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?
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)
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!
My non-smart fridge makes annoying noises when the door is open too long, and slightly less annoying noises the first time you open it after a power cut causes it to get too warm (I believe it depends on interior temp when the power comes back on?). It might make noises when it breaks in other ways, that thankfully hasn't become relevant yet
I guess connecting to home automation would be useful for a fridge that's sufficiently out of the way that you might not hear it. But for the fridge in my kitchen the noises emitted by the fridge itself are enough for me
:( 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.
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.
There’s a pretty active ecosystem of ESP32 projects for HVAC systems. I’ve been working on a few that add Apple Home and Matter support with a local web interface for basic and advanced features. The web interface exposes things Apple Home or Matter protocols don’t represent well. Vane positions, different modes (dry, fan only), equipment diagnostics, remote sensor control, and even peer to peer communication for fully offline control (no WiFi needed). Plus it gets me a control path independent of the home automation stack. There’s a lot you can put on the device itself these days.
With another project, based on the ESP32-S3, I was able to cram grafana style scrubbable charts, full activity log, refrigerant circuit diagram, equipment topology, etc. I'm pretty excited for the ESP32-S31 since it's a nice jump up in performance.
Shelly switches are great for this. For example, I have one behind a regular wall light switch for the front porch. The physical wall switch always toggles, but the local on-device script turns it on (if it’s not on) at sunset, and off (if it’s on) at 10:30pm. All configurable from the local web server. Only needs ntp (your choice) in case of restart after power failure.
Local web interfaces are great, yes. And it's pretty cool for me when devices have them. But they don't meet the no network requirement for people around me. Gotta have direct control of things. I mean, the landscaper isn't going to connect and use a web browser to adjust the sprinklers.
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.
There are some things that just don't have a dumb equivalent. I haven't run into any of those yet, but it's early days. At least having the one light not working for a while is probably not a disaster.
> 1. No network, no problem. Anything core to the house has to work with Wi-Fi or automation server shut down.
This is one of the reasons I love Zigbee, the device network is entirely independent of my computer network, can not directly communicate to the internet in any way, and every device I've put on the network so far can be configured to directly communicate with other devices so basic functionality remains operational even if the controller goes down.
For example, in my living room I have a Lutron Aurora dimmer dial locking a standard switch in the on position, then plugged in to the outlet that switch would normally control is a Leviton dimmer module. Both connect to the network controlled from a PoE Zigbee dongle, but have been directly paired as well so pressing the button on the dial makes it directly tell the dimmer module to turn on, rotation commands are sent direct, etc.
The controller also receives these commands when it's active and can add to the functionality for inputs like double-click, but if it's down the basic functionality still works entirely as a normal person would expect. You walk in to the room and see a knob on a switch plate, you turn the knob and the light brightness increases regardless of the state of the internet, my home server, or even my network as a whole.
The other major advantage compared to Matter/Thread is that it's not IP based so it's impossible for any Zigbee device to ever have a cloud dependency. No matter how much some marketing dingus might want to, it can not be ruined.
I haven't done any direct pairing yet, but Zigbee certainly rocks. Everyone is pushing Matter and Thread, but with the basic premise of devices being globally addressable, I just have no confidence of being in control of them.
I'm a strong proponent of everything-over-IP in an ideal world, and love the fact that Matter mandates IPv6 which forced a lot of embedded vendors to finally support it, but the IoT vendors have demonstrated that if they're given an internet connection they WILL abuse it to lock you in to their platform.
Matter over Thread is nice because it functionally provides an "IoT VLAN" by default and can only reach the internet or even the LAN through a gateway, which can be fully controlled by the user making it the best of both worlds.
As for Matter over WiFi, for the reason in the first paragraph I've maintained a rule of not buying WiFi IoT devices unless the device has a real need for direct access to the internet or high-speed data from the LAN to do what I want it to do, and Matter doesn't change that. It is nice though that hypothetically I could have a Matter-over-Thread switch cause a WiFi device to do something directly without needing to go through Home Assistant.
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.
same story as a meat thermometer someone told me about. the hardware outlives the server and still stops working. At least with opensource projects, if they are no longer maintained, its possible to fork and keep they going. hardware can be tough unless the device's firmware can be updated.
That's some good rules but if most products arent like that or these products with your requirements are more expensive, then is it really "OK"? This is a real big issue and isnt a technical one. Its not about if you have some obscure alternative or can hack it away in a weekend. The point is that 99% of home appliacnes or whatever, dont need a server or any concept of cloud or its jsut a feature which you turn on and off and nothing much changes besides backup or such. Also surely accounts or connecting your email isnt needed either.
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.
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
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.
I will be happy when this genre of whining fades away. Everyone has suddenly turned into the most obnoxious 7th grade English teacher imaginable. "I wanted to like your ideas, but you used lines from the reading assignment without quotes and 3 sentences had dangling participles. Plagiarism and grammar offenses, straight to jail."
It's not missing quotes, it's "i didn't write this at all, I put zero effort into it, I don't know what it says, it may be made up". Claude is good for code, terrible at prose, it boggles my mind that people use it to write PR descriptions, let alone their own personal thoughts.
> It's not missing quotes, it's "i didn't write this at all, I put zero effort into it, I don't know what it says, it may be made up"
In this case the article starts with "AI;DR: I wrote the first draft, AI helped me edit and rewrite it, and I made the final pass. The rules, setups, and talks are my own" so none of that applies.
It is not about the grammar though, that's just proof. It's about the overwhelming feeling that whoever "wrote" the piece hasn't read it at all - so why should you?
i did draft it and i'm still getting better at the over zealous llm auto corrections. i dont like some of the phrasing either. i am actively cleaning it up and i do appreciate the feedback since there is value in refining my process.
I appreciate it! I am another person would have read more of your post, but also stopped after too many instances of "LLM voice" ( https://martinfowler.com/bliki/LLM-Voice.html ). It's nothing personal, and I agree with your post's premise (I run HA and local-only myself) and love that you are spreading the word for others to learn about this important aspect of technology. I just don't want to spend my valuable time reading machine output, and I'm evidently not alone.
By all means, pass it through an LLM for review if you really want, but you don't need to fix all the issues it finds, and you shouldn't let it remove the personality from your text by rewriting things. Because they will eagerly do that.
I would much rather read a post with a human voice - don't be afraid to just do it :)
The issue is about moving the burden of verification to the reader.
Its like you went to a potluck and brought a salad with you, you just closed your eyes and threw some things from the fridge together without thinking, it's all mixed together now and you don't know if something was old/moldy.
You are proud and happy that you brought a contribution, and you feel like your responsibility/duty ends there while the people who want to eat it have to be careful and fish things out, or just avoid it because they can't be sure if it was old or not.
Natural language is a non-verifiable domain due to the ambiguity of the intent, so we will never get away from this issue in LLM-land. You can't know intent of the writer unless they wrote it themselves.
"This is written by AI" has become the new "First Post!" The complainers might as well just run an automated bot that posts it to every HN article, to save them the trouble of typing.
Yes, but the reason we might as well is because there's so much slop.
I think it interesting that people are more upset about the complaint than the cause of the complaint.
If this place — a place where people are supposed to be sharing interesting things for techies and geeks — gives up on complaining about thoughtless nonsense, its culture dies.
To me it's like complaining that the site hijacks the scroll bar or has annoying cookie popups. Yes, we all hate it. Yes, web sites shouldn't do it. Yes, it's something that should be pushed back on... But complaining about it adds nothing to the conversation.
I used to complain when web sites had huge whitespace margins on the left and right side of the text. I hate that! But most web sites do it, and complaining about it here accomplishes nothing and adds nothing to the conversation.
There is a growing body of evidence that people prefer ai generated text when its not labeled as ai generated. However given the same text with ai generated label, all of the sudden people prefer human written text. Just a fun fact about our human biases towards "made by ai" disclosure.
I mean if you run this as the usual social psychology experiment you already know what answer you'll get right? Pay 20 college students a couple bucks to review a set of texts, some human written, some AI written, and score how much they "liked" each text. Make the texts representative (ergo half the human texts are fucking awful, all the AI texts are, well, AI texts), and there you go, evidence! People actually prefer AI writing!
Or, of course, make the human texts excellent blog posts that fit the readers' area of expertise and there you go, evidence! People hate AI writing! The HN hivemind was right all along!
I do not know how accurate that is. Assuming it is true, I think it's a similar phenomenon that happens with CGI.
Tell people it's CGI, they won't be enthusiastic about it. But tell them it's "practical effects" and they will absolutely love it. Even when the movie has even more CGI than average... they will still praise it, and will even praise the CGI scenes they failed to identify as computer generated.
I can't claim to be able to detect LLM writing with 100% accuracy, but when I do notice it, it's very off-putting. Some claim that's because it's bad writing and that I would be fine with a higher-quality generated text, but that isn't true. I dislike LLM writing purely because it's generated by a machine, not because it's bad (though the bad quality does make it more jarring). In my opinion, handing other people something that a machine generated for you and asking them to read it is extremely disrespectful. I want to read what the human wrote, imperfections and all, because humans have intrinsic value that makes their work matter to me. The product of a machine, on the other hand, is just a facsimile trying to appear like a human made it, and that holds no interest at all to me no matter how good the content would be were it man made.
was your comment written by an LLM to try to justify itself :p
I'd like to see this evidence. I have very limited time and attention. LLMs are usually used to take the user's prompt and flower it up to 10x to 100x fluff. Just give me the prompt.
- 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.
RPi hardware is nice because it's a fairly-consistent, known quantity. All copies of a given board have the ~same features and bugs, so an image made for someone else's Pi 4 will run on my own Pi 4. This grants appliance-like utility that we haven't really ever had much of in the x86 PC space.
Pi hardware almost never shows up as cheap as a person might wish for. It's scarce on the used market, even when prices are high. In that space, demand outstrips supply.
And it's never been "cheap". A decade and a half ago when it first showed up, folks were already comparing the price (plus the SD card, and the appropriate power brick, and the USB cable, and the wifi adapter, and some kind of housing) to that of a used PC or laptop, and the Pi seldom came out on top.
But if we're doing new vs new, then it does alright price-wise.
And it's tiny. It's got some hackability built in, with the not-USB camera interface and the not-HDMI display support, and some GPIO that can be useful for some stuff.
It's got places where it's useful. I use them in those places. (And despite that lengthy defense, I run Home Assistant on a cheap SFF Lenovo box that came to me with RAM and an SSD for $50, delivered, in pre-used form)
Its a do-everything solution with a do-not-break-compatibility history to its design and a lot of attention driving its price upwards. These are good things as well, so...
Examples:
- SD cards are very convenient for hobbyists "experimenting" with OS variants. They are not great for performance or reliability; a soldered on storage device is common in many other products.
- The price is more than it seems - you need a case and a power supply and an SD card.
- The case situation is bad. If you want to 'set and forget' you really want a robust case, potted. Many N100's are designed for more abusive environments.
- Most use cases don't need the HDMI, Camera, Audio, Ethernet, or all the USB ports. That is board space, purchase cost, and idle power draw that can be reduced or dropped.
- It doesn't have some modern connectors, e.g. M2, because there is no space on the board, and the PCIe lanes have gone to the connectors you probably don't use.
Many stick based computers, particularly second hand and older generations, are better for most use cases. These are typically Intel N series; but are increasingly ARM, particularly at the lowest prices.
Absolutely this. For pretty much all use cases I have, for performance, power consumption, robustness, etc. is much better the ESP option or the “big” N100. Rpi is there in the middle. For some playing or prototyping I have some from the ol good times…
I personally don’t like raspian at all, and how they change things. They try to be “compatible” with very old HW, while at the same time many SW things I’ve done, are systematically broken in each new release.
My 2ct experience.
If you want something that works out of the box without too much work. Home Assistant Green is also a great little device. We first ran on a NUC (more powerful than N100), but then we switched to a HA Green. Passively cooled, doesn't draw a lot of power and has been super reliable. And it's well-supported by HA for obvious reasons.
On a home network with dynamic addresses: No need to worry about IP addresses; no overlaps happen, no mistakes, no spreadsheet table to manage. Use names (as long as DHCP works, then so does DNS). Renumbering the network is simple when that's useful.
It still works when the internet is down. And when the router itself breaks, we fix it or replace it and things work again. It's good to have a functional spare.
---
On a home network with static addresses: Micromanage IP addresses, keep a spreadsheet table, potentially make mistakes. Avoid using names (DHCP may be eliminated but DNS is still a failure point), making things harder to remember six months from now. Renumbering is arduous.
It still works when the router is down, but: When the router breaks, we still fix or replace it anyway. It's good to have a functional spare.
---
In terms of SRE: I don't need 5 9's of reliability for my smart light bulbs' IoT controls at home. They all still have some manner of physical switch, and they all can still be toggled on and off the old-fashioned way. Parallel controls work well-enough.
Everyone will get by almost-normally while I fix the router, which is a task that I'll set forth to be doing anyway for all of the reasons that household internet access provides.
I've recently come up with what I think is an better solution. If someone's already done it, I have no idea. I just went ahead and implemented it because I thought it'd be interesting.
The concept is that, instead of assuming that IP and MAC addresses are at all reliable, to "fingerprint" devices on the LAN based on things such as SSH public keys, TLS SPKI fingerprints, UPnP device UUIDs, mDNS, common TCP port profiles, and IEEE OUI manufacturer in order to form mostly-stable identities. The IP and full MAC are only fallen back to when there's absolutely no other data to work off of. My software uses a daemon that scans the network in order find devices and update existing device identities if IPs change. It serves hostnames for these devices on the `.home.arpa` TLD and creates synthetic IPv6 addresses for each device in case one would rather use their own DNS server or hosts file.
I recently "finished" this project, and it's kind of awesome because I was able to swap to a new router and everything just kept working as-is without having to think about it. Of course mDNS solves a lot of these problems already, but not every device has mDNS.
What's really cool is I recently realized that this is also very useful for accessing your local network with a VPN like Wireguard. Even though I only planned on this being something you run on individual clients, I put it on my OpenWrt router and configured it with dnsmasq so that, when I connect to my network with Netbird, all of my local hostnames just work on my phone by virtue of connecting to the network and using its DNS server. That said, my project doesn't require router admin access at all. It can even run as rootless if TUN access (for synthetic addresses) is turned off.
Maybe I'll release it publicly at some point, but if anyone wants to steal this idea and do their own implementation, I am not territorial over it. It's made getting my LAN devices working together a joy.
i use DHCP reservations since im a bit lazy and it seems to work good. It requires a reboot of the device on the network. at least for openwrt based routers.
Hypothetically, a long DHCP lease should let you ride out a short Internet/router outage. After that, hypothetically of course, APIPA addresses would be assigned, and I suppose that mDNS should be working even after APIPA addresses take over. I was unable to make this happen. I also tried IPv6-only LAN, which my brand-new printer didn't like very much.
And don't forget to backup your router and switch configurations!
They make industrial sd cards. I use those for many porjects that need a long life. However, this one has been running so long: Samsung EVO/EVO Plus, 32 GB, made 11/2018 used since 2019 daily. Wear has written roughly 1.5 TB in the 263 days, about 6 GB a day. So good enough for production for me!
I have used industrial cards. Are much better, yes. In a real production setup, where a stop costs actual money, or the thing is not easy to access, distance or dismantle a system, no way in hell I would use one.
I found a few wifi and bluetooth devices that were cheaper than or didn’t have a zigbee equivalent. Window fan and air monitor specifically.
I didn’t do an exhaustive market search, just sharing my anecdotal experience. It’s a great feeling to be able to focus on hardware capabilities and price, then will the software you envision into existence via AI accelerated RE.
Thread over Matter is not fully local. In Thread < 1.4, devices can connect to internet servers through NAT64. Most Thread Border Routers (AFAIR Home Assistant + OTBR is the main exception) do not allow you to disable NAT64.
In Thread >= 1.4, devices can request globally routable IPv6 addresses. The Thread 1.4 white paper does not hide the intentions:
The Thread 1.4 Border Router acts as the
IPv6 router between the Thread network and external networks,
including the user's home network and the larger internet. It is by default
configured to support external communication [...]
Cloud Connectivity: Direct internet connectivity enables Thread devices
to easily connect to cloud services. This facilitates remote monitoring,
control, and firmware updates, enhancing the user experience and
enabling remote device management capabilities. Device vendors can
now directly service and support their deployed fleet of devices,
regardless of which application ecosystem each device is integrated
with.
This is what you get when you let Google, Amazon, Apple, and Samsung design a smart home standard. They'll use every opportunity extract analytics and increase control. You'll own nothing and be happy.
Zigbee, Z-Wave, and ESPHome over WiFi/Thread are the only major modern local and privacy-preserving smart home standards.
Matter over WiFi is also not fully local for obvious reasons (though you can try to block them on the router/firewall and/or put them on a separate VLAN).
But matter is a better network. More efficient and faster. Large zigbee networks with cheap devices often result in ghost devices not responding. Matter even with the cheapest 2 dollar bulbs from AliExpress have zero issues.
That’s not the case at all. Matter over thread is newer. The latest range of ikea smart home stuff is all thread.
Thread has the advantage of consuming very low power so battery powered sensors and remotes can run for years. Wifi is only used because it doesn’t require buying a thread border router.
The other advantage of Thread is that the devices don't need to directly communicate with a border router (I think Zigbee is also like that?) unlike Matter over WiFi, where it does need to communicate directly with an access point.
> AFAIR Home Assistant + OTBR is the main exception
Luckily this is my setup.
I am not opposed to the possibility of cloud connectivity (as long as it is deactivateable) and even if I weren't I think I still prefer Matter (thread/wifi) over nearly all non-matter devices I've ever used, especially all the problems I had with Zigbee devices will make me stay far away from it. (Even though I am aware that many people had gread experiences with zigbee.)
Bunch of sensors, lights and switches (sonoff (though those were wifi, much easier but energy hungry), ikea, aqara and some more I don't remember). They were often annoying to pair, dropped from the mesh, or simply didn't like to work with my bridge (first ikea, later smart things and also tried skyconnect with HA directly).
Matter wasn't flawless, especially because I fell into the trap that was mentioned here before by trying to separate things into clans without fully understanding the needs of matter. But now that it is fixed, I had little problem with adding new devices (like not having a lamp in short range to the bridge lol) or them staying connected.
Interesting. I also have many IKEA and some Aqara Zigbee devices and they don't have any issues. Did you check if the Zigbee channel didn't overlap with 2.4GHz WiFi? (Same issues as with Thread) Did you use a multi-PAN (Zigbee + Thread) coordinator? I am pretty sure that some SmartThings hubs use multi-PAN.
I think this interpretation is a bit too negative. Undoubtedly the companies want to spy on you, and with Thread 1.3 they were already perfectly capable of doing so - via your Thread border router, which connects to the normal LAN and can talk to the Internet (if you let it). Allocating Thread devices from global IPv6 blocks (GUAs) doesn't change that, it just removes the NAT between Thread devices and the border router.
Before Thread 1.4 the IPv6 addresses used by the Thread network were all ULAs - i.e. the devices were only routable locally and didn't have a global IPv6 address, so if you did happen to want to talk to your Thread devices over the internet, it was possible but the Thread border router had to do NAT for you.
With Thread 1.4 the border router can assign device IPs straight from the global IPv6 block assigned by your ISP, which means you can directly access your Thread device over the internet without the NAT provided by the Thread border router. If you don't want to locally host your smart home automations (you and I do, but the average consumer doesn't) this can be very useful.
IMO this fits perfectly with the philosophy of IPv6 - every device can have a globally routable IP address without needing NAT.
If you don't want your devices reaching out to the Internet (like I don't), the solution is to put all that stuff on its own VLAN with no Internet access (which I do).
> If you don't want your devices reaching out to the Internet (like I don't), the solution is to put all that stuff on its own VLAN with no Internet access (which I do).
Or, if there is already a perfectly functioning open standard that just works and it's private by default (ZigBee) boycott the new standard created and forced upon us by BigTech and don't buy that hardware.
I think this interpretation is a bit too negative.
You wrote a long message to say that you agree with me technically, but then said that it has good uses as well. I don't disagree with that, but the problem with these companies is - you give them one finger and they take your whole hand. Besides that,
If you don't want to locally host your smart home automations (you and I do, but the average consumer doesn't) this can be very useful. It's also really helpful if you want to access your own devices remotely over the Internet and not need to bother with port forwarding.
this implies that this is not possible with Zigbee. You can use your devices remotely with Zigbee, but it is fully governed by your Zigbee hub/coordinator. You can use Home Assistant, the IKEA hub, the Samsung SmartThings hub, Homey, etc. to control your Zigbee devices remotely. The very big, but relevant difference is that the devices themselves cannot connect to the mothership to do analytics or to lock away functionality. There were already some Thread + Matter devices that required talking to the manufacturer's servers to get all functionality (I think e.g. some Nuki devices, but I might misremember).
If you don't want your devices reaching out to the Internet (like I don't), the solution is to put all that stuff on its own VLAN with no Internet access (which I do).
Putting Thread + Matter devices on a different VLAN than e.g. your phone which you are commissioning it with is very fragile. So fragile that most Matter documentation just says "your phone needs to be on the same WiFi network as your smarthome hub". It is possible (I had such a setup for a while), but requires that you e.g. do allow mDNS advertisements across VLANs (since Matter relies heavily on mDNS). Since you are talking about average consumers, Zigbee will give them privacy without any extra work, whereas with Thread + Matter they have to set up their network in a way that their ISP-provided router most likely doesn't even support.
Thread + Matter has no point outside very large deployments and giving a bunch of companies the opportunity to start data mining your sensor data as well.
> There were already some Thread + Matter devices that required talking to the manufacturer's servers to get all functionality
The most I have seen that is rather the functionality not being available in matter (ie the standard does not support it) so they are only available on zigbee with the manufacturers app.
In my personal experience
> Zigbee will give them privacy without any extra work
Was exactly as fragile and dependent on manufacturer infrastructure for some fratures as you describe MoT. (the privacy part is true though, and as mentioned all the time: I have heard others had much better experiences than me with zigbee)
The most I have seen that is rather the functionality not being available in matter (ie the standard does not support it) so they are only available on zigbee with the manufacturers app.
No, it's also used for remote access, see e.g. Nuki and Tado.
Not super-nefarious, but this is probably just the beginning. If companies can gather data they will and if manufacturers can add features that require subscriptions, they will. The line always has to go up and Thread offers them the means to do so. If you give the opportunity, this always happens (see e.g. TVs, 'owning the glass').
Was exactly as fragile and dependent on manufacturer infrastructure for some fratures as you describe MoT.
I have over 50 Zigbee devices and never have a single issue (using HA and zigbee2mqtt, some on a separate Hue network). The Matter over Thread devices that I have, regularly drop off the network (no interference, etc., they just still seem very buggy). This also makes complete sense, since Thread adds a lot of complexity (IPv6, TREL, communication between border routers, etc.).
I tried to hop on the Matter bandwagon until I found out that it's basically impossible to keep Matter over Wifi devices on their own IoT VLAN.
I have therefore come to the conclusion that there's literally only downsides to adopting them over Zigbee.
Looks like this was published recently.
When I did research about 6 months ago there wasn't really much information other than a vague "you need to do mdns broadcast and a bunch of firewall hackery" at which point I gave up as it was not worth the effort.
Still not sure if it's worth the effort but at least there's a guide :)
Similar, with z-wave instead of zigbee. Except after installing two smart switches, a motion detector, and two smart outlets, I realized there aren't really any automations that I actually want. My porch light turns on and off at sunset and sunrise, through an electronic rube goldberg machine that could have just been a single photocell lightbulb.
Not contradicting or criticizing, just describing my lack of creativity.
i do like zigbee (now Thread). however, i want to some fancy stuff sometimes that i could never do with those devices. Such as get the weather if its available. Updated sunset/sunrise times. Those are local services i use BUT i've seen it where they reach outbound to the internet when needed.
Probably, unless chinese companies change their business model and a secure successor to the ESP family arises. Low conviction loosely held opinion, shared bc you asked.
But why does that matter? My smart home upgrade cycle is around once every 5 years. The world is changing on a dime every 6 months nowadays. Projections beyond about a year out in the smart home market are useless to me as a consumer.
Things are moving fast now, security issues are now on the agenda with various governments, these ESP32's may be patched by Claude sooner than you think.
I pity those who've yet to experience running uncensored LLMs locally. Sure, they're not as fast as Claude, but they're still quite capable if you're patient.
It was helpful for reverse engineering some perfectly good software that the publisher recently discontinued and bricked for existing users. Sure, you could probably try to fool Claude or GPT into doing it, but I'd rather not play mind games with my LLM to get it to do something it thinks is illegal. I also just like brainstorming with uncensored Qwen because it doesn't constantly spit cautionary verbiage and disclaimers like frontier models do. The Splash version of regular old Qwen-2.6 also just did a bang-up job creating a simple daemon for managing my Govee smart outlets over BLE (without Home Assistant or other crap).
Presumably you're not concerned about claude deciding you're a bad actor and having your account closed? It seemed to go through a spat of doing that when people were doing the tv rooting recently to remove trackers/ads.
Like I said, DS 4.1 was on track to solve the same things while locally hosted. Using claude is a convenience. If anthropic really cracked down on RE, I’d move that workflow to open weight providers for a few dollars worth of DS and Kimi API costs, or continue to invest in my local setup.
The amount of RAM sitting in my house unused today is enourmous compared to the amount of RAM I had available 30 years ago. My 10 year old laptop alone has 32G. That's 1000 times as much RAM as I had in my entire house in 1996, and 200,000 times as much as 1986.
When I first bought memory I would be shocked if an edge device had 4MB of memory, just like I'd be shocked if it had 32G today.
We did have a period of centralised compute/memory/storage and dumb terminals, then this went to the edge, especially with PCs being smart end devices communicating over a dumb network. Then cloud became the fashion and everything centralised again. I hope the fashion goes the other way to more distributed systems, far more resilient.
The specifics of whether a local LLM runs in a lightbulb or a "home hub" is less interesting to me than the soverignty aspects. My fear is that we'll end up with the worst of both worlds, where we have to purchase, look after, and power, the LLM hardware, but it won't work without permission from "the cloud".
What use cases do you envision that go beyond the usual simple triggers and conditionals?
I could see using an LLM to set up more complex home automation rules that I am too lazy to set up myself. But for what kind of situations is it useful to have an LLM run the device directly?
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 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.
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.
fair. the qr codes and nfc tags are for guests and family who won't install an app. they sit next to the wall switches, not instead of them. I put them inside the switch plate sometimes to make it hidden. kind of a neat approach BUT it does require access to my wifi and usually guest permissions
How does the WiFi wall switch work? If WiFi goes down you can still click a light on and off? This is tempting as I have to leave all my smart lights on rn but do I really need to rip out all my light switches? Ugh.
Also anyone found the fix to smart bulbs constantly “updating”?
Yes it does, smart switches have buttons that control the relay directly.
Also aside from maintenance there's no reason your WiFi should go down, an inexpensive UPS can keep it up for hours in a power outage. Even if it can't access the outside world, local routing works fine.
Bought our home almost 20 years ago, and the former owner (apparently) owned a now-defunct home automation business. The upside was 100% X10/Insteon power and light controls, top alarm panel, environmental controls and so on running off of a wall-mounted "SmartLinc" touch panel. I expanded the system with Axis cameras and a Mac Server running Indigo and SecuritySpy (non-subscription), which enabled remote monitoring and control, but it all operates locally.
In a moment of weakness, I bought a D-Link wireless (wifi) water detector for the sump pit (it was on clearance for $30); it relied on a D-Link Home cloud service (free) that was closed down a few years ago.
Not wanting to toss it, I repurposed a spare Chromebook 14 that I was going to recycle and turned it into a Alpine monitor for the water detector, which I'm powering from the Chromebook's USB port (if power is lost in a rainstorm, I need to know about it before the sump pit fills up).
An immediate issue was that the D-Link firmware tried to "phone home" every 24 hours and would drop from the local network, since the server was kaput, and I had no clue what response it was expecting from the server to keep it alive. Solution; "reboot" the unit by interrupting USB power (with a Pico and MOSFET) every 23 hours. There's no question that home automation should work without the internet.
every single sentence was a level of jank that i could never imagine subjecting my parents to. could you code custom firmware for the dlink btw? your fix sounds like good ol' redneck engineering but for the smart home :p
The neat thing about X10 and Insteon is that everything acts "normally" at the device level, but you can program everything (remotely?) via the Indigo server. Your parents wouldn't even have to know that their house was running schedules (except for lights turning on an off).
If you're referring to the water sensor, it's definitely not parent-friendly, but I sometimes have to bail water during power outages from thunderstorms. It has to work during a power outage and an old Chromebook was cheaper than a UPS.
I kind of ran out of time for messing around with the sensor, but I did think of either dumping the Winbond chip on my reader, or using a .bin file that was used for OTA updates and pushing a modified version down locally.
That likely would have been "cleaner" vs. a 23-hour reboot cycle. I likely would have wanted a dump first in case I borked the firmware. The Python monitoring script just uses standard HNAP to read status codes from the sensor in 30 second intervals - I think there's some way to push updates via HNAP as well.
Thanks for the reminder - may be a good project for winter, when there's no chance of needing the sump.
Any sota llm of today can basically reverse engineer and fix any issue on such devices with a one-shot prompt. Recent emudhra CA leak makes interception traffic from most IoT devices trivial too
I haven't figured out how to recieve pings from my weather stone, I've always got to get up and look out the window, which, then I can just see the weather.
You don’t need to root your weather stone. You’ll want to root your weather tree. Rooted weather trees give you better long term data (if the tree is flourishing, the weather was nice all season). Unrooted weather trees just let you know if it was raining in the last couple days (still an upgrade over a weather stone, but not by much).
"...And so that's how I ended up with a $14,000 basement repair. Yup, those damn chromebooks are more waterproof than you'd think! I mean it makes sense, they're designed to be vomited on by children all day at school."
This sounds stupid but as a belt and suspenders approach I put a $65 2560x1440 resolution, H265 capable PoE IP camera on a piece of cheap cat5e utp literally aimed at my sump. Camera is joined to the same frigate server that runs the rest of the house security cameras including doorbell. If the sump alarm goes off while I'm away from the house for vacation or work, I can easily login and check if there's really a flood condition going on.
Frigate supports training states and classifications.
Perhaps you could train it on some of the images and set up notifications for a state you don't want?
For example, I have my doorbell camera trained for gate open/closed and car present/not present.
You select a few of the pre-captured images in the UI of gate open, gate closed, and then our car, no car, someone else's car, and it'll publish an event to MQTT or whatever (don't want to assume you use Home Assistant), then you can send out notifications on it.
Yes, if I wanted to put more work into it I could certainly train it on what it looks like with the sump mostly dry (the default state, 99.999% of the time) and then have it alert me if it diverges from that. Right now I keep an eye on weather alerts for $home and glance at the cameras at least once a day when I'm away, which is good enough to catch a failed sump pump and then dispatch professional repair services.
Every cloud service will eventually get shut down. The problem is that nobody wants to put the effort into reverse engineering the product to fake out the cloud service until after the cloud service is gone and the reverse engineering is much more difficult.
I have a Dell Chromebook that's been running (Debian) without issue (battery removed) in a similar capacity for over a decade, and four more that have been up for the last five years. The former I bought new, the latter are all second-hand survivors from some public school program. They're solid.
My personal preferred infra there would be for all devices to be meshed together, and a smart water shutoff valve able to cut off automatically if a water leak sensor trips.
I installed a water siphon backup for the sump pump. It creates a water siphon if the level gets above the electric sump and drains the pit. Is basic plumbing to install. Other than wasting a half gallon for every 20 gallons pumped out, it doesn't really on any technology nonsense.
Are you sure siphon is the right term? It sounds like a pumping scheme utilizing the venturi effect where a pressurized flow of water induces suction on a second pipe. Am I right?
And yet the words siphoning out the water is much easier to explain a simple concept to humans in a situation where the science doesn't actually matter. I'm not trying to be hateful, but HN slays me with this kind of needless pedantry sometimes.
It seems directly relevant in this case as you were detailing a technical solution to a problem in a thread about tradeoffs thus the implementation details were the entire point of the conversation.
It was confusing because you spoke of efficiency (doesn't apply to a siphon) and another commenter asked to where it drained, presumably because a sump will typically be well below grade meaning you can't siphon anything.
Your solution is a form of active pumping that doesn't require any electronics. A useful option and probably not something that will occur to most people to consider.
It's better than nothing sitting between you and knowing your cellar is flooding, and I could think of worse solutions than a device with a built-in UPS that's capable of powering a sensor.
It's a risk, not a guarantee, but there's nothing about macbooks that make them more susceptible to it than any other laptop with a li-ion battery. The batteries age. If they're in a machine that runs hot they age a bit faster (and my 2020 Intel MBP runs hot and is plugged in near 24/7), if they're charged to 100% all the time they age a bit faster (but battery management software like Apple has had for years nearly eliminates this risk factor).
There was, however, a batch of macbooks in 2019 that got recalled for a defect that increased the chance significantly, and perhaps that's what the parent poster has in mind.
I was referencing my own experience with 2010-2016 Intel Airs and Pros ending up in the Hazardous Materials bin, but good to know my current M1 Max will be up to the task whenever I can afford to upgrade.
What? What macbooks is this? My mac book air has been plugged in most of the time since 2013... I've replaced the battery once I think, but only because the battery life had decreased to 10 minutes.
my 2017 macbook pro got pretty bad spicy-pillowitis from being plugged in all the time. It got so bad that it wouldn’t sit flat on the desk, and I thought the trackpad was broken, but it was actually just unable to depress since it was right above the battery.
I just got rid of a 2013 MacBook Air earlier this year because I discovered two of the battery pockets inflated. But I never did a battery replacement, which I probably should have when I stopped using it as a laptop.
B-hyve has them. They talk to a hub over low power Bluetooth. The hub connects to my VPC-siloed WiFi AP on its own subnet with no routes to anything else in the house. But I still get phone alerts because it connects to its cloud service.
A leak sensor is the simplest home automation sensor possible. Its literally just a bare wire connected to ground, another connecting to a digital IO. You put the two bare wires on the floor a little distance apart and write some code to fire off an alert if the IO ever reports being connected, because if it is that means something conductive is sitting between the two wires.
especially if you keep your Chromebook in the cellar ;)
but isn't the alternative to have some third party on the Internet sitting between you and knowing that your cellar is flooding? third party services in the cloud have downtimes (not necessarily less than your Chromebook), tend to shut down entirely over time, tend to give up on you if you're not on a premium plan, tend to stop supporting the version of your device, etc
I recently had a similar issue. But instead I decided to go with industrial controls. I bought a Contactor with a NC and NO connections. I put a float switch in series with the activator coil circuit. So when the float triggers: NC->O and NO-C, which turns off the water and turns on the pump.
Contactors are built to run giant industrial motors for years, nevermind my little 1/4hp pump.
The way I built it lets my mother bypass it by unplugging the pump from the control box and plugging it right into the wall outlet.
So, hopefully it lasts years. If it doesnt, anyone can bypass it
Insteon remains incredibly solid. The new company is super slow with new product because it's tiny, but the hardware remains incredibly good long-term. It doesn't clutter common wireless bands and by nature of being stupid, Insteon devices cannot really be hacked.
My rules for home automation are largely the same as the article, first and foremost the existing physical UX of switches and stuff should always work, and ideally automations should be nearly invisible: They should subtly be convenient and monitoring should be useful, but nobody should be confronted with a weird change of lifestyle because a house has automation. And nobody should ever need an app to use their house.
They can also develop 1 to 1 relationships before a controller is even introduced. If I turn on my kitchen lights, the living room turns on to 50% without any controller, they talk amongst themselves over PowerLine and RF. Great 20 year old tech that's still relevant IMO.
Yeah app access can be handy, it just shouldn't be foundational to everyday use. If I want to change something remotely I wouldn't normally, apps are great. Each time my parents get their phone app out to turn on a light in the room they are in, I die a little inside.
> Solution; "reboot" the unit by interrupting USB power (with a Pico and MOSFET) every 23 hours
Reminds me of a possible apocryphal anecdote I heard about Alan Turing. He loved going on long bike rides, but one day his bike developed a distressing habit of slipping the chain. After the third time it happened, he realized it was due to a broken tooth on both the front and the back gear. He went back home and calculated how many revolutions of the pedals it would take for both broken teeth to line up when they were at the critical part of the rotation, and he built a little mechanical counter that would click on each link in the bike chain, and at the calculated time, a little arm would reach down and clasp the chain, holding it in place long enough for the teeth to re-engage.
He showed the device to his friend, who said "Why didn't you just buy new gears? You could have gotten both for a dime." And Alan replied "Oh, that's clever, I didn't think of that."
A similar anecdote about Turing appears in Neal Stephenson's "Cryptonomicon", except without the clever solution or the "absent-minded professor" stereotype.
A student came up to Von Neumann on campus and decided to prank the great scientist. "Two trains are on the same track 50 miles apart and they hurtle towards each other at 50 mph. As they start, a really fast fly on the nose of one train flies towards th eother at 60mph. When it reaches the other train it immediately flies back to the first, then back to the second, and so on until the fly is crushed in the crash. How far will the fly fly?
Von Neumann thought for a moment, and then replied, "30 miles!" The student grinned. "I thought you'd know this one! The trains take half an hour to hit each other, therefore the fly will fly 30 miles."
Von Neumann started. "I didn't think of that, I summed the series."
———
Source: "The Prisoner's Dilemma" by William Poundstone
A tennis match between two players, A and B, has finished. You are given the complete sequence of points played, in order. Each point is recorded as the letter of the player who won it.
Write a program that reads this sequence and works out which player won the match.
there are multiple ways to compute this but not obvious way is to look at the end of the list and see who scored last point.
Yes, and this (knowing a priori how tennis in particular is scored being useful for solving a problem) is the exact example that gets used as to why the SAT used to be somewhat biased in favor of people with the "right" background/upbringing/culture.
I'm building a tool that can monitor sump pits, more expensive, $99 early bird special, but supports local MCP and no hacking needed.
you can email me if you're interested, tyler@oculette.ai
Point your AI of choice at it for a bit, I wouldn't be surprised if you could write some custom firmware for it pretty simply. Or at least mock D-Link Home well enough to resolve the phone home problem.
Again, bulk of the blame goes to the techie hipsters who made coffee makers with cloud subscriptions and DRM popular. So did the BMW owners who paid $xx a month to get access to hardware THEY ALREADY OWNED.
"hey man, have u seen my coffee machine..it connects to the internet..so cool"
If you bought a GPU anytime in the last 20 years, the difference between a workstation GPU and a consumer GPU is only the driver. The hardware is the same with at most single contact changed so they can refuse to run the workstation driver on the consumer GPU. Similarly, many printer models use (used to use) the same head and the only difference in quality was the driver.
I vaguely remember this stuff and wondering at the time why I'm driving a BMW and have no subscriptions I'm aware of and the only one offered is the map of road conditions that relies upon all BMWs with radar mapping the roads and sending the data back to BMW servers you then load the map from. That seemed entirely reasonable as a subscription. I realized at some point this probably means I opted into paying more upfront at purchase to just have every feature turned on, so I never got offered a subscription after the fact to turn it on later.
I'm imagining now responses telling me I'm a slave to capitalism because I probably got upsold by ten grand or something I wasn't paying attention to when I signed everything. This recalls a discussion last night about grocery prices and everyone at the table talking about the comparison shopping they do and I blurted out I have no idea what anything costs. I just buy the same stuff every week and scan my card. I don't look at the stickers. I don't look at the receipt. I just look at whether I'm hitting my desired savings rate every year and remain on track for retirement goals, which I do, and that's all I actually care about, not whether I'm getting the absolute best deal possible on every transaction I ever make.
It brings my mind back to the article from the Atlantic posted here yesterday about anxiety rates going through the roof in spite of the modern western world being by far the safest it's ever been. It seems the difference now is people are so acutely aware of any and every risk, alerts to which are constantly being shoved in your face, that it overwhelms our ancestral ape brains as if dying with ten grand less in the bank is equal to having a tiger stalking you through the bush 10 meters to your right.
I'm tired. There's a podcast I listen to hosted by exercise scientists debunking stupid press headlines about the latest study or TikTok trend, with one of the hosts always pointing out how exhausted she is by all of these articles telling here she's doing something wrong. "Everything you knew about X is wrong." "You've been doing X wrong your entire life." It exhausts me too, Lauren. I bought the wrong car. I bought the wrong TV. I guess I should be living in fear that my appliances will stop working if I lose network. I should be worried that LG is telling every data broker in the world that I've spent the past year watching nothing but Buffy the Vampire Slayer and Sabrina the Teenage Witch. Shit, some agent probably just scraped Hacker News, correlated my user name with my real identity using stylometry, and now I'm going to be double-triple served nothing but ads for whatever consumer segment that puts me in. I'm going to die even poorer when some cleverly placed sponsored link in a search result makes me impulse buy some whimsigoth living room art I didn't really need.
The world of 30 years ago was a world in which six of my friends from school were killed, the National Guard occupied my city for half a week, 55 people were killed in a riot, a man dragged from his truck and beaten half to death on live television, a shootout with the LAPD that lasted three hours also live televised. We weren't allowed to wear red or blue for fear of kids getting shot by the crips and bloods. My middle school had three drive-bys while I was there. We regularly had recess and any other outdoor activities canceled because of dangerously unhealthy air quality, called "smog days." My little sister's best friend's mom lived under an assumed identity because she survived an attack by the nightstalker Richard Ramirez. My older sister's best friend was murdered in her sleep by her deadbeat mom's boyfriend.
But the world of 30 years ago still felt full of wonder and possibility. I saw a television and wanted to tear it apart to learn how it worked, to understand the miracle of how such a thing was possible. I see a television t...
I’m all about reducing e-waste but that sounds like insanity when you can get a sensor that’s better in every way for $25 without wasting electricity powering an entire laptop.
I have a few of Ikea's other Matter over Thread devices and have had spotty results with them so far using Google devices as thread border routers. Planning to move to a border router from SMLight which I hope will provide a more reliable connection.
I have none of that stuff. My home was purchased new with some 10baseT wired to 5 spots in the home and to every portal (window, door). we have a solar power system with enough battery backup watt hours for backup even if it rains for two weeks
But I DO have wildlife like bobcats and hares and raccoons that are autonomous and require remediation occasionally. Luckily, their activity is limited to the out of doors.
My water heater sits on top of a rubber pan and has a standalone battery operated water detector sitting on the edge. It will sound an alarm if water is detected. It's about $12 for two in Amazon. Low tech and it works.
My washing machine sits on the floor in the bathroom, but Danish apartment building regulations require there to be a floor drain (which is just the shower drain in this case) in case of a leak.
I don't know what a "Chromebook 14" is. Perhaps you mean HP's so-named 14-inch model.
In the year 2000, I lost all my possessions, especially the computers and storage media. Subsequently, I enjoyed NPR and Coast-to-Coast on a radio, including the Hurricane Katrina coverage. I had no idea what New Orleans looked like in 2005.
By that time, I was living across the street from a large thrift store. Eventually in 2006, I was weak-willed enough to begin building a home network again, and the thrift store provided two essential products: keyboards and WiFi routers.
The WiFi router was probably obsolete and EOL when I purchased it, but it definitely worked 100%, especially because I took care at the thrift store, to select a compatible AC adapter brick for it. It was a basic Netgear model, with 4 Ethernet ports, original firmware, probably 802.11n; even though an acquaintance said he would hack into it, he was inebriated most of the time, and I would never know if the router or network was hacked. You wouldn't know either, in 2026. Most consumers don't know, and could never find out, if their edge equipment is compromised. But anyway, even while running Ubuntu 6.04, I accepted the router's default firmware, and that Netgear thing lasted at least 5 years while I had an ADSL connection, and one and a half computers.
---
In the mid-1980s when I hung out at Radio Shack after church, I snapped up a lot of basic X10 kit. I collected the Dimmer Modules, the Appliance Modules, and later there was an RF remote unit available! The crown jewel for 12-year-old me was a central control unit that looked like a clock radio.
The control unit had cyan fluorescent display, simply the current time and alarm settings. And there were a bunch of control buttons where you could select an addressable Module, dim the lights, on/off, set timers, etc. And it was kind of amazing that everything plugged into mains outlets, and no other installation necessary!
My parents were constantly annoyed when I would sneak one of the modules onto a living-room floor lamp, or my favorite trick was cueing up a record on the turntable and then activating the Appliance Module to wake me up for school with Siouxsie and the Banshees or something. I was not allowed to persistently wire the entire house. I did not get into the journeyman electrician's job of replacing wall switches with X10 modules (those were slick). So, not even my shared bedroom was totally automated, but it was fun to experiment. It hopefully did not affect any neighbors. The pulse signals were said to stop at junction boxes, but there was still a House Code to set, in case all your flat neighbors wanted independent X10 networks.
I also never purchased the PC Interface that came with software. I believe I didn't yet have a compatible PC, anyway, and it seemed not in the spirit of a robust, predictable home automation. So yeah, my smart home devices worked without the Internet. How smart do you wanna get? My television set didn't spy on me, but everyone else does!
> I had no clue what response it was expecting from the server to keep it alive
That is the worst part: being explicitly designed to fail un-safe. You should be able to declare the unit unfit for the purpose it was sold for and get a full refund unless they open the specs so you can keep the device running as intended using your own resources.
I totally agree, and if you asked consumers, most would care and prefer for their smart home devices to communicate as little as possible with the outside world. And yet the market keeps offering invasive products, the discrepancy is frustrating.
There should be a system like the nutritional information on food and beverage products, as time went by most people read those and choose accordingly, which then incentivize companies to adapt their offering. We could imagine something similar which would define clearly the capabilities of the hardware or software sold, clearly labelled on the physical package or on the online shop with an easy classification for consumers to understand.
I disagree. If you asked any Random Customer they don't care if it's local first. They want it to work immediately with as little set up as possible and they want to access it from their phone.
If they could get away without even having a Wi-Fi router and just use their cellular plan they would. Why? Because there are plenty people who are happy with just whatever internet they get on their cell phone and don't want to spend extra for a router and an ISP
This is the real problem. The non-HN crowd just wants to turn on the tv for the dog from work and watch them. They don't really care about security because they assume it must be good.
They don’t know what “local first” means. But if you asked them “Would you like your lightbulbs to send data to a tech company or would you like them to not use the internet at all” I think you’d find all of them would prefer the no internet option.
With offline solutions like the ikea smart home stuff you can access it from your phone with even less setup than the wifi based options.
Unless it's law or AI becomes capable of rooting every device, it will never happen, most hardware gets out of the door with minimally functional software that only covers the immediate business need.
1. Buy zigbee/zwave
2. Prefer things you can flash tasmota if zigbee/zwave is not possible.
2. Maybe use eMudhra for mitm and flash the unsupported devices.
"eMudhra" lol, joke aside, i think there is value capturing leaving your network to see what it could be sending. i know its possible on some devices but can be pretty hard from my understanding.
everything I buy will work without internet for sure, and all of the above is much better with Home Assistant at the helm controlling things (only issue is the constant updates which you have to keep on top of)
If you accept and thus consented to a computer, or a device with a computer in it, without first verifying you can write to the media of its software, control the boot process, and will have sufficient access to that software, you get what you consented to.
Not sure why people are downvoting you, this is true.
My home automation rules are even simpler: home automation devices are firewalled off the Internet. If something does not work because of this, it gets returned, period. I do my research and only buy Zigbee or WiFi devices that can be hacked, so I haven't had problems so far.
ev chargers are the worst case of this imo. a lot of them only do the off-peak schedule through the vendor app, so if the cloud goes away you end up charging at peak. on some time-of-use plans that's ~$500 a year for the same car
Smart home devices should not freeze up whenever internet is down. A device like Google Nest should have the minimum decency to allow you to interact with it via local Wifi (which continues un-interrupted). It is fine if half the functionality doesn't work. It is fine if the "AI" doesn't work, but at least allow me to fcking change the temperature manually
Same thing with devices like Google Wifi which freezes up or run hot loops like headless chickens when internet is down.
If they can work without the internet, how TF do I scrape your data and sell it for a hefty profit? I made the equipment cheap so that everyone wants to buy it!
Oh I also have a under-the-table contract with some intelligence community for some backdoors. You know it's a lot more expensive for them to actually send someone to test physical security, than just pay a contractor to sit before the desk collecting data, right? RIGHT??
We've backed ourselves into a fundamentally silly situation: inventions intended to improve convenience and save time don't do either. They work "OK" then there's an update, and then you need to figure out what changed and how. And then it breaks, and you need to go troubleshoot what still works (if anything) and what no longer works. It's not more convenient, just more complex.
> They work "OK" then there's an update, and then you need to figure out what changed and how.
I don't allow my devices to update their firmware. If the device does not work out of the box, I return it, end of. My car has also been taken off the grid.
This is well put, and something I’ve been trying to put into words in regards to the trillion dollar AI industry. The convenience is very tempting, but the complexity doesn’t make it a reassuring dependency for decades into the future.
Do you have any favorites for single smart plug or multi switch smart plug strip? I used to be a heavy tp link kasa user since they have smart plug strips that worked locally (polling I think and not push) but I am now doing my best to move everything to plugs that don't even need setup through an app, just home assistant. Wish I had thought about this more when I first started with home automation but I think the migration will be worth it.
Look at Zooz, they have a variety of plugs and multi-port strips with control of individual outlets (both at the outlet and via zwave). They have options rated for being outdoors (holiday lights, etc), heavy duty ones for large motors, in wall switches and relays that go in the electrical box if you want something more permanent or invisible, etc. Solid support in home assistant. Push, 100% local, no shenanigans, no apps, direct to home assistant even for the updates.
I also have some eve energy plugs that do matter/thread (also push/local). There are apps here but i generally haven't needed them once i got things set up in home assistant.
I have a ton of Kasa. Local polling, but since it's more for on/off, I don't care about a 5s poll. I had to buy about ~30 light switches for my house, and using Kasa saved thousands over Lutron.
Smart Plugs, sensors, etc: Ikea is amazing. You need to buy a hub, but then accessories are 3x cheaper. Their plugs are $9 and have energy monitoring which is amazing. They have water leak sensors for $12. Incredible value.
Except VARMBLIXT. I can't connect it to HomeAssistant no matter how I much try. It sort of begins discovery, but then can't connect to it. Could be some weird networking issue, but all IKEA lamps connect without any issue.
IKEA is a bit hit and miss. Their light bulbs are great for the price. Their power plugs also work pretty well, but at some point I started shaking them before use because two of them had pretty large solder balls that would have shorted things out:
Most of their sensors are meh. Their PIR sensors are known to be mediocre compared to the competition. Their air quality sensor is ok to get trends, but e.g. the CO2 measurements are quite a bit off compared to quality sensors and are way off if you live at a higher altitude (due to using a thermal conductivity sensor and not NDIR like better sensors).
Also, the IKEA smarthome app (used in conjunction with the Hub) uses Play Integrity and therefore blocks e.g. GrapheneOS. But you are better off with Home Assistant and their ZBT-2 antenna (can be flashed for Zigbee or Thread) anyway.
> Their plugs are $9 and have energy monitoring which is amazing.
That's 2nd generation, first one was just a basic on/off smar plug. I still got these and so far I had no issues even after Dirigera hub got that firmware update which enabled Matter. Ikea 'gear' overall work much better than Meross one I had before.
Switchbot smart plugs around$5 in a four pack and you can flash them over the air to ESP home or tasmota so they run completely local if you want a Wi Fi option
Esp32c3
For single plugs, IKEA INSPELNING are really great if you can still get them (I stocked a bunch of them before they sold out here). Zigbee, affordable and very reliable.
GRILLPLATS is the successor. I have one, but haven’t tried it much yet because it is Matter over Thread. They have a hidden Zigbee mode, but it doesn’t expose power use:
But if Matter over Thread fits into your system, it’s an affordable option.
For power strips, we are now using the Gen4 Shelly Power Strip. It supports pretty much anything: its own API over WiFi that is supported by HA, Shelly cloud, Matter over WiFi, Zigbee, and Bluetooth. It can even work as a WiFi range extender and as a Bluetooth proxy for HA. Initial setup is by connecting to it as a WiFi AP and the connection to the built-in web interface. There were some issues with Zigbee flooding, but I think that was improved with firmware updates.
"Anything that can run Tasmota" for me. See here: https://templates.blakadder.com/plug.html Straightforward web interface, scriptable, native HA, supports every protocol under the sun. Works fine here.
Currently using a bunch of Nous A8T (EU plug) because they come with Tasmota ootb.
I fully stick to ZigBee and ZWave devices wherever possible, the reliability is just much better in my experience + they don't clog up my wifi. Bonus aspect of replacing all my lightswitches with ZigBee smart switches is they all act as mesh routers for the network, so signal is good through the entire house.
FAFAIK neither Zigbee nor ZWave uses IP networking in their protocol stack which makes them infinitely better than Matter, especially given the context of this article.
Matter-over-Thread also keeps devices away from the Internet if you run your own OTBR (e.g. via Home Assistant). With off-the-shelf border routers you can't be sure.
This is true theoretically, but in practice, out of the tens of smart devices i've set up, the Zigbee devices have always been more reliable than the Matter over Thread ones. Also easier to set up. Sure it's an older standard but it seems like its simplicity makes it work much better.
Matter over Thread is a bit more complicated just because there are typically two parts: a Thread border router and a Matter server, compared to ZigBee where you might just use something like Zigbee2MQTT which does both functions.
That's also part of it's advantage though: I have a Mikrotik router which has a Thread radio, so I run the router software on that, and Matter server on another machine - so no USB dongle needed in my server rack where the signal would be poor.
Also Matter over Thread supports multiple Thread routers, multiple Matter servers, and devices being paired to multiples of each - maybe that doesn't help with the complicated part, but it opens it up for commercial deployments as you can have redundancy. I would not be surprised if it replaces DALI in some instances in the not too distant future.
There are still some teething issues, which is because it's still kind of early in its adoption (ZigBee felt the same before Zigbee2MQTT). Personally I feel it's the way everything will go simply because it is more interoperable. Already today you can get routers, TVs and fridges (Samsung) with Thread hardware, so it's going to become mainstream as regular people will be able to just buy a smart bulb, pair it, and control it with their phone without buying or setting up anything else.
IKEA released a bunch of new smart home devices that support Matter over Thread and ZigBee this year if you want to try it out.
Yeah, I do have the IKEA Matter-over-Thread air sensor. Whenever the electricity goes down, it refuses to come back up until I restart it multiple times. I never saw this issue with any Zigbee device. I also had some dimmers that operated both on MoT and Zigbee, and ended up switching them all to Zigbee to resolve this exact issue.
I get that MoT has more features, but it seems like it makes the whole thing less reliable. I also hate the pairing process with Matter-over-Thread. Enable IPv6 in your home network, scan the QR code with your phone... What if the QR sticker falls in a few years? With Zigbee I just press the reset button and I see it directly in the HA web interface. No need for a phone, no need for camera, and doesn't touch your router. Less features, but my devices have been working for years without a single failure.
I have had the exact same experience. Zigbee has been working for years (in HomeAssistant) without any problems.
With Matter/Thread though, i frequently run into issues. Particularly these were power switches from Ikea by the way, but i don't thing that's relevant.
Also the range of Zigbee devices is much better / bigger, since each device "expands" the network range so to speak.
Matter over Thread is such an insane protocol. Devices need to have an additional bluetooth radio on top of the Thread IEEE 802.15.4 radio, just for the initial pairing, then it's never used again.
I'm disappointed that Ikea decided to switch their smart home gadgets from using Zigbee to Matter over Thread.
IIRC they're still offering Zigbee that you can switch to via a manual switching sequence, though some people say it still runs a little worse than the older Zigbee line(s).
As far as I know, pretty much all commercially available Thread SoCs also support Bluetooth LE. They both operate in the 2.4 GHz band, so supporting both protocols isn't that hard.
I thought so too until I found out Thread currently (AFAIK) doesn't support multicast requests while Zigbee does, which can be hugely important for smart lighting with 30+ bulbs. Until that gets added to the spec (and supported) you might be better off putting the lighting part of the setup on Zigbee.
> Group messages are also important as they allow simultaneous control of several Matter Nodes through Multicast. In order to route this traffic into the Thread network, both Matter and Thread implement the Unicast Prefix-based IPv6 Multicast Addressing Scheme defined by RFC 3306.
> Thread defines a series of features to support multicast across a heterogeneous network (Thread and Wi-Fi/Ethernet Network segments) for multicast addresses with scope larger than realm local.
I got the phrasing a bit wrong, but it's Matter over Thread that doesn't support multicast on an application level. It's part of the spec that just shipped (1.6.1) but no controllers actually support it right now. It's about to ship with HA:
Zigbee: an open spec, no licensing fees, cheaper devices. Can be locked down by OEMs to their hub (which can then demand an internet connection). In recent memory Hue attempted this, and only walked back after huge backlash. I couldn't find a thermostat back when I did my home without hub lock.
ZWave: the consortium demands interoperability. Testing and licensing fees means more expensive devices. Things will work together, though sometimes you will need to figure out what parameters actually do.
Use both. Home assistant is perfectly happy with two dongles plugged in.
In practice the answer ends up being both. There are some devices that only exist in one protocol (sometimes for technical reasons) and as a sibling noted, the antennas are cheap and home assistant exists.
ZWave stuff tends to be rock solid (there is expensive certification involved) but that means the devices are more expensive and the selection can be limited because the volumes aren't always there to make sense to get certified.
The only real problem I've found with the "all of the above" situation is that device bindings only work within one protocol. If you want to bind a lightbulb to a wall switch such that they don't require a trip through the hub they need to both be zigbee, zwave, or matter. This is unfortunate because there are a lot of really nice ZWave wall switches but not a lot of ZWave lightbulbs.
Zigbee is definitely at 2.4 GHz. Zwave is 900 MHz only IIRC. I can't recall if Zigbee is at 900 MHz although some form of it (not called Zigbee) likely is.
Devices using that band are like devices that don't have USB-C. There are plenty of useful ones in my home, but they are targets for eventual replacement and any new product still using the older standard is should not be brought into my home if it can be avoided at all. A micro USB port on a new device, is a reason to return the product
I've gone one level further and only use open source smart devices (excluding the big ticket items such as the robot vacuum or air purifier, where it's often impossible to find a 'Local' option or the difference in quality of its main purpose is just not negotiable...).
ESPHome or Tasmota work really well for all kinds of things.
The last one I replaced was an infrared blaster for controlling my AC. The previous one, even though it was in an IoT VLAN decided to spam the network with requests to its mothership (never succeeding, fortunately).
440 comments
[ 771 ms ] story [ 404 ms ] threadI guess connecting to home automation would be useful for a fridge that's sufficiently out of the way that you might not hear it. But for the fridge in my kitchen the noises emitted by the fridge itself are enough for me
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.
Unfortunately, almost no lan connected devices sport such an interface.
With another project, based on the ESP32-S3, I was able to cram grafana style scrubbable charts, full activity log, refrigerant circuit diagram, equipment topology, etc. I'm pretty excited for the ESP32-S31 since it's a nice jump up in performance.
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.
This is one of the reasons I love Zigbee, the device network is entirely independent of my computer network, can not directly communicate to the internet in any way, and every device I've put on the network so far can be configured to directly communicate with other devices so basic functionality remains operational even if the controller goes down.
For example, in my living room I have a Lutron Aurora dimmer dial locking a standard switch in the on position, then plugged in to the outlet that switch would normally control is a Leviton dimmer module. Both connect to the network controlled from a PoE Zigbee dongle, but have been directly paired as well so pressing the button on the dial makes it directly tell the dimmer module to turn on, rotation commands are sent direct, etc.
The controller also receives these commands when it's active and can add to the functionality for inputs like double-click, but if it's down the basic functionality still works entirely as a normal person would expect. You walk in to the room and see a knob on a switch plate, you turn the knob and the light brightness increases regardless of the state of the internet, my home server, or even my network as a whole.
The other major advantage compared to Matter/Thread is that it's not IP based so it's impossible for any Zigbee device to ever have a cloud dependency. No matter how much some marketing dingus might want to, it can not be ruined.
I'm a strong proponent of everything-over-IP in an ideal world, and love the fact that Matter mandates IPv6 which forced a lot of embedded vendors to finally support it, but the IoT vendors have demonstrated that if they're given an internet connection they WILL abuse it to lock you in to their platform.
Matter over Thread is nice because it functionally provides an "IoT VLAN" by default and can only reach the internet or even the LAN through a gateway, which can be fully controlled by the user making it the best of both worlds.
As for Matter over WiFi, for the reason in the first paragraph I've maintained a rule of not buying WiFi IoT devices unless the device has a real need for direct access to the internet or high-speed data from the LAN to do what I want it to do, and Matter doesn't change that. It is nice though that hypothetically I could have a Matter-over-Thread switch cause a WiFi device to do something directly without needing to go through Home Assistant.
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.
__
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.
Regulation, taxes, law. Can also just take over their business. Oh and they have military.
The state can do literally anything it wants.
It just often shouldn't, because a lot of escalative actions have quite the collateral.
But you don't need any of those to steer here.
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
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.
In this case the article starts with "AI;DR: I wrote the first draft, AI helped me edit and rewrite it, and I made the final pass. The rules, setups, and talks are my own" so none of that applies.
I will not be happy because we will be lost as a culture when we give up on this.
By all means, pass it through an LLM for review if you really want, but you don't need to fix all the issues it finds, and you shouldn't let it remove the personality from your text by rewriting things. Because they will eagerly do that.
I would much rather read a post with a human voice - don't be afraid to just do it :)
So you are just signaling that you aren't reading specific types of LLM language that you can detect, but not all of them.
Its like you went to a potluck and brought a salad with you, you just closed your eyes and threw some things from the fridge together without thinking, it's all mixed together now and you don't know if something was old/moldy.
You are proud and happy that you brought a contribution, and you feel like your responsibility/duty ends there while the people who want to eat it have to be careful and fish things out, or just avoid it because they can't be sure if it was old or not.
Natural language is a non-verifiable domain due to the ambiguity of the intent, so we will never get away from this issue in LLM-land. You can't know intent of the writer unless they wrote it themselves.
I think it interesting that people are more upset about the complaint than the cause of the complaint.
If this place — a place where people are supposed to be sharing interesting things for techies and geeks — gives up on complaining about thoughtless nonsense, its culture dies.
I used to complain when web sites had huge whitespace margins on the left and right side of the text. I hate that! But most web sites do it, and complaining about it here accomplishes nothing and adds nothing to the conversation.
How do you envision that pushback if it's not people complaining about it?
Or, of course, make the human texts excellent blog posts that fit the readers' area of expertise and there you go, evidence! People hate AI writing! The HN hivemind was right all along!
Tell people it's CGI, they won't be enthusiastic about it. But tell them it's "practical effects" and they will absolutely love it. Even when the movie has even more CGI than average... they will still praise it, and will even praise the CGI scenes they failed to identify as computer generated.
I'd like to see this evidence. I have very limited time and attention. LLMs are usually used to take the user's prompt and flower it up to 10x to 100x fluff. Just give me the prompt.
- 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.
But yeah, true.
RPi hardware is nice because it's a fairly-consistent, known quantity. All copies of a given board have the ~same features and bugs, so an image made for someone else's Pi 4 will run on my own Pi 4. This grants appliance-like utility that we haven't really ever had much of in the x86 PC space.
Pi hardware almost never shows up as cheap as a person might wish for. It's scarce on the used market, even when prices are high. In that space, demand outstrips supply.
And it's never been "cheap". A decade and a half ago when it first showed up, folks were already comparing the price (plus the SD card, and the appropriate power brick, and the USB cable, and the wifi adapter, and some kind of housing) to that of a used PC or laptop, and the Pi seldom came out on top.
But if we're doing new vs new, then it does alright price-wise.
And it's tiny. It's got some hackability built in, with the not-USB camera interface and the not-HDMI display support, and some GPIO that can be useful for some stuff.
It's got places where it's useful. I use them in those places. (And despite that lengthy defense, I run Home Assistant on a cheap SFF Lenovo box that came to me with RAM and an SSD for $50, delivered, in pre-used form)
Examples:
- SD cards are very convenient for hobbyists "experimenting" with OS variants. They are not great for performance or reliability; a soldered on storage device is common in many other products. - The price is more than it seems - you need a case and a power supply and an SD card. - The case situation is bad. If you want to 'set and forget' you really want a robust case, potted. Many N100's are designed for more abusive environments. - Most use cases don't need the HDMI, Camera, Audio, Ethernet, or all the USB ports. That is board space, purchase cost, and idle power draw that can be reduced or dropped. - It doesn't have some modern connectors, e.g. M2, because there is no space on the board, and the PCIe lanes have gone to the connectors you probably don't use.
Many stick based computers, particularly second hand and older generations, are better for most use cases. These are typically Intel N series; but are increasingly ARM, particularly at the lowest prices.
I paid ~$150 for a fanless 4-port router with 16GB of RAM a few years back. Now they are ~$400 and up.
Your router is a point of failure, so an internet outage making it go weird doesn't mean you want all your stuff becoming on manageable).
IoT is honestly one of the best ways to rapidly learn SRE concepts.
On a home network with dynamic addresses: No need to worry about IP addresses; no overlaps happen, no mistakes, no spreadsheet table to manage. Use names (as long as DHCP works, then so does DNS). Renumbering the network is simple when that's useful.
It still works when the internet is down. And when the router itself breaks, we fix it or replace it and things work again. It's good to have a functional spare.
---
On a home network with static addresses: Micromanage IP addresses, keep a spreadsheet table, potentially make mistakes. Avoid using names (DHCP may be eliminated but DNS is still a failure point), making things harder to remember six months from now. Renumbering is arduous.
It still works when the router is down, but: When the router breaks, we still fix or replace it anyway. It's good to have a functional spare.
---
In terms of SRE: I don't need 5 9's of reliability for my smart light bulbs' IoT controls at home. They all still have some manner of physical switch, and they all can still be toggled on and off the old-fashioned way. Parallel controls work well-enough.
Everyone will get by almost-normally while I fix the router, which is a task that I'll set forth to be doing anyway for all of the reasons that household internet access provides.
The concept is that, instead of assuming that IP and MAC addresses are at all reliable, to "fingerprint" devices on the LAN based on things such as SSH public keys, TLS SPKI fingerprints, UPnP device UUIDs, mDNS, common TCP port profiles, and IEEE OUI manufacturer in order to form mostly-stable identities. The IP and full MAC are only fallen back to when there's absolutely no other data to work off of. My software uses a daemon that scans the network in order find devices and update existing device identities if IPs change. It serves hostnames for these devices on the `.home.arpa` TLD and creates synthetic IPv6 addresses for each device in case one would rather use their own DNS server or hosts file.
I recently "finished" this project, and it's kind of awesome because I was able to swap to a new router and everything just kept working as-is without having to think about it. Of course mDNS solves a lot of these problems already, but not every device has mDNS.
What's really cool is I recently realized that this is also very useful for accessing your local network with a VPN like Wireguard. Even though I only planned on this being something you run on individual clients, I put it on my OpenWrt router and configured it with dnsmasq so that, when I connect to my network with Netbird, all of my local hostnames just work on my phone by virtue of connecting to the network and using its DNS server. That said, my project doesn't require router admin access at all. It can even run as rootless if TUN access (for synthetic addresses) is turned off.
Maybe I'll release it publicly at some point, but if anyone wants to steal this idea and do their own implementation, I am not territorial over it. It's made getting my LAN devices working together a joy.
And don't forget to backup your router and switch configurations!
1. buy the Zigbee version
2. boom, local only smart home is done
I guess if there's a device that's actually wi-fi only then I'll have to go through your flow but so far there's been no need.
I found a few wifi and bluetooth devices that were cheaper than or didn’t have a zigbee equivalent. Window fan and air monitor specifically.
I didn’t do an exhaustive market search, just sharing my anecdotal experience. It’s a great feeling to be able to focus on hardware capabilities and price, then will the software you envision into existence via AI accelerated RE.
In Thread >= 1.4, devices can request globally routable IPv6 addresses. The Thread 1.4 white paper does not hide the intentions:
The Thread 1.4 Border Router acts as the IPv6 router between the Thread network and external networks, including the user's home network and the larger internet. It is by default configured to support external communication [...]
Cloud Connectivity: Direct internet connectivity enables Thread devices to easily connect to cloud services. This facilitates remote monitoring, control, and firmware updates, enhancing the user experience and enabling remote device management capabilities. Device vendors can now directly service and support their deployed fleet of devices, regardless of which application ecosystem each device is integrated with.
https://threadgroup.org/Portals/0/Documents/Thread_1.4_Featu...
This is what you get when you let Google, Amazon, Apple, and Samsung design a smart home standard. They'll use every opportunity extract analytics and increase control. You'll own nothing and be happy.
Zigbee, Z-Wave, and ESPHome over WiFi/Thread are the only major modern local and privacy-preserving smart home standards.
Matter over WiFi is also not fully local for obvious reasons (though you can try to block them on the router/firewall and/or put them on a separate VLAN).
Thread has the advantage of consuming very low power so battery powered sensors and remotes can run for years. Wifi is only used because it doesn’t require buying a thread border router.
Luckily this is my setup.
I am not opposed to the possibility of cloud connectivity (as long as it is deactivateable) and even if I weren't I think I still prefer Matter (thread/wifi) over nearly all non-matter devices I've ever used, especially all the problems I had with Zigbee devices will make me stay far away from it. (Even though I am aware that many people had gread experiences with zigbee.)
In contrast, Matter + Thread devices in our network drop out very frequently (and no, I am not using multi-PAN, etc.).
Matter + WiFi devices are probably much more reliable (if close enough to an AP), but a security and privacy nightmare.
Matter wasn't flawless, especially because I fell into the trap that was mentioned here before by trying to separate things into clans without fully understanding the needs of matter. But now that it is fixed, I had little problem with adding new devices (like not having a lamp in short range to the bridge lol) or them staying connected.
Before Thread 1.4 the IPv6 addresses used by the Thread network were all ULAs - i.e. the devices were only routable locally and didn't have a global IPv6 address, so if you did happen to want to talk to your Thread devices over the internet, it was possible but the Thread border router had to do NAT for you.
With Thread 1.4 the border router can assign device IPs straight from the global IPv6 block assigned by your ISP, which means you can directly access your Thread device over the internet without the NAT provided by the Thread border router. If you don't want to locally host your smart home automations (you and I do, but the average consumer doesn't) this can be very useful.
IMO this fits perfectly with the philosophy of IPv6 - every device can have a globally routable IP address without needing NAT.
If you don't want your devices reaching out to the Internet (like I don't), the solution is to put all that stuff on its own VLAN with no Internet access (which I do).
Or, if there is already a perfectly functioning open standard that just works and it's private by default (ZigBee) boycott the new standard created and forced upon us by BigTech and don't buy that hardware.
You wrote a long message to say that you agree with me technically, but then said that it has good uses as well. I don't disagree with that, but the problem with these companies is - you give them one finger and they take your whole hand. Besides that,
If you don't want to locally host your smart home automations (you and I do, but the average consumer doesn't) this can be very useful. It's also really helpful if you want to access your own devices remotely over the Internet and not need to bother with port forwarding.
this implies that this is not possible with Zigbee. You can use your devices remotely with Zigbee, but it is fully governed by your Zigbee hub/coordinator. You can use Home Assistant, the IKEA hub, the Samsung SmartThings hub, Homey, etc. to control your Zigbee devices remotely. The very big, but relevant difference is that the devices themselves cannot connect to the mothership to do analytics or to lock away functionality. There were already some Thread + Matter devices that required talking to the manufacturer's servers to get all functionality (I think e.g. some Nuki devices, but I might misremember).
If you don't want your devices reaching out to the Internet (like I don't), the solution is to put all that stuff on its own VLAN with no Internet access (which I do).
Putting Thread + Matter devices on a different VLAN than e.g. your phone which you are commissioning it with is very fragile. So fragile that most Matter documentation just says "your phone needs to be on the same WiFi network as your smarthome hub". It is possible (I had such a setup for a while), but requires that you e.g. do allow mDNS advertisements across VLANs (since Matter relies heavily on mDNS). Since you are talking about average consumers, Zigbee will give them privacy without any extra work, whereas with Thread + Matter they have to set up their network in a way that their ISP-provided router most likely doesn't even support.
Thread + Matter has no point outside very large deployments and giving a bunch of companies the opportunity to start data mining your sensor data as well.
The most I have seen that is rather the functionality not being available in matter (ie the standard does not support it) so they are only available on zigbee with the manufacturers app.
In my personal experience
> Zigbee will give them privacy without any extra work
Was exactly as fragile and dependent on manufacturer infrastructure for some fratures as you describe MoT. (the privacy part is true though, and as mentioned all the time: I have heard others had much better experiences than me with zigbee)
No, it's also used for remote access, see e.g. Nuki and Tado.
Not super-nefarious, but this is probably just the beginning. If companies can gather data they will and if manufacturers can add features that require subscriptions, they will. The line always has to go up and Thread offers them the means to do so. If you give the opportunity, this always happens (see e.g. TVs, 'owning the glass').
Was exactly as fragile and dependent on manufacturer infrastructure for some fratures as you describe MoT.
I have over 50 Zigbee devices and never have a single issue (using HA and zigbee2mqtt, some on a separate Hue network). The Matter over Thread devices that I have, regularly drop off the network (no interference, etc., they just still seem very buggy). This also makes complete sense, since Thread adds a lot of complexity (IPv6, TREL, communication between border routers, etc.).
Your post reinforces that feeling.
I used this guide if you're interested - https://ndo.dev/blog/hassio-thread-setup/, it's been working without fuss.
Looks like this was published recently. When I did research about 6 months ago there wasn't really much information other than a vague "you need to do mdns broadcast and a bunch of firewall hackery" at which point I gave up as it was not worth the effort.
Still not sure if it's worth the effort but at least there's a guide :)
Not contradicting or criticizing, just describing my lack of creativity.
Custom firmware?
These are cheap chinese adware driven sensors. Security is not their priority.
But why does that matter? My smart home upgrade cycle is around once every 5 years. The world is changing on a dime every 6 months nowadays. Projections beyond about a year out in the smart home market are useless to me as a consumer.
The future of smart devices is a local LLM running on every single one.
We did have a period of centralised compute/memory/storage and dumb terminals, then this went to the edge, especially with PCs being smart end devices communicating over a dumb network. Then cloud became the fashion and everything centralised again. I hope the fashion goes the other way to more distributed systems, far more resilient.
The specifics of whether a local LLM runs in a lightbulb or a "home hub" is less interesting to me than the soverignty aspects. My fear is that we'll end up with the worst of both worlds, where we have to purchase, look after, and power, the LLM hardware, but it won't work without permission from "the cloud".
I could see using an LLM to set up more complex home automation rules that I am too lazy to set up myself. But for what kind of situations is it useful to have an LLM run the device directly?
Also, QR codes as a good UX? I like what you are smoking, it is very effective in disconnecting from reality.
Also anyone found the fix to smart bulbs constantly “updating”?
Also aside from maintenance there's no reason your WiFi should go down, an inexpensive UPS can keep it up for hours in a power outage. Even if it can't access the outside world, local routing works fine.
In a moment of weakness, I bought a D-Link wireless (wifi) water detector for the sump pit (it was on clearance for $30); it relied on a D-Link Home cloud service (free) that was closed down a few years ago.
Not wanting to toss it, I repurposed a spare Chromebook 14 that I was going to recycle and turned it into a Alpine monitor for the water detector, which I'm powering from the Chromebook's USB port (if power is lost in a rainstorm, I need to know about it before the sump pit fills up).
An immediate issue was that the D-Link firmware tried to "phone home" every 24 hours and would drop from the local network, since the server was kaput, and I had no clue what response it was expecting from the server to keep it alive. Solution; "reboot" the unit by interrupting USB power (with a Pico and MOSFET) every 23 hours. There's no question that home automation should work without the internet.
If you're referring to the water sensor, it's definitely not parent-friendly, but I sometimes have to bail water during power outages from thunderstorms. It has to work during a power outage and an old Chromebook was cheaper than a UPS.
I kind of ran out of time for messing around with the sensor, but I did think of either dumping the Winbond chip on my reader, or using a .bin file that was used for OTA updates and pushing a modified version down locally.
That likely would have been "cleaner" vs. a 23-hour reboot cycle. I likely would have wanted a dump first in case I borked the firmware. The Python monitoring script just uses standard HNAP to read status codes from the sensor in 30 second intervals - I think there's some way to push updates via HNAP as well.
Thanks for the reminder - may be a good project for winter, when there's no chance of needing the sump.
Anyone made any progress on rooting it?
You can then flash new firmware over USB.
Perhaps you could train it on some of the images and set up notifications for a state you don't want?
For example, I have my doorbell camera trained for gate open/closed and car present/not present.
You select a few of the pre-captured images in the UI of gate open, gate closed, and then our car, no car, someone else's car, and it'll publish an event to MQTT or whatever (don't want to assume you use Home Assistant), then you can send out notifications on it.
And yet the words siphoning out the water is much easier to explain a simple concept to humans in a situation where the science doesn't actually matter. I'm not trying to be hateful, but HN slays me with this kind of needless pedantry sometimes.
It was confusing because you spoke of efficiency (doesn't apply to a siphon) and another commenter asked to where it drained, presumably because a sump will typically be well below grade meaning you can't siphon anything.
Your solution is a form of active pumping that doesn't require any electronics. A useful option and probably not something that will occur to most people to consider.
No issues whatsoever so far, and it holds the battery charge totally fine (which I get to test 2-3 times a year when i travel with it).
There was, however, a batch of macbooks in 2019 that got recalled for a defect that increased the chance significantly, and perhaps that's what the parent poster has in mind.
Good luck hoping it will be existing in a few years. No service - no alerts. The whole thing becomes trash then.
but isn't the alternative to have some third party on the Internet sitting between you and knowing that your cellar is flooding? third party services in the cloud have downtimes (not necessarily less than your Chromebook), tend to shut down entirely over time, tend to give up on you if you're not on a premium plan, tend to stop supporting the version of your device, etc
Contactors are built to run giant industrial motors for years, nevermind my little 1/4hp pump.
The way I built it lets my mother bypass it by unplugging the pump from the control box and plugging it right into the wall outlet.
So, hopefully it lasts years. If it doesnt, anyone can bypass it
I need an alarm because I've had a few pumps wear out, and also because I've had the power go out long enough to flood.
Your solution automates something that shouldn't have a computer involved in automating.
My rules for home automation are largely the same as the article, first and foremost the existing physical UX of switches and stuff should always work, and ideally automations should be nearly invisible: They should subtly be convenient and monitoring should be useful, but nobody should be confronted with a weird change of lifestyle because a house has automation. And nobody should ever need an app to use their house.
Reminds me of a possible apocryphal anecdote I heard about Alan Turing. He loved going on long bike rides, but one day his bike developed a distressing habit of slipping the chain. After the third time it happened, he realized it was due to a broken tooth on both the front and the back gear. He went back home and calculated how many revolutions of the pedals it would take for both broken teeth to line up when they were at the critical part of the rotation, and he built a little mechanical counter that would click on each link in the bike chain, and at the calculated time, a little arm would reach down and clasp the chain, holding it in place long enough for the teeth to re-engage.
He showed the device to his friend, who said "Why didn't you just buy new gears? You could have gotten both for a dime." And Alan replied "Oh, that's clever, I didn't think of that."
A student came up to Von Neumann on campus and decided to prank the great scientist. "Two trains are on the same track 50 miles apart and they hurtle towards each other at 50 mph. As they start, a really fast fly on the nose of one train flies towards th eother at 60mph. When it reaches the other train it immediately flies back to the first, then back to the second, and so on until the fly is crushed in the crash. How far will the fly fly?
Von Neumann thought for a moment, and then replied, "30 miles!" The student grinned. "I thought you'd know this one! The trains take half an hour to hit each other, therefore the fly will fly 30 miles."
Von Neumann started. "I didn't think of that, I summed the series."
———
Source: "The Prisoner's Dilemma" by William Poundstone
https://archive.org/details/prisonersdilemma0000poun
Also one train was a boy coming home (stative) and the dog was running
Write a program that reads this sequence and works out which player won the match.
there are multiple ways to compute this but not obvious way is to look at the end of the list and see who scored last point.
https://www.thedrive.com/news/bmw-commits-to-subscriptions-e...
https://www.tastingtable.com/2264769/customer-tricks-keurig-...
Explanation: https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/...
I'm imagining now responses telling me I'm a slave to capitalism because I probably got upsold by ten grand or something I wasn't paying attention to when I signed everything. This recalls a discussion last night about grocery prices and everyone at the table talking about the comparison shopping they do and I blurted out I have no idea what anything costs. I just buy the same stuff every week and scan my card. I don't look at the stickers. I don't look at the receipt. I just look at whether I'm hitting my desired savings rate every year and remain on track for retirement goals, which I do, and that's all I actually care about, not whether I'm getting the absolute best deal possible on every transaction I ever make.
It brings my mind back to the article from the Atlantic posted here yesterday about anxiety rates going through the roof in spite of the modern western world being by far the safest it's ever been. It seems the difference now is people are so acutely aware of any and every risk, alerts to which are constantly being shoved in your face, that it overwhelms our ancestral ape brains as if dying with ten grand less in the bank is equal to having a tiger stalking you through the bush 10 meters to your right.
I'm tired. There's a podcast I listen to hosted by exercise scientists debunking stupid press headlines about the latest study or TikTok trend, with one of the hosts always pointing out how exhausted she is by all of these articles telling here she's doing something wrong. "Everything you knew about X is wrong." "You've been doing X wrong your entire life." It exhausts me too, Lauren. I bought the wrong car. I bought the wrong TV. I guess I should be living in fear that my appliances will stop working if I lose network. I should be worried that LG is telling every data broker in the world that I've spent the past year watching nothing but Buffy the Vampire Slayer and Sabrina the Teenage Witch. Shit, some agent probably just scraped Hacker News, correlated my user name with my real identity using stylometry, and now I'm going to be double-triple served nothing but ads for whatever consumer segment that puts me in. I'm going to die even poorer when some cleverly placed sponsored link in a search result makes me impulse buy some whimsigoth living room art I didn't really need.
The world of 30 years ago was a world in which six of my friends from school were killed, the National Guard occupied my city for half a week, 55 people were killed in a riot, a man dragged from his truck and beaten half to death on live television, a shootout with the LAPD that lasted three hours also live televised. We weren't allowed to wear red or blue for fear of kids getting shot by the crips and bloods. My middle school had three drive-bys while I was there. We regularly had recess and any other outdoor activities canceled because of dangerously unhealthy air quality, called "smog days." My little sister's best friend's mom lived under an assumed identity because she survived an attack by the nightstalker Richard Ramirez. My older sister's best friend was murdered in her sleep by her deadbeat mom's boyfriend.
But the world of 30 years ago still felt full of wonder and possibility. I saw a television and wanted to tear it apart to learn how it worked, to understand the miracle of how such a thing was possible. I see a television t...
https://www.thesmartesthouse.com/collections/water-sensors/p...
I have a few of Ikea's other Matter over Thread devices and have had spotty results with them so far using Google devices as thread border routers. Planning to move to a border router from SMLight which I hope will provide a more reliable connection.
I have none of that stuff. My home was purchased new with some 10baseT wired to 5 spots in the home and to every portal (window, door). we have a solar power system with enough battery backup watt hours for backup even if it rains for two weeks
But I DO have wildlife like bobcats and hares and raccoons that are autonomous and require remediation occasionally. Luckily, their activity is limited to the out of doors.
are you sure you are getting what you want?
Even lower tech!
In the year 2000, I lost all my possessions, especially the computers and storage media. Subsequently, I enjoyed NPR and Coast-to-Coast on a radio, including the Hurricane Katrina coverage. I had no idea what New Orleans looked like in 2005.
By that time, I was living across the street from a large thrift store. Eventually in 2006, I was weak-willed enough to begin building a home network again, and the thrift store provided two essential products: keyboards and WiFi routers.
The WiFi router was probably obsolete and EOL when I purchased it, but it definitely worked 100%, especially because I took care at the thrift store, to select a compatible AC adapter brick for it. It was a basic Netgear model, with 4 Ethernet ports, original firmware, probably 802.11n; even though an acquaintance said he would hack into it, he was inebriated most of the time, and I would never know if the router or network was hacked. You wouldn't know either, in 2026. Most consumers don't know, and could never find out, if their edge equipment is compromised. But anyway, even while running Ubuntu 6.04, I accepted the router's default firmware, and that Netgear thing lasted at least 5 years while I had an ADSL connection, and one and a half computers.
In the mid-1980s when I hung out at Radio Shack after church, I snapped up a lot of basic X10 kit. I collected the Dimmer Modules, the Appliance Modules, and later there was an RF remote unit available! The crown jewel for 12-year-old me was a central control unit that looked like a clock radio.The control unit had cyan fluorescent display, simply the current time and alarm settings. And there were a bunch of control buttons where you could select an addressable Module, dim the lights, on/off, set timers, etc. And it was kind of amazing that everything plugged into mains outlets, and no other installation necessary!
My parents were constantly annoyed when I would sneak one of the modules onto a living-room floor lamp, or my favorite trick was cueing up a record on the turntable and then activating the Appliance Module to wake me up for school with Siouxsie and the Banshees or something. I was not allowed to persistently wire the entire house. I did not get into the journeyman electrician's job of replacing wall switches with X10 modules (those were slick). So, not even my shared bedroom was totally automated, but it was fun to experiment. It hopefully did not affect any neighbors. The pulse signals were said to stop at junction boxes, but there was still a House Code to set, in case all your flat neighbors wanted independent X10 networks.
I also never purchased the PC Interface that came with software. I believe I didn't yet have a compatible PC, anyway, and it seemed not in the spirit of a robust, predictable home automation. So yeah, my smart home devices worked without the Internet. How smart do you wanna get? My television set didn't spy on me, but everyone else does!
That is the worst part: being explicitly designed to fail un-safe. You should be able to declare the unit unfit for the purpose it was sold for and get a full refund unless they open the specs so you can keep the device running as intended using your own resources.
https://github.com/alaudet/raspi-sump
https://raspisumpdocs.linuxnorth.org/
There should be a system like the nutritional information on food and beverage products, as time went by most people read those and choose accordingly, which then incentivize companies to adapt their offering. We could imagine something similar which would define clearly the capabilities of the hardware or software sold, clearly labelled on the physical package or on the online shop with an easy classification for consumers to understand.
If they could get away without even having a Wi-Fi router and just use their cellular plan they would. Why? Because there are plenty people who are happy with just whatever internet they get on their cell phone and don't want to spend extra for a router and an ISP
With offline solutions like the ikea smart home stuff you can access it from your phone with even less setup than the wifi based options.
Something anyone can use.
You were warned.
If you accept and thus consented to a computer, or a device with a computer in it, without first verifying you can write to the media of its software, control the boot process, and will have sufficient access to that software, you get what you consented to.
And, you deserve it.
My home automation rules are even simpler: home automation devices are firewalled off the Internet. If something does not work because of this, it gets returned, period. I do my research and only buy Zigbee or WiFi devices that can be hacked, so I haven't had problems so far.
Smart home devices should not freeze up whenever internet is down. A device like Google Nest should have the minimum decency to allow you to interact with it via local Wifi (which continues un-interrupted). It is fine if half the functionality doesn't work. It is fine if the "AI" doesn't work, but at least allow me to fcking change the temperature manually
Same thing with devices like Google Wifi which freezes up or run hot loops like headless chickens when internet is down.
All Nest thermostats allow that, even the ones that are no longer supported by the Google cloud.
Oh I also have a under-the-table contract with some intelligence community for some backdoors. You know it's a lot more expensive for them to actually send someone to test physical security, than just pay a contractor to sit before the desk collecting data, right? RIGHT??
I don't allow my devices to update their firmware. If the device does not work out of the box, I return it, end of. My car has also been taken off the grid.
I only use "Local Polling" or better, and highly prefer "Local Push".
I also have some eve energy plugs that do matter/thread (also push/local). There are apps here but i generally haven't needed them once i got things set up in home assistant.
Smart Plugs, sensors, etc: Ikea is amazing. You need to buy a hub, but then accessories are 3x cheaper. Their plugs are $9 and have energy monitoring which is amazing. They have water leak sensors for $12. Incredible value.
+1 on the IKEA stuff being amazing!
https://community.home-assistant.io/t/ikea-inspelning-qa-iss...
Most of their sensors are meh. Their PIR sensors are known to be mediocre compared to the competition. Their air quality sensor is ok to get trends, but e.g. the CO2 measurements are quite a bit off compared to quality sensors and are way off if you live at a higher altitude (due to using a thermal conductivity sensor and not NDIR like better sensors).
Also, the IKEA smarthome app (used in conjunction with the Hub) uses Play Integrity and therefore blocks e.g. GrapheneOS. But you are better off with Home Assistant and their ZBT-2 antenna (can be flashed for Zigbee or Thread) anyway.
That's 2nd generation, first one was just a basic on/off smar plug. I still got these and so far I had no issues even after Dirigera hub got that firmware update which enabled Matter. Ikea 'gear' overall work much better than Meross one I had before.
GRILLPLATS is the successor. I have one, but haven’t tried it much yet because it is Matter over Thread. They have a hidden Zigbee mode, but it doesn’t expose power use:
https://www.zigbee2mqtt.io/devices/E2435_E2491.html
But if Matter over Thread fits into your system, it’s an affordable option.
For power strips, we are now using the Gen4 Shelly Power Strip. It supports pretty much anything: its own API over WiFi that is supported by HA, Shelly cloud, Matter over WiFi, Zigbee, and Bluetooth. It can even work as a WiFi range extender and as a Bluetooth proxy for HA. Initial setup is by connecting to it as a WiFi AP and the connection to the built-in web interface. There were some issues with Zigbee flooding, but I think that was improved with firmware updates.
You can even flash ESPHome: https://devices.esphome.io/devices/shelly-power-strip-4-gen4...
Currently using a bunch of Nous A8T (EU plug) because they come with Tasmota ootb.
That's also part of it's advantage though: I have a Mikrotik router which has a Thread radio, so I run the router software on that, and Matter server on another machine - so no USB dongle needed in my server rack where the signal would be poor.
Also Matter over Thread supports multiple Thread routers, multiple Matter servers, and devices being paired to multiples of each - maybe that doesn't help with the complicated part, but it opens it up for commercial deployments as you can have redundancy. I would not be surprised if it replaces DALI in some instances in the not too distant future.
There are still some teething issues, which is because it's still kind of early in its adoption (ZigBee felt the same before Zigbee2MQTT). Personally I feel it's the way everything will go simply because it is more interoperable. Already today you can get routers, TVs and fridges (Samsung) with Thread hardware, so it's going to become mainstream as regular people will be able to just buy a smart bulb, pair it, and control it with their phone without buying or setting up anything else.
IKEA released a bunch of new smart home devices that support Matter over Thread and ZigBee this year if you want to try it out.
I get that MoT has more features, but it seems like it makes the whole thing less reliable. I also hate the pairing process with Matter-over-Thread. Enable IPv6 in your home network, scan the QR code with your phone... What if the QR sticker falls in a few years? With Zigbee I just press the reset button and I see it directly in the HA web interface. No need for a phone, no need for camera, and doesn't touch your router. Less features, but my devices have been working for years without a single failure.
With Matter/Thread though, i frequently run into issues. Particularly these were power switches from Ikea by the way, but i don't thing that's relevant.
Also the range of Zigbee devices is much better / bigger, since each device "expands" the network range so to speak.
I'm disappointed that Ikea decided to switch their smart home gadgets from using Zigbee to Matter over Thread.
* https://handbook.buildwithmatter.com/how-it-works/thread/#ip...
> Thread defines a series of features to support multicast across a heterogeneous network (Thread and Wi-Fi/Ethernet Network segments) for multicast addresses with scope larger than realm local.
* https://openthread.io/codelabs/openthread-border-router-ipv6...
?
https://github.com/OpenHomeFoundation/roadmap/issues/216
But this lack of (application-level) multicast is what I was referring to:
https://community.home-assistant.io/t/wth-popcorn-effect-wit...
ZWave: the consortium demands interoperability. Testing and licensing fees means more expensive devices. Things will work together, though sometimes you will need to figure out what parameters actually do.
Use both. Home assistant is perfectly happy with two dongles plugged in.
ZWave stuff tends to be rock solid (there is expensive certification involved) but that means the devices are more expensive and the selection can be limited because the volumes aren't always there to make sense to get certified.
The only real problem I've found with the "all of the above" situation is that device bindings only work within one protocol. If you want to bind a lightbulb to a wall switch such that they don't require a trip through the hub they need to both be zigbee, zwave, or matter. This is unfortunate because there are a lot of really nice ZWave wall switches but not a lot of ZWave lightbulbs.
They kind of do, if your WiFi uses 2.4GHz bands
ESPHome or Tasmota work really well for all kinds of things. The last one I replaced was an infrared blaster for controlling my AC. The previous one, even though it was in an IoT VLAN decided to spam the network with requests to its mothership (never succeeding, fortunately).