only capable of playing the stations that were set as presets. Then the preset radio stations go offline, or the URLs for them change.
Love your rules.
My personal situation imposes even stricter versions of your #1 and #2:
1. No network, no problem. Anything core to the house has to work with Wi-Fi or automation server shut down.
2. Smart functionality must not remove dumb functionality.
As one example: having a physical switch that controls a light and the switch can be operated by automation is good. Having automation control a light and the switch just tells automation what to do is bad.
This does make things hard sometimes. For instance, your sprinkler controller doesn't pass that bar. In fact the only one I've found so far that has a local API for automation plus full "dumb" on-device control (no Wi-Fi, no app) is RainMachine, and they are gone. If only OpenSprinkler hardware had a touch screen for local UI.
I tried to stick to this philosophy, specifically the part about it working even if my local automation server is down, for a long time. But I finally found an automation case where I wanted functionality where I could not simultaneously have the functionality and have the switch continue working when the server was off.
Specifically, I wanted the switch to turn on different lights depending on the time of day. In the evenings in early morning I wanted it to turn on the dimmer floor lamps, and during the day, I wanted it to turn on the brighter Center room light. This was not possible without turning the switch from a light turning on device, to an automation activating device.
I thought really long and hard about if I was worth the trade off, and in the end I decided that my automation server was only likely to be down for significant periods of time if the power was out and the lights weren't working anyways. If I run into a situation where this becomes a problem for me, I'll reconsider my decision. Which is a great thing about the local control. If I make a decision that I don't like I can just change it.
The best devices have a local web interface so you can control them with your computer (like how you control your router).
Unfortunately, almost no lan connected devices sport such an interface.
Basically any appliance is completely ridiculous to require an internet connection imo. When are you going to need to use a stove or dishwasher without being home? What value do you actually get from having a smart fridge that you can view analytics about from your phone?
the only useful thing about a smart fridge is whether if it fails.... i hope it would make a noise. One trick i used was a cheap ble temp sensor in the fridge that would report to the BLE adapter of a tasmota device that sends mqtt messages to grafana. then i would get alerts if it was open or something broken. happened a few times!
This is absolutely great stuff, but I think I caught you red-handed there. Alexa Routines break your fundamental rule!
Alexa is really convenient. i still think my rules still follow as long as core systems are still controllable offline. Which I really try to achieve through use of tablet on the wall (local wifi is still offline access in my head)
I'm not really sure what this is doing on HN tbh, as it reads like 2018 entry-level IoT talking points, but it's 2026 now, but regardless of that.
__
I suspect that for resilience reasons alone, we might see a shift towards connecting less stuff to the public internet.
Not because consumers would complain. No one cares about consumers - including consumers themselves. But because of the threat to the stability of the system, internet connected stuff poses.
The fridge meltdowns are an example of that, but you can equally stun an enemy economy by just messing with the smart lights in people's homes, as it is the people who work.. or don't because they're tired.
The internet is not the weak link, the internet is a security nightmare. You don't want a Chinese guy shut you off because all your Tuya devices can be controlled by them. You don't even want a random hacker to set your fridge to heat, like the recent France incidents.
Also, QR codes as a good UX? I like what you are smoking, it is very effective in disconnecting from reality.
Decades ago who would’ve thought that a computer could fit in the palm of your hand.
The future of smart devices is a local LLM running on every single one.
I think a smart home hub of sorts should be something you can install in any home, like a water heater or an internet router. The hub connects to your wifi and manages all of your smart devices. You register your devices with it using a web interface that is only just like configuring a router
The web interface part would be the really complex piece I think. You would have to have some way for devices to tell it what features they have, so it could dynamically produce the appropriate controls per device
I dunno if this is really feasible to actually build but it's one of the only ways I can think that would allow average non-technical people to do this without having to administrate their own hardware and network settings and such
You have described Home Assistant. That’s exactly how it works, including offline if you need it and auto discovery of devices on the network.
Home assistant would be similar to the software side of things sure
I'm talking about a fully built piece of hardware that comes in every house like plumbing and wiring does
my next iteration i talk about migration to home assistant. at the time i started this process HA was very difficult to use and not flexible enough for me. that has changed.
Home Assistant sells hub devices with various connectivity radios, including the newest one they’re announcing tomorrow.
https://x.com/home_assistant/status/2115496455876038961
Cool, I guess I'm not alone in thinking this is a valuable and useful thing
Thanks for sharing this, I had never heard of home assistant before