FR version is available. Content is displayed in original English for accuracy.
Advertisement
Advertisement
⚡ Community Insights
Discussion Sentiment
53% Positive
Analyzed from 4925 words in the discussion.
Trending Topics
#software#car#don#auto#cars#updates#more#update#android#code

Discussion (140 Comments)Read Original on HackerNews
1. You write release notes thet contain technical details. Less than 1% of your customers understand them. More than 90% probably won't even care, let alone understand the document. And then there are the folks who get confused or scared and reach out to customer support with weird questions. This generates extra workload.
2. You explain nothing. The release simply is. The technically minded people are mildly annoyed. A few customers affected by open issues wonder if it's fixed now. The rest of them doesn't even care that there is an uodate and carries on with their lives. Customer service continues to complain about the usual bunch of random and weird customer issues.
It's quite natural to start doing (2) in a consumer facing business, isn't it?
The costs will be recouped by sacking more developers...
Absolutely.
But also don't break my stuff.
Or businesses would close to do inventory.
We made do.
Not my own device on my own machine.
I don't remember a time my CD player on my car would stop working due to an update.
"Bug fixes and performance improvements "
Even worse than the "reformat" commit message that your bisect landed on.
Today there’s no disc, no recall - that cost to shipping broken software is gone. We the users pay the price.
Arguably the peak quality after a few months/years of fixes is better, but that doesn't really matter because usage is front-loaded
The disks were very expensive and if you didn’t update them it wouldn’t know about new roads (slower trip, missed turns, etc).
The article is a lovely cathartic rant against agile software development methodologies applied in the wrong place in the wrong way, whether or not the software(s) in question used such methods. On of the worst assumptions, I believe, is that the end-user is willing and able to function as testing/QA without detriment to the product and company.
See also self-checkout etc.
To exit a climate control modal on the screen you have to find and tap a tiny red "X" box in the furthest corner of the screen from the steering wheel.
I've driven the 2025 and the 2024. The 2024 not only had all the crappy UI, but the driver assistance features were also alerting you constantly, and they were terrible. I was amazed how much they toned it back and improved it with the 2025.
Still wouldn't buy either of them personally. The constant nags and alerts are so fucking annoying and distracting. The seat belt chimes that don't shut off and get louder and louder make me want to rage so hard. I am religious about wearing my seatbelt, so there's absolutely no reason for it to piss me off so bad.
While that particular issue isn't solvable, I am open to any advice on coding tools that might allow one to unlock other settings or make changes like "ensure auto-stop is fully disabled across restarts and drivers".
They are far better.
BYD reliability and support leaves European cars for dead.
Don't put discrete, isolated HW functions behind a SW powered screen. It's that simple.
It's practically impossible to test every permutation of code against every system. Maybe AI can help, but practically it'll just mean the software gets more complicated, with more features. And to top it all off, more and more features get regulated, so they have to be there. The rear-view camera requirement in particular, since you need a screen to see the output. And if you have a screen... well it's an already paid cost, so, might as well display other things too.
We should kill the reg.
I don't think they're saving that much by ditching them and going to SW.
They also didn't need updates (well, the Bluetooth module may have, but nothing else).
It definitely was nice not to have to worry whether the climate control may stop working because the radio was modified. Or because of any update.
As a driver, dumping everything into one SW system has significantly degraded my experience. What I gain ("Ooh, I can now use Waze on a bigger screen!") is minimal.
> The latter can be done with a CI pipeline for pennies worth of compute in a fraction of a working day.
I'd appreciate the point if they were successful at it. As it is, they're not. It's rare to find a non-buggy car.
Many cars have AC controls on the screen only.
Older cars (that still had touch sensitive displays), simply had separate HW modules and HW interfaces (buttons/switches) for the rest. You never had to worry that modifying/repairing/updating one would impact the other.
All the hardware pieces are installed at build time
As a licensed driver who resides in the Sonoran Desert, can you even imagine the horrific visions that just flashed before my eyes?
We often joke around here that wearing oven mitts is a good way to get our cars started in the late afternoons. It's not really a joke.
I personally have several pairs of gloves, and I never fail to don those gloves when I go out, whether I am walking, riding an e-Scooter, or driving, because even as a pedestrian we must touch so many metal objects that bask all day in the direct sunlight.
Heated steering wheels. What a world we live in today!
Man, when it's freezing outside, it's awesome. I wouldn't buy a car without it now.
So exactly like software licensing? Most apps nowadays don't even require a purchase to download. The download is free but you need to pay $4.99/month subscription to use, or $99.99 for a "lifetime subscription". The code's are all there. The author just doesn't want you to use it.
This is something far more heinous, you bought a thing for a lot of money and just in order to extort even more money from you, they simply disable/lock away a feature that you technically already possess.
A better analogy in software might be that you bought a video game for $60,000 and the only way to beat a lower level boss without spending 2,000 hours trying to, is to pay the developers another $5,000 for a super weapon.
You'd prefer they get nothing for the effort they put into developing the software?
I've been lately into mobile apps and i am finding that there is no system which combines these 3
1. AOT 2. JIT (for hot paths) 3. Interpreter for non JIT paths or where you explicitly do not want jit.
Imagine, a system which compiles your app to AOT but when you push OTA update, part of the app are selectively replaced to JIT or Interpreted mode.
it's theoretically possible but nobody seems to be doing it. I found react native / expo eas update but i don't think it's like this, it has a Hermes VM which runs bytecode but it has no JIT so you'll write native code for hot path then you'll need to upload a full update to Android. So, only toy level code performance can be can actually be written in JS?
Much better, patch the parts where AOT calls into JIT or interpreter.
Currently i am using react native and flutter. Flutter's UI framework code is in Dart if you load this whole code into JIT, it will consume a lot of resources on mobile device as the framework is big and does lot of work. If all framework code was AOT and your custom patchable code also comes with AOT but upon OTA replaced by JIT or Interpreted code, crazy performance!
But what if we could run the most of the code in AOT and only run changed code in JIT or interpreted mode? arguably it would perform as good as it does not being complete AOT while also providing react native like fast updates.
AOT will be in base app and it will include JIT or Interpreted OTA updates.
For Apple, JIT can simply be disabled and OTA update can run patched part in interpreter.
But JIT works on Android (well), so this Hybrid system is capable of being much faster than React Native where your JS code only runs in Hermes VM which isn't JIT.
In this system, all your crazy math and algorithms (on hot path) stay easily updatable as this small part can be run on JIT with snapshot saved for next load if it has not changed!
JIT on average is 100x faster than dumb interpreter for language like Dart.
Does it already exist? otherwise i am seriously thinking about building something like this.
That said shipping JIT / interpreter with your program to recompile updates / parts of it sounds silly to me.
Which is never, unless something is broken.
Having rolling releases for a CAR is absolutely stupid.
Amen. If you go to a bakery, you expect them to care. If you hire a photographer, you expect them to care. Software isn’t (usually) a factory line; CRUD may be similar concepts throughout, but everyone is making it themselves.
Give a shit about what you make.
Rather than hamstring all software by requiring DOT testing before firmware updates are published, follow Tesla's model which has been very reliable within the industry
Please don't take this as suggesting any of them are good at software, mind you.
Let's face it, this really is about Tesla, vs the rest of the major players (ford, kia, VW etc)
(This software building code is more necessary for software used in critical infrastructure. But it should also be applied to consumer devices as basic protection for consumers against manufacturers breaking functionality the consumer paid for)
If Android Auto stopped working I'd also be livid because I don't use the built in crap.
An update which advertised, amongst other features, that it "rectifies errors and prevents security gaps" and stated "This update is recommended for everyone."
Borderline insane to refer to the user as "complicit" in that case.
Someone in auto industry decided that plugging device, and dependency on core functionality of the car to 3rd party device, that might be lost, have battery died, used for something else, etc is a good way to save money and not do proper software. It's even more bizare now, mid 2026, when software is solved with AI.
It's good that there are some companies, that ban android/apple car since that's an ugly experience for the user.
On the contrary, having cars stop trying to provide a bespoke more-proprietary outdated piece of software you have less control over, probably have surreptitious telemetry reporting back from, and might have to pay a subscription fee for, and instead just delegate to the smartphone you already have, is a huge and surprising win.
> It's good that there are some companies, that ban android/apple car since that's an ugly experience for the user.
It's a terrible user-hostile loss when cars do that, typically because they want to maintain more control or try to extract more revenue from the user.
If you don't want to use it, don't use it; there's nothing forcing you to do so.
there are companies with amazing software experience, Rivian, Tesla, Nio, Lucid, even gm is start moving into that direction, and WV is buying software from rivian.
I don't want an amazing software experience. I want an unsurprising experience, ideally the one I already have.
The only thing better than Android Auto would be to just provide a standardized port (and perhaps a wireless standard) for a combination of video output, audio output, touchscreen input, and charging, with optional standardized sensor inputs. Then you wouldn't need two different standards (Android Auto and Apple Carplay), just one, which would also work with any new device that came along to break that duopoly.
I own a Tesla, and a Ford. Amazing is not how I would describe the Tesla software experience. It lacks features like iMessage for group and for non-phone recipients that I am able to use in my Ford. Even though many people would say the Ford software is otherwise inferior. And if history is anything to go by, there are features in CarPlay today that Tesla will never add to their infotainment system.
Are you fucking serious? Tesla's head unit software is barely passable. It's shit.
Nearly half of the screen is taken by useless toy car depictions, and navigation can't even render the full street names because the width of the input field is fixed.
I'd agree if it worked.
Android Auto sucks. And I don't like that my auto manufacturer can wash their hands off it by pointing at Google.
> If you don't want to use it, don't use it; there's nothing forcing you to do so.
As long as the car manufacturer gives me basic functionality (radio, stereo, Bluetooth, etc). Nominally they do, but it sucks in a different way from Android Auto. So I have to ping pong between these two.
My prior car's aftermarket Bluetooth receiver was fantastic. The fact that I can't install something like that on modern cars is a huge regression.
The entire idea that everytime you sit in the car you need to pair your devices, what if you have several devices in the car etc ? it's such a horrible, broken, neurotic idea.
No, it doesn't. It's a very simple streaming protocol.
It's literally a gRPC-encapsulated stream of h264 frames over a USB connection. With touch events and some car-related telemetry streamed back. You can implement it in a weekend: https://github.com/mrmees/open-android-auto
You can create whatever you want, including just streaming videos onto the head unit or making it play Doom while driving (with steering wheel for input).
This isn't the auto industry deciding that you need to use your phone. On the contrary, GM and others tried hard to push back on Carplay and AA. This is the buyers telling the auto makers that they want Carplay and AA since they know that that actually works, and they know that the software the car actually comes with will be garbage, or at the very least unfamiliar and not really worth dealing with when you can hook up your phone and let that actually solve the problems the user wants to be solved.
It's insane to me that anyone could be of the opinion that it's good that some automakers ban/don't implement Carplay and AA. It's just taking away user choice. It's hard to believe anyone could have this opinion without either never having driven a modern car, or just being an industry plan.
They even fixed once a semi broken hardware for me. Camera power started acting up. I called tesla they said you can come to service to replace or wait a bit we will release OTA that will decrease a power consumption, in 3-4 weeks they fix my custom problem without going to serive
Uh? I can literally count the times my Model 3 2019 software broke something on one hand:
- when they redesigned the AC controls to make them more visually appealing but less functional (no button borders and no fill)
- when they decided to put air recirculation under auto-control and ignore the user's settings
- when they optimized the cellular connectivity and it took them a while to get back proper reconnect on loss of signal (garages etc)
- when they tuned sentry's sensitivity and there was some back-and-forth for a couple cycles between "record everything" and "record nothing"
None of this made the car undrivable or totally useless. I did hear of reports of early HW4 cars bricking their FSD computer, and Tesla replaced it.
In my opinion it's still a much better experience than the absolute guesswork of "what will my screen display today when I connect the phone? and where will I find Maps again?", based on software updates on the car AND the phone.
EDIT: also agreeing with the sibling comments: my 7 year old car got a lot of extra features since release, and most of them working very well at the first try.
Absolutely nonsensical. Both Android Auto and Apple CarPlay are better experiences than any first party car interface I’ve experienced.
In many ways the auto industry stumbled when they allowed this connectivity, just like phone networks stumbled when they let Apple dictate the iPhone from top to bottom. Good news is those stumbles worked out great for users. We get iPhones without bundled crapware apps and we get cars that don’t require monthly subscriptions for basic functionality your phone provides.
Presumably every car manufacturer can use AI. Yet there are still bugs. If all bugs are solved with AI, and therefore every car manufacturer with access to AI writes bug-free software, the only remaining conclusion is that some car manufacturers don’t have AI yet.
This suggests the supply of AI is too limited, and there isn’t enough AI to go around. Solution: build more AI data centres.
Why doesn't op simply ask AI to write software to fix his problem?
Hello, Elon
Seriously this is so wrong. I love being able to carry all my preferences from my phone directly to the car without any additional configuration. Before this, we had to do stupid stuff like entering individual contacts in the cars system.
The FAA has historically done a good job on this front with aviation (notwithstanding the 747 MAX debacle, which was a failure implicating flight control software but not a failure of separation of assurance levels). Automotive software and standards and NHTSA is far behind on this front. Cost is certainly a factor; it is not economical to apply aviation safety and engineering standards and processes directly to automobiles. But in the meantime, it is an absolute Wild West for any contemporary drive-by-wire car.