It started innocently. I wrote a little program that prints news headlines and random facts when you press a button. Which is objectively cool, I suppose. But after the novelty wore off, I realized what I really wanted wasn’t just information…

I wanted mail!!

Because getting mail is mostly fun. Especially when it’s not junk. And there’s something satisfying about a machine physically spitting out a message into the real world like, “Congrats, you have been perceived.”

That idea became TinyMail: a tiny web app that lets anyone send a message (and optionally a photo) to my receipt printer.

And after I started posting the printouts on Instagram, people began using it more… and more… and more. Apparently, we as a society are not done enjoying the thrill of “I sent a thing and it showed up somewhere.”

What TinyMail Does

TinyMail is simple on purpose:

• Visitors type a name (optional-ish, you can put whatever you want or nothing at all)

• Write a message

• Snap a photo (optional)

• Hit send

• The receipt printer does what receipt printers do best: prints and cuts.

It’s basically a public guestbook, but instead of living on a website forever, it gets printed into REAL LIFE existence like a tiny monochrome postcard.

The Surprisingly Good Economics of Tiny Paper

Thermal receipt paper is cheap, and a roll lasts way longer than you’d expect.

A common roll size for 80mm printers is about 262 feet of paper (80 meters). How many TinyMail messages does that become? It depends on how long each print is, but here’s a real-world estimate that I spent way too long working out:

• If an average TinyMail print is 6 inches (15 cm) long

  • 80 meters ÷ 0.15 meters = ~533 message

    s

• If it’s a little longer, say 8 inches (20 cm)

  • 80 meters ÷ 0.20 meters = 400 messages

So a single roll is roughly 400–530 messages.

In other words: many many many manyyy messages per roll is not an exaggeration.

Building It Was the Fun Part (But Also: The “Oh No, The Internet” Part)

Once you make something public-facing, especially something that prints on physical paper, you have to understand that there is plenty of room for abuse.

So over time, TinyMail grew up a little. It gained features and protections to keep it fun without becoming a streamer generator.

Bot Protection: Cloudflare Turnstile

I added Cloudflare Turnstile (a CAPTCHA alternative) to reduce automated spam. It’s a nice balance: it keeps bots out without making humans solve six puzzles about traffic lights like they’re studying for a CDL.

Rate Limiting

Even without bots, one enthusiastic person with free time can turn any public form into chaos.

So TinyMail rate limits requests to prevent someone from hammering the printer nonstop. The goal isn’t to kill the vibe… just to ensure it doesn’t become a paper waterfall.

Privacy: “Cute Project” Doesn’t Mean “Creepy Project”

Privacy matters. A lot. I get it.

TinyMail is designed to be lightweight and respectful:

• Photos are only stored long enough to be processed and printed, then they’re removed.

• The app doesn’t collect identifiable information beyond what the user voluntarily types into the name or message field.

• The user’s IP address is logged only so rate limiting can be applied.

• Once the rate limit window passes, that record is purged.

That’s it. No profiles. No tracking. No targeted ads.

Why This Project Rules

TinyMail scratches a very specific itch:

• It makes online interaction feel physical

• It turns messages into artifacts

• It’s simple enough that anyone can use it

• And it creates this tiny moment of connection where someone out there goes “I just printed that to his printer.”

It’s silly, wholesome, and just the right amount of over-engineered. (The best amount)

Give It a Shot!

Use the link below to send me some TinyMail!

Subscribe now

Send TinyMail

The dishwasher finishes the moment you forget it exists, and laundry only becomes urgent once it smells slightly suspicious.

To solve this, I built a set of Home Assistant automations that watch power usage, determine appliance state, and then politely notify us when something is done.

No smart appliances required.

The Hardware: Energy Tells No Lies

At the core of this setup is a Refoss energy monitor, installed at the electrical panel using CT clamps. I found it easy to install, but I’m going to tell you to hire an electrician.

Each appliance (dishwasher, washer, dryer) has its own circuit monitored, giving me real-time power readings straight to Home Assistant.

The Secret Sauce

Instead of relying on a single “power dropped = done” trigger, each appliance uses a three-state lifecycle tracked with a helper:

• Off – Appliance is idle

• Running – Appliance is actively doing its thing

• Done – Appliance finished and is waiting for human intervention

This prevents false alerts when someone opens the door mid-cycle or power briefly dips.

Dishwasher Logic

▶️ When it Starts

• Power rises above 5W for 1 minute

• Status flips from Off → Running

⏹️ When it Finishes

• Power stays below 5W for 5 minutes

• Status flips Running → Done

• A green indicator on our AWTRIX Kitchen display turns on, a message is displayed and a quick tone is played.

✅ Acknowledgement

• Zigbee tilt sensor on the dishwasher door is at 90 degrees for 5 minutes

• Status moves Done → Off

• Indicator light turns off.

Dryer Logic

Dryers are less subtle, so they get different thresholds.

▶️ When it Starts

• Power above 40W for 1 minute

• Status flips from Off → Running

⏹️ When it Finishes

• Power stays below 50W for 5 minutes

• Status flips Running → Done

• A cyan indicator on our AWTRIX Kitchen display turns on, a message is displayed and a quick tone is played.

✅ When It’s Acknowledged

• Clear button located in the laundry closet is pushed

• Status moves Done → Off

• Indicator light turns off

I’m only monitoring a single leg of the dryer’s power. The first leg I tried would drop to 0W during cool-down, which caused the notification to fire early. This lead me to believe the drum motor and controls are only powered on the other leg. Moving the CT over made sure the alert only triggers when the dryer is actually finished.

Washer Logic

While working on this project, I found that washers fluctuate wildly. Spin cycles, fills and seemingly random pauses. Our new front load washer is a bit less dramatic since there are no pulses from an agitator.

Additionally, we have a small peltier dehumidifier and the Dyson vacuum charger in the laundry closet which shares a circuit with the washer. I just needed to tune the threshold to account for the power draw of those devices.

▶️ When it Starts

• Power above 80W for 1 minute

• Status flips from Off → Running

⏹️ When it Finishes

• Power below 80W for 5 minutes

• Status → Done

• A cyan indicator on our AWTRIX Kitchen display turns on, a message is displayed and a quick tone is played.

✅ When It’s Acknowledged

• Clear button located in the laundry closet is pushed

• Status moves Done → Off

• Indicator light turns off

Just One Laundry Button?

I didn’t see a need for separate reset buttons for the washer and dryer. Instead, a single button handles both. When pressed, it only resets appliances that are currently marked as Done, returning them to Off.

For example, if the washer is still running but the dryer has finished and been emptied, pressing the button will only reset the dryer. If both appliances are marked Done, they’ll both be reset—so yes, that’s your cue to empty both.

Notifications That Exist in the Real World

We’re all pretty desensitized to push notifications and text messages. So to ensure it gets our attention in a somewhat subtle way, I use the following:

• Indicator lights on the AWTRIX Display in the Kitchen

• Audible alerts using RTTTL tones

Why This Works So Well

✔ No smart appliances required

✔ No cloud dependencies

✔ Highly configurable thresholds

✔ Scales easily to other appliances

Final Thoughts

This is one of those automations that quietly becomes essential. Once it’s running, you stop thinking about it and your life adapts to it.

Next up?
I am considering adding a reminder with timing depending on the appliance. A reminder to empty the washer is far more time sensitive. Meanwhile, those dishes can hang out in the dishwasher until tomorrow

Things like:

• A guest has arrived

• The laundry is done

• Someone is on the nice list

• CO₂ levels are high, open a window

I wanted visibility without chaos. Awareness without anxiety. Information without my phone because I’m ignoring most of those notifications anyway.

Why AWTRIX?

The moment I stumbled across the project, I ordered one immediately. This is my version of “doing research”.

Its compact size, clean pixel display, and ridiculous flexibility make it perfect for ambient notifications. Ours lives on top of the kitchen cabinet, clearly visible from the living room but not screaming for attention like a billboard in Times Square. (But I wouldn’t say no to a Vesta Board if one showed up at my door.)

When it’s idle, it simply shows the time. Calm. Respectful. Mind-your-business energy.

How It Works

The AWTRIX is configured to subscribe to an MQTT broker that’s already running alongside Home Assistant. Whenever something worth knowing happens, Home Assistant publishes a message.

The flow looks like this:

Thing happens → HA publishes MQTT message → AWTRIX sees it → AWTRIX displays it

If I wanted to add more AWTRIX displays later, it would take almost no additional configuration. Additional displays would just need to be configured to subscribe to the same MQTT topic. Scales beautifully.

The Sounds of Notifications

Many of the notifications include a short chime. The AWTRIX supports RTTTL tones only, which means…

Yes.

It sounds exactly like an old-school Nokia ringtone.

I didn’t plan this. I would have denied wanting this. But now? I love it.

It’s nostalgic, unmistakable, and somehow less stressful than modern notification sounds.

(Shoutout to Nokia for accidentally inventing the perfect home automation alert format. Miss you!)

The Weird Nostalgia I Didn’t Expect

This setup unlocked a memory I didn’t know I still had.

When I was a kid, my dentist’s office had this non-verbal communication system. Whenever a patient arrived, was ready, or needed something, a staff member would press a combination of colored buttons.

An analog tone would play.

Colored lights would blink on a board somewhere.

I had no idea what any of it meant, but I was obsessed with trying to guess.

Now? Same vibe.

Whenever the AWTRIX makes its little tone, we both glance up, read the message, mentally file it away, and go back to what we were doing. No disruption, no discussion required.

Exactly like the hygienists did.

Subtle, Informative, and Slightly Secretive

The display also uses small indicator lights to represent status at a glance. Laundry? Dishwasher? Guest Access?

Only we know what those colors mean.

To everyone else, it’s just a neat little glowing rectangle.

To us, it’s a shared language.

Which, apparently, completes my transformation into a dentist’s office from 1997. Don’t forget to get a prize from the toy chest on your way out!

Dangerous Ideas (Proceed with Caution)

Here’s a thought that definitely won’t spiral:

If my husband is sick in bed…

I could set up a Zigbee button that sends a notification when pressed.

“Assistance requested.”

Displayed quietly.

With a Nokia chime.

…Okay.

We’ll see about that one. 😄

Final Verdict

This has been a game changer.

The AWTRIX doesn’t interrupt. It doesn’t nag. It doesn’t demand attention.

It just… informs.

And somehow, that makes all the difference.

If you’re drowning in push notifications but still want to stay informed, this little pixel display might be exactly what you didn’t know you were missing. At least, while your at home...

The goal was simple: AirPlay everything! Music, AppleTV audio and party playlists to all speakers at once, Sonos and Nuvo alike.

Why Not Just Buy the Sonos Amp?

Sonos sells an amp that would’ve solved this in, like, ten minutes. But… Sonos pricing. It would have sounded great, I’m sure, but the project wasn’t exactly didn’t demand the premium route.

So instead, I dug out a Raspberry Pi gathering dust, plus a 3.5 mm → RCA cable. This led me down a rabbit hole I now affectionately call PiTunes.

Shairport Sync

I stumbled on Shairport Sync, which is an open-source AirPlay audio player for Linux. Basically, it lets computers pretend to be AirPlay speakers. It supports:

• AirPlay 1 & AirPlay 2

• Sync’d multi-speaker playback

• Various audio backends (ALSA, PulseAudio)

• Metadata display and timing corrections

➡️ Full project: https://github.com/mikebrady/shairport-sync

I fired up Shairport Sync on the Pi and initially had it broadcasting as an AirPlay 1 device. Great, but that doesn’t support multi-speaker sync. The meant I couldn’t play the same audio in sync with my Sonos speakers.

I poked around and found out that multi-speaker playback is only available for AirPlay 2 devices. Luckily, Shairport Sync does have AirPlay 2 support, and once enabled, PiTunes showed up as an AirPlay 2 target. That meant I could:

✔️ Select PiTunes + all Sonos speakers

✔️ Play music everywhere in sync

✔️ Not have echos with timing mismatches

Magic!

Volume Woes... Because Why Would It Be Easy?

Turns out the Raspberry Pi’s built-in audio output is… modest. Probably giving the same power as a walkman, it needs to be fed into something that amps the signal. The Nuvo system does amplify it, but it wasn’t enough, especially to drown out my singing in the shower.

Solution: I added a small headphone preamp (20 dB gain) between the Pi and the Nuvo input. Audio became loud enough and the Pi’s output can be turned down for cleaner headroom.

But here’s the twist: over time, the volume started drifting lower, even with the preamp. Reboots would reset the Pi’s mixer, and ALSA settings wouldn’t always stick. After some wrestling with alsamixer and persistent config files, it’s mostly stable. Occasionally a reboot still requires a quick manual volume check.

Because the Pi is plugged into a UPS, reboots are rare. So I’ve let this one sit for now. That’s a problem for future me.

Is the Nuvo System Dying?

I have noticed the home speakers sound slightly quieter than before, even with the same PiTunes volume settings.

My current theory is that the Nuvo amp is maybe… fading? Aging electronics do that. I haven’t run any formal tests yet, but it’s on the to-do list after I finish ~30 other half-baked projects.

For now, the system works well enough, our place has synced audio throughout, and that’s good enough for us.

Future Adventures: Paging with SIP

At one point I was working on a SIP client on the Pi that would be used to auto-answering and auto-broadcasting into the system for paging.

That experiment deserves its own post. Stay tuned for another post that I will probably call.... PiPage!

TL;DR

• Built a Raspberry Pi–powered AirPlay 2 target with Shairport Sync

• Integrated it with an old Nuvo system via a preamp

• Syncs with Sonos speakers for whole-home audio

• Volume quirks solved (mostly)

You’re expecting someone
You miss the call box call
Your guest is left waiting

I’ve lived in many places that have had callboxes and it’s always been just a little annoying:

• We had to answer every call box call to let a visitor in and most of the time we didn’t know which of us the visitor was going to call

• Hosting a party? Don’t loose track of your phone, otherwise someone will be stranded

• We occasionally left our phones and key fob inside, locking us out of the building until we could follow a neighbor inside

So naturally, instead of accepting this as “how buildings work” or forcing myself to remember my phone and/or key fob, I built Callbox Concierge.

How it started…

My building’s call box has an entry for both me and my husband. Normally, each of these go straight to our respective cell phones.

I asked the property manager to change my number to a dedicated phone number that routes to my FreePBX server instead.

I was nervous at first, but was feeling better knowing that the number associated with my husband’s name is still his actual cell phone number. So if my plan didn’t work out, we could at least have that until I could switch it back.

Version 1: Just for Us

When a call comes in, FreePBX does the following:

1. Inbound route checks caller ID

   • If it matches one of the known call box numbers in my building, it’s sent to a special dial plan.

2. The dial plan checks Home Assistant

   • If you’re new to my life, I run Home Assistant for automation.

   • There’s a virtual switch called Guest Access.

   • This switch can be toggled:

     • From Apple Home

     • From a physical button on our fridge (yes, really)

If Guest Access is ON

• The system immediately plays the sound of a “9” button press

• That’s the unlock key for our building

• Door opens, no interaction required

If Guest Access is OFF

• The caller hears a prompt:

  • Enter an access code or

  • Press 0 if you don’t have one (for the unexpected guest or delivery driver)

Outcomes:

• Valid access code → unlock key plays → door opens

• Press 0 → call forwards to a ring group:

  • Both of our cell phones

  • Desk IP phones (who doesn’t love a corded phone?)

Notifications

Any time someone is granted access, whether by code or Guest Access, we get:

• A visual notification on a display in our kitchen

• A paging announcement over the various IP phones

• A text message heads-up

At this point, we were both sold.

We barely use access codes anymore. If we’re expecting someone, we just turn Guest Access on.

I honestly can’t imagine not having this now.

“Other People Might Want This…”

As you can imagine, the functionality is sometimes a topic of conversation when people come over. Those who saw the value expressed an interest in having something like this.

The first version, however, was very much built for us, not others.

So I started thinking... Phone numbers from my provider are cheap. Like, shockingly cheap.

So I did what any reasonable person would do:

I rebuilt it from scratch…

Version 2: Callbox Concierge

The new model is built around subscribers.

Each subscriber gets their own dedicated DID (phone number) that they give to their property manager. I call this number their “RainbowLink Number”.

Incoming Call Flow

When a call comes in:

1. FreePBX identifies the DID as belonging to a subscriber

2. The dial plan matches the DID to a subscriber record in the database

3. It checks the subscriber’s current Concierge Mode:

   • Allow All

   • Code / Normal

   • Forward

Concierge Modes Explained

Allow All (Guest Access)

• Plays the subscriber’s configured unlock key (usually “9”)

• Door opens immediately

Code / Normal Mode

• Caller is prompted for an access code or to press 0

• Valid code → unlock key plays

• Press 0 → call forwards to the subscriber’s configured phone number

  • They can answer, talk, and press the unlock key themselves

Forward Mode

• Concierge is bypassed entirely

• Call forwards straight to the subscriber’s phone

Self-Service

Subscribers can manage their system in two different ways.

1. Dial-In Management

• Call your RainbowLink Number from your registered cell phone number

• Caller ID is verified

• Enter a PIN (just in case someone out there is doing Caller ID spoofing)

• You’re dropped into an admin menu where you can:

  • Change the current mode

  • List access codes

  • Add or remove codes

2. Concierge Portal (Web)

Manage:

• Current mode

• Access codes

• Subscriber details

Log in using your registered email address

Building the Portal

The original Concierge system was built with:

• MariaDB

• Python

• PHP AGI (the “brain”)

• Asterisk dial plans

That handled dial-in management just fine.

But of course… that wasn’t enough.

Concierge Portal

I built a full web interface using Laravel.

This was my first time working with Laravel, and while there was a learning curve, I was pleasantly surprised by how efficient and well-structured it is.

That said:

• I am still wrestling with CSS

• Overrides don’t always behave how I expect

It looks okay right now. There is absolutely room for improvement. That’s a problem for another day.

Currently:

• One user per subscriber, but this would be easy to expand with the current structure

• Google-only authentication (no shared passwords)

• Multiple admin phone numbers per subscriber are already supported

Multi-user portal access is on the roadmap.

The Admin Side

Once subscribers existed, managing them cleanly became essential. I was not about to be over here manually adjusting things in the database.

Anyway, that lead me to building:

• A separate internal Admin Portal, just for me

• Used for:

  • On-boarding

  • Off-boarding

  • Managing subscribers

At this point, Concierge became something I could easily manage for as many subscribers as I need.

“But What If the Server Dies?”

Great question, and believe me... I’ve gone through the scenarios in my head...

Could I build but a robust, multi-node, multi-site platform to drastically reduce the downtime risk? Absolutely. But I’m not at the point where full high availability makes sense yet, but:

• My PBX already has a warm spare on a VPS

• It can handle basic call functionality if my primary site is down

• The web portal wouldn’t be available, but calls into the admin menu would still work

Last-Resort Fallback

My phone provider allows a failover forwarding number if both PBXs are unreachable.

Right now, I configure this to forward calls directly to the subscriber’s cell phone, which should prevent almost all full-service interruptions.

Final Thoughts

Callbox Concierge started as a way to stop answering the same phone call over and over.

It turned into:

• A custom dial plan

• A subscriber platform

• A web portal

• An admin system

• And a fridge button (still my favorite and the most used part)

It works beautifully for us and now it’s built in a way that others can use too.

More to come!

The North Pole Department of List Verification Hotline was born!

You call in.
You wait.
Clementine Bell consults the official records.
And you find out where you stand.
Nice. Naughty. Or… Undetermined.

The Inspiration

The idea came from a North Pole communicator that both my sister and brother had when their kids were younger. It was a little device that let them “call” the North Pole and hear messages from Clementine Bell herself.

This hotline was my attempt to recreate that feeling. Something fun, slightly mysterious, and charmingly over-the-top just in time for Christmas.

How the Hotline Works

When you dial the hotline, the call lands in a dedicated dial plan. From there, the experience is carefully paced:

• The phone rings briefly (otherwise it answers right away and you might miss something!)

• The call is answered

• Clementine introduces herself

• She starts looking you up to see your status

Behind the scenes, that look up is powered by a little randomness:

• A random “look up” audio clip plays (one of many)

• A random number between 1 and 20 is chosen

• That number determines your fate:

  • 1–10 → Nice (50%)

  • 11–15 → Undetermined (25%)

  • 16–20 → Naughty (25%)

So yes… statistically speaking, it is easier to be Nice. Clementine believes in hot coco and optimism.

Avoiding Holiday Limbo

Early on, I realized there was a flaw:

No one wants to call twice and get Undetermined twice in a row and callers that did didn’t call back. I’d call that a poor experience.

So in a later version, I added logging for each call, noting their phone number and what their result was. With that in place, I built logic to inspect the caller’s most recent result at the beginning of the call.

If your last call resulted in Undetermined, your next call is forced to be decisive and the odds are updated:

• 50% chance for Nice

• 50% change for Naughty

• No back-to-back ambiguity

Clementine’s keyboard may be full of gingerbread crumbs, but she is not inefficient…

I actually wish I had started doing this much sooner because this added functionality allowed me to run reports and provide statistics on the outcomes.

It seemed like I was making improvements almost every day after launch. (I even kept a change log!) With enhancements such as:

• Removing unnecessary keypresses

• Tightening timing (including keeping most calls under 1 minute)

• Improving audio pacing

• Making outcomes feel fair but fun

Home Assistant Integration (Because I Had To)

Eventually, I wanted to keep tabs on what Clementine was declaring in real time.

• When a caller is declared Nice or Naughty:

  • A Home Assistant text field called “last_decision” is updated as soon as it is determined.

  • A virtual button called “send_decision” is pressed right after Clementine finishes reading the decision.

• Undetermined results are intentionally ignored… I only wanted to hear if someone got Naughty or Nice.

A Home Assistant automation listens for that button press and sends the result to an AWTRIX display in our kitchen, so we get a festive little announcement when Clementine reaches a verdict, complete with a monophonic chime of Jingle Bells.

It’s subtle. It’s unnecessary. It’s perfect.

I even brought the AWTRIX display along to our family’s Christmas Eve gathering, which was fun for us to watch as everyone called to check their status.

Writing the Voice of the North Pole

All of the prompts were written specifically for this project and recorded using Google’s Text-to-Speech voices.

It was surprisingly time-consuming, but also a lot of fun.

ChatGPT helped me get started with the scripts, especially in finding the right tone. Clementine needed to be authoritative, cheerful, and just official enough to feel real without breaking the illusion.

A few revisions later (and some aggressive silence trimming), Clementine sounded exactly how she should.

Popular Demand

What started as a fun little seasonal project turned out to be… very popular.

So popular, in fact, that I had to move the hotline to a different VoIP service provider just to handle the number of concurrent incoming calls.

That was not a problem I expected, but I’ll happily take it.

Final Thoughts

Obviously, the North Pole Department of List Verification Hotline was never meant to be serious.

It was meant to:

• Spark nostalgia

• Make people smile

• Add a little magic to December

• Engage people using a slightly more analog technology

• And, naturally, give me an excuse to build something ridiculous with FreePBX/Asterisk

Mission accomplished.

If you were one of the 250+ people that called in, thank you! I will begin working on 2026 Clementine soon so she will be better then ever next season! Please let me know if you have any suggestions!!

And remember… if your status couldn’t be determined… call back!

Some track sleep.

I track the vibes of the air in our home.

This project started innocently enough: I wanted to understand how temperature and humidity varied across the space and how well (or poorly) the HVAC system was doing its job. Naturally, this escalated.

Phase 1: Temperature, Humidity and HVAC

I began with Zigbee temperature and humidity sensors placed all over the place:

• One in each room

• One on the balcony for outdoor reference

• One at the HVAC return

• One directly on a supply vent

All of this feeds neatly into Home Assistant, where I built a dashboard to visualize room-by-room temperatures and the HVAC delta between supply and return (inspired by my experience with commercial building automation systems).

This alone was enlightening:

• Rooms don’t heat or cool evenly (shocking, I know).

• The HVAC delta tells me all about system performance.

I checked this dashboard constantly… at least until the novelty wore off and it became “background knowledge.”

Obsessive? Absolutely.

Sustainable obsession? Apparently not.

Phase 2: Enter the AirGradient (aka: The Point of No Return)

Just when I thought I was done, I added an AirGradient indoor air quality sensor.

And wow. That escalated quickly.

Now I’m tracking:

• CO₂

• PM0.3, PM1.0, PM2.5, PM10

• VOC Index

• NOx Index

  

What the Data Taught Me (Besides Humility)

The biggest takeaway? Indoor air circulation is… not great.

CO₂ levels rise faster than you’d expect, especially with doors and windows closed. Elevated CO₂ is known to impact cognitive performance, and seeing it climb in real time is a powerful reminder that sometimes the smartest automation would be… opening a window.

Honestly, I think everyone should have a CO₂ monitor. Not to induce panic… but to gently say:

“Hey, maybe let the outside in for a minute.”

The particulate and gas sensors were equally revealing:

• Cooking spikes PM levels fast (especially frying).

• Laundry and mopping have a noticeable VOC impact.

• Candles are tiny pollution machines.

• Leaving the door open while the Turkish restaurant nearby is cooking onions? Immediate confirmation. No notes needed.

Mitigation

We do have a standalone air purifier, and it absolutely works. You can see particulates drop in near real time once it detects the pollutants and the fan turns up. That said, the data made me think about how many people don’t have these things..

• Air quality monitoring

• Active purification

• Or even awareness that indoor air quality can vary this much

You can’t fix what you can’t see. But I guess seeing everything is not always great either.

Final Thoughts

Is this setup overkill?

Yes.

Do I regret it?

Not at all.

The combination of distributed temperature sensors and the AirGradient has turned my home into a living, breathing dataset. One that gently nudges better habits, including a subtle AWTRIX notification when CO₂ levels get too high.

And if nothing else, it’s deeply satisfying to say:

“These onions aren’t just strong… they’re measurable.”