Should You Break Up With Alexa?

Your Alexa is in the news lately, and it’s not because the device has suddenly become good for something more than setting timers and turning smart bulbs on and off. Your Amazon Echo devices that are powered by Alexa are getting a souped up AI mode called Alexa+ later this year, but they’re disabling the feature that allowed users to specify that all queries should run locally at the same time. After March 28th, all queries on these devices will now be sent to Amazon’s cloud for processing (and storage and future retrieval by Amazon employees or nefarious hackers).

The local processing feature was only available on relatively new hardware, the Echo (4th gen), Echo Show 10, and Echo Show 15, so if you were using any other Alexa devices in your home, they were already sending all of your queries to Amazon’s servers.

Yes, it’s bad that Amazon is pulling an advertised and supported feature from existing hardware and they should have to refund anyone who bought one of these devices. But for the vast majority of people using Alexa devices, this is no change. It’s exactly how Alexa has worked from its launch in 2014 until these newer devices rolled out.

The more important question is: “Should you have ever put one of these devices in your house in the first place?”

When Amazon rolled out the Alexa, way back in the heady days of 2014, I was captivated by the little black cylindrical computer that you could talk to and kind of get to do useful stuff, when everything worked right. Amazon played up the Star Trek library computer vibe of the thing, and it was a fun, if mostly useless, way to interact with a computer.

I enjoyed playing 20 questions with the Akinator and I used it to listen to music in the rooms where I didn’t have a Sonos setup. Eventually I connected Alexa to my smart home hub, and it quickly supplanted the tetchy web of sensors I was using to turn lights on and off automatically in my home. Has it ever been easier to say “Alexa turn on the bedroom lights” than it was to flip a switch? Maybe when I had both hands full of a squirmy baby, but not since then. But that’s where it stopped. In 2025, I use voice agents mostly the same as I did in 2015, for music playback, setting timers when I’m cooking or doing chores, and controlling smart home stuff.

Why do we keep these things around?

At a certain level, we’ve become inured to the high level of surveillance that we experience in the service of selling us new shoes and underwear. Do an idle search or spend too many seconds looking at a paid influencer spot on Instagram, and suddenly every ad you see will be for the advertised product. Whatever you do, don’t click on this affiliate link for a 40 pound bag of dehydrated cereal marshmallows, or you’ll start seeing marshmallow ads everywhere.

In the beginning, I felt OK with these things because the promise has always been “It detects the wake word on the device and then sends everything after that to the cloud”. And while there were going to be some errors with that system, sometimes you take the good with the bad, right?

My thoughts on this in the beginning were that we should assume best intent—that is, believe that these devices are actually working the way they claim to be. But then Amazon had a series of problems with data privacy—they stored recordings of kids indefinitely, they let employees and contractors access data from Ring cameras inside people’s houses, they handed user data over to the police without requiring warrants for a decade, they pay employees to listen to recordings to make the service work better, rampant allegations of improper data security, and the list goes on.

I got rid of my Alexa devices a few years ago¹ , but I kept speakers from Google in the house, and still have Siri turned on my Apple devices. Given the rise of authoritarianism in the United States, I’m not sure having these speakers around is such a good idea. When people are being turned away at the border, seemingly because they criticized the current administration, having an always-on, digital snitch connected to servers that are accessible to the US government seems iffy at best.

What are the other alternatives?

Now that it seems like getting rid of all of the corporate-controlled voice assistants is probably the right choice, I’d hate to lose the convenience of voice control for things like kitchen timers, music playback, and smart home control. There are a lot of options from a variety of companies, but as is often the case, the Open Source folks are here to save the day.

Home Assistant has a voice module that you can use to replace some of the functionality of an Alexa, Google Home, or Siri with, and run it locally (or in Home Assistant’s Nabu Casa cloud instances). From the reviews I’ve read, it’s not a one-to-one upgrade, the audio quality on the HA Voice speaker is such that you won’t use it for music, and it lacks access to the deep information graph that Google, Amazon, and Apple’s voice assistants can tap into by default.

In true Home Assistant fashion, they’ll let you replace any of the three core components—speech-to-text, the conversation agent that parses your voice commands and infers your intent from them, and text-to-speech with pretty much anything you’d like. If you want to put a giant NVIDIA GPU in a box in your garage and run a local model as the conversation agent, they’ll let you do that.

If you want to hook it up to a cloud-based model from ChatGPT or Google Gemini you can do that too. And of course, it’s much easier and cheaper to get the cloud-based solutions running than to do the same thing locally. (Please don’t do the work to ditch Google or Amazon’s voice assistant and then run ChatGPT in its place, that completely defeats the purpose.)

The unfortunate reality is that running these kinds of complicated, interconnected services locally is difficult. There are a lot of moving pieces, it’s likely to be fiddly, and some stuff just won’t work the same way it does with the commercial solutions. There are other contenders in this space, using varying levels of Open Sourced-ness, but the only ones that are turnkey have the same (or worse) potential privacy problems as the solutions from Apple, Google, and Amazon.

Or you could just keep using Siri and hope that Apple doesn’t roll over for the US government the same way they just did for the UK government.

What’s Next?

I’ve been at the Game Developer’s Conference this week and have had a bunch of lovely chats with people from around the world who work in all aspects of game dev. But after a week of non-stop days and nights, my social battery is empty and I’m looking forward to coming home this weekend and spending some quality time with some of the games I saw. I talked about some of the different things I saw on this week’s episodes of The Full Nerd and the Tech Pod (The Full Nerd is up now and this week’s Tech Pod will drop early Sunday morning).

I’ve also been enjoying Om Malik and Fred Vogelstein’s newsletter, Crazy Stupid Tech. Om and Fred have both covered emerging tech for investors for a long time and they have a good track record of focusing on the products and teams that end up making an actual positive impact for people, not just investors and shareholders. Om’s latest post about cheap Chinese IoT devices that phone home has left me rethinking the devices I want to give the ability sniff around in my private subnets.

1  Technically, they’re aging in the garage until they’re ripe enough for e-waste