Dabbling with Home Automation

TL;DR

Kasa smart plugs and light switches work well and are very affordable. If you are curious about home automation, you can experiment with little effort and at a cost of $15-$100. The plugs are super-simple, the light-switches require very basic electrical wiring competency.

More

I've begun dabbling with home automation, mostly for lighting. I have purchased the Kasa EP25 smart plugs (4 for $40) and Kasa Smart light switches (<$30) that are well-reviewed by Wirecutter.

One of the light switches is motion-sensitive. I'm still fine-tuning it, but seems to work well. In hindsight, I would say if you think there is any possibility you want motion-detection, get it, because it doesn't cost much more, and you can turn it off if you don't want it.

Setup of the smart plugs was reasonably easy. The hardest thing is the fact that they only operate on 2.4 Ghz wifi, and require the setup device (smartphone) to be on the same band. Since phones will generally default to 5.0 Ghz, this requires going into your router control panel (not hard, but you may not remember how to access it or the password), and temporarily turning off 5.0 Ghz (don't forget to turn it back on).

Installation of the switches was not hard, and I rarely do electrical work. There is one gotcha, which is the volume of the plug is>> than a dumb plug, and you may have a hard time jamming it all back in the box.  I'm sure a pro could manage it, perhaps by trimming excess ground and neutral, but I wasn't up for that. After some searching, I came up with a pretty good work-around: these $5 box extenders.

The Kasa app is pretty good, and the plugs, though not the switches, are HomeKit-compatible.

My #1 need was to provide good & reliable wall-switch illumination of rooms that didn't have overhead lighting. Each room had the typical (1) switch-controlled outlet. So that takes care of one light, and doesn't require any smart stuff to work as expected. That was important, I don't want an over-complicated home where a guest can't turn on a light without installing an app. Then the other lights in the room are connected to a smart plug, which is triggered to go on with the smart switch. All works well, albeit with a ~3-second latency of the additional light to go on.

I also automated the fan in our bedroom that provides "white noise". It is on the floor in the corner, and my aging back dislikes stooping to turn it on and off. So I put it on a timer that corresponds to bedtime + 1 hour.

Similarly, I have a foot-warming carpet in my office that I also have to stoop to turn on. And that I sometimes forget to turn off. So I hooked that up to a smart plug, and put a turn-off schedule for the end of my workday, in case I forget.

Our Christmas tree lights are also very inconvenient to reach, so those also went on a timer (though given they are LED, I think the energy savings is negligible). On that note, these plugs include energy monitors. It is mildly interesting to see how much energy different appliances consume.

Last, I have a Corsi-Rosenthal fan setup in the basement, where the cat lives. I like to run that a couple of hours a day, not for cooling but for filtering. That was easy to put on a schedule

So that's it for implementation experience so far. In the process, I did quite a bit of research--much more than was needed for my relatively straightforward use cases--so I will document that next.

Apple & Subscriptions

I hate the fact that Apple rakes in up to 30% of subscriptions through the app. BUT--they sure do make it easy to monitor and cancel. I routinely check my subscriptions (mostly for streaming services), and set calendar reminders for when I need to cancel. Or if you are pretty sure you just need the one month already subscribed, you can proactively cancel, and you get the service till the end of what you have paid for(!) Quite a breath of fresh air in that regard.

Here's a feature idea: save me the manual step of the calendar entry--let me set reminders tied to subscriptions, by clicking on the subscription. (For extra credit--this is asking a lot probably--create calendar entries as reminders.)

I want a smart light switch that has up/down positioning

I don't live alone. It is absolutely critical that when a person stumbles into a room, when they fumble on the wall to flick the light switch, they get the expected illumination. The smart behavior can't get in the way of that.

Many of the smart switches, including the Kasa ones, don't work this way. They have a push interface that toggles between on and off. But you have no way, just by looking at the switch, to know which is on and which is off. 

This can be made to work nearly the way I want, replicating dumb switches. I simply connect the switch's ON and OFF states to the "toggle lights" scene. So the user who hits the switch to change the state of the lights gets what they expect. But it is still less elegant, and slightly less obvious, than if UP meant ON and DOWN meant OFF.

One counter-argument is that the state of the switch can become inconsistent, if the lights are turned OFF, say, via a smart app. To me, the solution is that toggling the switch DOWN and back UP will solve this problem. That is probably the most natural thing to do anyway. 

Another alternative would be to extend the behavior of the 3-pole dumb switch (2 switches controlling the same light). UP and DOWN don't necessarily correspond to state, but flipping the switch guarantees a toggle. The problematic detail here is what to do if state is mixed, for multiple connected lights.

Business idea: Customized service to notify a fan it is worth watching the game

Recording sports and watching them on my own time & terms is transformational. I think I would watch 90% less sports if I had to suffer through them in real-time.

I am also very much in the anti-spoilers camp. I know not everybody has the same viewpoint, but I am hardcore about it. The downside, though, is you lose the flexibility of entirely skipping a worthless game.

I can imagine a service that uses a combination of machine-learning and user-configuration parameters to notify a user if a given game is worth watching on the DVR. The user could set various parameters such as:

• My team wins (I personally don't include this one, but some people do).
• The score is close at the end.
• The score is close most of the game.
• The game involves a reasonable amount of scoring.
• The scoring is not all on penalty kicks (I'm looking at you, futbol) or home runs (baseball).

Also, the result wouldn't be a pure yes-no, although that would be an option. There would be an overall rating underneath the yes-no. So if you are me, a spoiler hater, you might want just the yes-no, because the score could give something away. But if you are on the fence or particularly busy, maybe you want an indication if it is a really good game, not just good enough to be a generally enjoyable watch.

Then there are all kinds of ways you could incorporate statistical and machine-learning elements, to adjust the rating. Just scratching the surface:

• Higher rating if many other people with profiles similar to yours rated the game highly.
• Lower the rating if you identify that you accidentally found out the result (lower still if you know the score).
• Use ML to adjust the rating based on your own ratings of past games.

Then also enhanced features such as telling you how far to skip ahead, or to skip to the last few minutes.

Email feature Flag R for reply expected

Sometimes when you send an email, you definitely expect a reply. But people's email habits being what they are, said reply does not always arrive. Google actually tries to use AI to remind you of emails that seem like they need a reply but haven't. That's nice, as far as it goes. I would like email to have a feature where you can set a "Reply Expected" flag.

Some E-Commerce Apps Should Be More Careful About Default Stores

Over the years, I have had a few cases where online apps wind up placing my order at the wrong location. I can't universally rule out user error on my part, but I have some clear-cut cases where the app messed up. In one case (can't remember now which app), when I identified my use case as In-Store Pickup, and an item was not available in my default store, but was available in a nearby store, it auto-switched my order to the nearby store. 

Recently with Panera, the app had a flat-out bug.

Either way, I wonder if more thoughtful preference configuration could help. What jumps to mind is asking the customer to identify which use case best describes their needs:

1. I almost always order for the same location. If an order is ever different than my default location, please ask me to confirm.

1. I order for many different locations. Do not ask me to confirm when the location changes or is new.

For extra-credit, AI could probably be brought to bear, to refine the options, or to default them.

Google Calendar Feature Idea: Filter Recurring

 I tend to use Google calendar for a lot of recurring reminders (renew drivers license, perform monthly maintenance tasks, etc, etc). I would like a UI affordance to toggle my view to show All - Recurring Only - Non-Recurring Only.

Central Mexico Trip Report

My wife, adult son and I just returned from a teriffic 11-day vacation in Mexico. Not beach Mexico, but historic, central Mexico. 6 days in Mexico City, 5 days in Morelia--an old, colonial city 4 hours west of Mexico City. We would never have chosen this vacation, if not for the opportunity to visit old friends, who were living in Morelia for 4 months, while on university sabbatical. But are we ever glad we discovered the delights of central Mexico!

First thing nobody told me about is the climate. I vaguely knew Mexico City was at a high elevation, but I never put two and two together. I dumbly looked at a map, thought to myself southern Texas is really hot, Mexico City is way south of that, it must be a blazing inferno. What a misconception! The mile-high elevation means it is very temperate, all year round. Think Hawaii or San Diego, except maybe a bit more comfortable--75-78F and very dry.

The next recalibration I had to make was the Mexico City environment. I absorbed the idea, 25+ years ago, that Mexico City air quality was horrendous. I think that was true, but like LA, it has improved a lot. It isn't great, but it also isn't a smoggy hell-hole. Most days we were there the weather app rated it "acceptable".

It was also very clean for a large city. The various street vendors were always sweeping up in their downtime. Compares favorably to US cities.

Now on to the question of security and safety. We have all read the chilling tales of drug-cartel violence haunting Mexico. It is real, but it is confined to certain areas, mostly near the border. Mexico City and Morelia both seemed very safe. I am sure there are dangerous areas of Mexico City, just as there are in Chicago, but no reason to go near them as a tourist.

Cost-of-living is very favorable for rich-country tourists. We stayed in the expensive part of the biggest city in Mexico, and overall costs were maybe one-third what they would be in the US (granted, the current exchange rate of nearly 20:1 is historically favorable). 

We spoke zero Spanish, and unquestionably stood out as gringos (being of mostly northern European ancestry, and some of us being pretty tall). Nevertheless, the people were very friendly, especially by megacity standards. 

Finding English-speaking service workers was spotty, a bit more than I might have guessed. Google Translate was indispensable. Get familiar with it before you leave, and download the Spanish language pack for offline translation.

We did not even think of renting a car. In Mexico City, like Manhattan, a car is the last thing you want. Uber is as reliable as in the US, and costs maybe one-third as much. For traveling between cities, there are extremely nice first-class buses. A four-hour trip to Morelia cost $33, and the bus was shiny, new and very comfortable.

Credit cards are widely accepted. Not quite as universally as current day in the US, maybe the way it was 10 years ago (i.e., cabs and street vendors don't tend to have Stripe swipers). So you will want to carry some cash. Including 5 pesos for public toilets.

Also, on the subject of currency, be aware they use the $ symbol to denote cost in Mexican pesos. Given the exchange rate while we were there, that meant dividing by ~20.

Bottom-line, central Mexico is a vastly under-rated tourist destination. It would also be an amazing place to snowbird. Or if so inclined, even to fully retire to.

Induction ranges should have one conventional electrical burner

Induction burners are the future of stovetops. One thing that causes consumer resistance is the concern that all your cookware has to be replaced. In many cases, that concern is overblown. But to ease the transition, it could be helpful to equip the induction stove with one conventional burner.

I guess one downside is that consumers will become accustomed to the cooktop instantly cooling, and may have a mishap with the conventional burner that remains hot (hopefully not actually touching it, but maybe placing a heat-sensitive item over it).

Cut tree stumps off below ground?

Is it really worth the cost of grinding away a stump? Can't you just cut it off good and low, and then do something creative with it?

Idiomwatch: Begs the Question

In modern usage, "begs the question" is "misused" more often than used in its correct/traditional meaning. Until recently, I was among the guilty. I, too, believed it means "begs (for) the question (to be posed)." I came to know that was incorrect (though I still think it has a lot of merit and utility). But I still did not realize the correct definition. I thought the real definition was "begs (off) the question", though I see now that doesn't quite fit, since begging off is active, versus passively trying to avoid the question. Wikipedia does note this usage.

The Wikipedia examples of the logical fallacy are not usages I commonly see.

Reminder Feature for Google Calendar

When receiving notifications (especially email notifications), include an option for "terminate any more notifications for this event". 

When I was in my teens, I picked up the idea of  "triple redundancy" as the NASA standard for fail-safe systems. The idea being, your backup also has a backup, so a catastrophic failure would require 3 (hopefully) independent things to fail at the same time. Without litigating whether this was actually a NASA design principle, or whether a more apt description is "triple failure, double redundancy, I can say the concept of redundancy made an impression on my youthful self.

Think about losing or misplacing your car keys. If you lose track of your primary key one day a month (1/30), adding a backup key takes that to once in 3 years (1/30^2). Add another backup, and you are at something like once over an entire lifetime (1/30^3 = 1/27000). And that is ignoring the fact that, once down to your second backup, you are probably going to be really, really careful.

Anyway, this week I experienced a rare quadruple redundancy failure--this time related to housekeys, not car keys.

We are having our siding replaced, so the garage keypad was removed (and resting inside the garage, where the workers left it). I idly noted that, as I walked out the garage, and walked to my car. What was I getting in the car to do? I was going to drop it off for service, and bike home. So that took out another source of access, when I returned home (via bicycle--that is how I minimize the pain of vehicle drop-offs)--no garage door remote. 

But I have never been one to rely on the garage for access, since if the power goes out (relatively more likely), or the opening system has a problem (even less likely), you are SOL. I keep a housekey on our keychains. BUT, we have a new car, and I haven't installed the housekey yet.

Still, I wasn't worried, because I also have a Tailwind smart garage door opener. I can easily open the garage door from my phone. Except it was this night my Tailwind chose to fail me (I think because of a required update). So had my wife not been at home, I would have experienced 4 system failures, resulting in lockout.

(I have since placed the keypad on the back deck, as a temporary backup, until the siding work is done.)

Inflation's Distorting Effects on Taxation of Interest

We've been in the era of 2% inflation so long that many of the distortions it causes have faded from memory. While reading up on iBonds, I remembered one of them: taxes are assessed against total returns, not real returns.

High-inflation example: 

• Inflation is 8%
• Your aggregate marginal tax rate is 32%
• You earn 10% interest (2 points above inflation)

So your after-tax yield is 6.8%--somewhat below inflation, so you are not even quite preserving your capital.

Low-inflation example:

• Inflation is 2%
• Same aggregate marginal tax rate of 32%
• You earn 4% (2 points above inflation)

Your after-tax yield 2.76%--somewhat above inflation.

Ordering Wine online Vivino

TL;DR

• I placed 2 orders to Vivino, for 8 bottles per order, each a mix of 2 different wines.
• There were under-$20 wines reviewed in the New York Times, and hard/impossible to find in St. Paul, MN.
• The retail price was competitive with the price listed in the article. Shipping was free. One order evaded sales tax.
• The delivery arrived promptly and was fully satisfactory.
• My cursory investigation suggests that buying directly from the distributor, to cut out Vivino's "middleman" profits, is not an option.

My (Limited) Experience

I've used Vivino (the wine rating/discovery app) lightly for a few years. I am a wine-lover but a dabbler.  I am not building a cellar and have a modest budget ($15-20 a bottle), so I really haven't gotten that much value out of the app, in terms of curation and discovery (and like many others, I consider the ratings nearly worthless, being crowd-sourced and utterly un-calibrated).

I finally got around to using Vivino to actually order wine, and I think that might wind up being my most valuable Vivino use case. I will frequently read some article in the general press--sometimes New York Times, sometimes the local paper--reviewing and recommending wines in my price range. I'm often curious to try what is recommended. The problem is finding it. The NYT stuff is almost impossible to find locally. And even the local stuff may only be available in a wine shop across town. That's way too much friction.

So recently I ran though a NYT article with recommendations. Of the 10 I looked for, I successfully found 4 on Vivino. Not bad. The prices were competitive, if not especially a bargain--never less, sometimes 5-10% more, than mentioned in the article.

Where Vivino shines is reasonable shipping cost. It appears that for < 9 bottles, they charge $15 flat rate. That actually isn't bad, for say 6 bottles. But it seems like there is a tipping point, I believe at the odd number of 9, where shipping becomes free. That makes Vivino a very convenient, price-competitive way to try and obtain wines that are not locally available.

And here's the thing--while Vivino doesn't support pure mix-and-match to get free shipping, it also doesn't have to be 9 bottles of the same wine. The way it seems to work is that Vivino has relationships with a number distributors. I don't know what that number is, exactly, but the same few kept coming up in my query vs the NYT list.

I don't have enough data to know if each distributor has their own requirements, but the two that fulfilled my order both seemed to set 9 as a breakpoint. As far as I could tell, any 9 bottles from that shipper could be combined to get the good price. I split my orders 4-5.

Both my shipments arrived promptly enough--4-5 business days--and were accurate. Very well-packed.

Comments About Vivino on Reddit

On Reddit, I have seen comments that put Vivino in the same category as DoorDash, and claim that they take a 15-20% cut. Given the competitive prices, I don't see how there is room for that much margin. But as a consumer, I am sympathetic to the idea that the aggregator should not eat the lion's share of the profits; I make sure to subscribe to anything I can directly, not through the Apple store.

So what I found interesting was that Vivino was absolutely transparent about who the distributor was. I looked up one of them on the web (Corkery Wine & Spirits). The price on their website was the same as Vivino, but then I noticed that they warned they only ship to 3 states (MN not one of them). I found this very curious, so I called them. I explained my situation, and they confirmed they only ship to those 3 states under their own name, but through Vivino they can ship elsewhere. I didn't try asking for a fuller explanation.

(The distributor for one of my two orders didn't charge tax for my state, giving Vivino an additional cost edge. I do not feel that this kind of online tax avoidance evasion is ethical or fair, so I don't consider this a pro for Vivino, but others may feel differently.)

Vivino vs Wine.com

Some Redditors speak more favorably of Wine.com, since they maintain inventory, rather than acting purely as a middleman. From the perspective of supporting your local wine ship, I don't see how this makes a difference. From the purpose of economic efficiency, I believe--for a highly fragmented market such as fine wine--maintaining inventory is unlikely to deliver net economic efficiency, and is instead a source of overhead and inventory limitations. The longstanding dream of online wine sales, after all, is to be able to buy any wine you hear of, for a price comparable to the normal brick-and-mortar retail price that prevails in its traditional physical distribution footprint.

In my limited experience with Wine.com, I found their inventory hit-and-miss, with a roughly similar hit-rate to Vivino (keep in mind, sample size with both is very small) . But I found their prices consistently on the high side--15-20% more than the typical retail price.

And that is before shipping. I haven't used Wine.com in a few years, so I did some quick investigation. It looks like their shipping price for 6 bottles is about $4.50 per--nearly double Vivino's $2.50. At 12 bottles, the per-bottle price decreases to about $3.25. It seems to very slowly decrease with volume from there. But clearly they are going for the Amazon Prime approach, by offering a $60 annual fee to get free shipping. Which is not a bad deal. 

Venmo Feature: Advanced Functionality for Splitting the Bill

Venmo has a split-the-bill feature, where the user can assign payment amounts. It's a good start. But a common use case is multiple people go to a restaurant, one person pays, and the payor doesn't know who ordered what. The basic functionality I would like to see:

• Attach a copy of the bill.
• Enter a total amount, invoke the split request.
• Each user examines the bill, calculates their share of the base bill. Venmo apportions tax & tip.
• Venmo keeps track of the total reimbursements received.
• Venmo provides reminders for parties who haven't responded.

Advanced functionality:

• If total is off by more than a set amount (e.g., 10%, above or below), Venmo notifies all users.
• Venmo scans receipt, and lets users click on their line-items.

Email Is Better than Texting

 

(Mainstream) Reasons email is better

Option to use a full-size monitor and real keyboard (yes I know Mac users)

Much easier to bring someone into the conversation mid-stream

Subjects are helpful

Vastly better search

(Mainstream) Reasons texting is better

Many people have developed bad habits about email, and tend to ignore/postpone 

I. Hate. Organized. Gambling.

Alarming and depressing article on the explosive arrival of app-based sports betting. I hate organized gambling. I hate it intellectually--it is the triumph of hope over math. I hate it aesthetically--the trappings of gambling, from garish smoke-filled casinos to grimy OTBs to convenience store lottery lines--is soul-crushing. I hate its social effects. And now, also hate it for overshadowing the sporting events themselves.

Keyless Entry Strategy for iOS or Android

I recently installed a Tailwind iQ3 smart garage door opener. It is very impressive, pretty simple to install, and downright cheap at $60. Just the fact that it will alert you if the door is left open at night, and if desired, will automatically close the door, if probably worth the price.

It also eliminates the need for a keypad. In my experience, keypads are a hassle and a bit of security risk. Because the code is typically very hard to change, people don't change it often, or use the temporary code feature. 

You can open the door remotely, which comes in handy in various situations, such as your neighbor wants to borrow some tool in the garage when you are away on vacation. Even better, you can assign opening privileges to other Tailwind app users. Great for guests, or families with more than 2-3 garage users, who would otherwise need to obtain extra garage door modules for their car.

Which brings me to my main point. While the sharing privileges feature is great, it comes with the substantial friction of each guest user having to install and provision the Tailwind app. It would be SO much better if this were nearly frictionless--i.e., if it were built into the mobile OS.

(Note: I know that a garage-oriented view of home access is a very suburb-centric viewpoint. There is a clear analog with smart door locks, more on that in a moment.)

Apple or Google should acquire Tailwind, and partner with the lock industry, to build keyless entry access-sharing into the OS. 

Apple is probably the more obvious candidate, at least in the US. If this were an Apple-only feature (think iMessage), it provides a distinct source of competitive advantage and lock-in through network-effect. Given that households in the US that have garage doors, and can afford smart locks skews upper-income, this fits well with Apple's customer base. Moreover, Apple's good image regarding security in general should transfer well to this use case.

I think the immediate first step is acquiring Tailwind (who I believe is the market leader, certainly the functional leader), and making it Apple-only going forward. The smart garage door market is new enough that I think that a play for total dominance is realistic.

The door lock market is much more established and fragmented. It also has a heavy decorative dimension. So I am doubtful that a total domination play is viable. Instead, Apple could move quickly, and leverage domination of the smart garage market, to establish its standard, open for adoption by heterogenous door lock manufacturers.  

******************

NOTES

As a consumer, I would ideally prefer to see an industry standard. But from the perspective of business strategy, it seems like a great opportunity for Apple. Also, I am doubtful about the standards-based approach happening any time soon.

A consumers should never rely solely on anything electrical, let alone "smart", for critical access. I cringe when I talk to someone whose only access to their house if via the garage door opener. So always carry a housekey--that applies to smart locks as well. But for many use cases, you can tolerate the pretty low risk of an electrical outage. E.g., if you have a weekly housecleaner for 10 years, and there is a 25% probability that once over those 10 years, the power will be out, and they won't be able to get in--that doesn't seem like a crisis (unless of course Murphy's Law strikes, and it happens to be the day before you are planning a big house party!).

My Original Covid "Modeling"

I recently came across some back-of-the-envelope modeling I did early in the Covid ordeal (05/18/2020). Conclusion was that the duration of "Flatten the Curve" was a theoretical minimum of 214 days. I didn't even bother posting it at the time, because it seemed to depressing.

The core assumption underlying the model was that a large percentage of the population was going to get Covid, it was only a question of when (my working figure was 60%). Then the next key metric requiring an assumption was--what proportion of Covid patients would require hospitalization? My guess at the time was 10%. This turned out to be way too high (by a factor ~5X), although per this paper, I think that was a figure that was circulating at the time.

Anyway, from there it was just a matter of calculating how many "inventory turns" of hospital beds would be required to cycle the entire Covid population through, and how long an inventory turn required (on average). My guess for duration (5 days), turned out to be very low, so partially offsets the high guess for proportion of population infected.

Keep in mind that 214 days is the theoretical best. It assumes the rate of Covid sickness is perfectly optimized, never exceeding hospital capacity (which would imply excess deaths), but never leaving any slack (to ensure the shortest possible duration).

Optimization is always hard, but optimizing rate of acquisition of a novel disease is impossible. So the real-world outcome probably would have been both excess deaths, and still a duration of longer than 214 days. That is, if the 10% of Covid patients require hospitalization assumption had been accurate. Thankfully that was not the case.

0.0028	Hospital beds per capita
0.5	% devoted to Covid patients
0.0014	Steady-state hospital beds
0.60	proportion of population who gets Covid
0.10	proportion of sick requiring hospitalization
0.06	Proportion of population requiring hospitalization at some point
42.9	"Turns" of beds, to cover entire hospitalized population
5	Duration of hospitalization
214	Total days required to work through entire population, if curve flattening is perfectly optimal.