What's new
By:

Software Issues are Universal

G

Goodbye

Member

Joined
Nov 1, 2024
Messages
49
Reaction score
51
Cars
Tesla Model S
Air DE Number
0
I hear a ton of people complaining about software issues with their Lucid. I hate to say it, but that is just a sign of the times. We now live in an age where the culture of software engineering is to "just get it working as fast as possible". Systems are written poorly, then each fix is just a band-aid so that the engineer can mark his/her ticket/request done.

There is SO much poorly written code out there today that it's a miracle anything works. And it's only going to get worse as these bandaids start rearing their ugly head.

The old culture of writing solid, reusable, documented, and debuggable code is long gone. And if Lucid has fully embraced this new culture (whether intentional or by accident), it will be nearly impossible to turn it around.

Being in the software industry for 30 years, I've seen first hand the crappy code some people promote because of the pressure for a quick turnaround. It's bad out there right now.
 
Goodbye said:
I hear a ton of people complaining about software issues with their Lucid. I hate to say it, but that is just a sign of the times. We now live in an age where the culture of software engineering is to "just get it working as fast as possible". Systems are written poorly, then each fix is just a band-aid so that the engineer can mark his/her ticket/request done.

There is SO much poorly written code out there today that it's a miracle anything works. And it's only going to get worse as these bandaids start rearing their ugly head.

The old culture of writing solid, reusable, documented, and debuggable code is long gone. And if Lucid has fully embraced this new culture (whether intentional or by accident), it will be nearly impossible to turn it around.

Being in the software industry for 30 years, I've seen first hand the crappy code some people promote because of the pressure for a quick turnaround. It's bad out there right now.
Click to expand...
Software is hard even for the very experienced like Boeing who erroneously got the Starliner internal clock incorrectly in 2019 and failed to reach the space station.

NASA had to come in to make make sure Boeing changed the way how the software was done.
 
Goodbye said:
I hear a ton of people complaining about software issues with their Lucid. I hate to say it, but that is just a sign of the times. We now live in an age where the culture of software engineering is to "just get it working as fast as possible". Systems are written poorly, then each fix is just a band-aid so that the engineer can mark his/her ticket/request done.

There is SO much poorly written code out there today that it's a miracle anything works. And it's only going to get worse as these bandaids start rearing their ugly head.

The old culture of writing solid, reusable, documented, and debuggable code is long gone. And if Lucid has fully embraced this new culture (whether intentional or by accident), it will be nearly impossible to turn it around.

Being in the software industry for 30 years, I've seen first hand the crappy code some people promote because of the pressure for a quick turnaround. It's bad out there right now.
Click to expand...
And AI is only going to make this worse, as folks stop looking at their code entirely and just trust the machine to "vibe code" it.

AI's immaturity, combined with good old fashioned human laziness, is going to lead to a massive disaster or three in the coming years.
 
doubletwist said:
ALL software is faulty.
Click to expand...
This. It is essentially impossible[0] to write provably bug-free and provably secure software.

[0] Before anyone comes at me, yes, provably bug-free software exists, but the proofs are always relative to assumptions. You can have high confidence, not absolute certainty. Formal verification is expensive; often 5-10x the development cost. It requires specialized expertise doesn’t scale well with complexity. Most verified systems are small kernels (seL4), compilers (CompCert), or crypto primitives (s2n, HACL*), not full applications.

Moreover, when I say software is “provably correct,” I mean it matches a formal specification. But:
1. The specification itself might be wrong or incomplete; you’ve just pushed the bug up a level.
2. Proofs rely on models of hardware, compilers, and the environment. If those models are inaccurate, the proof doesn’t guarantee real-world behavior. For cars, this is essentially impossible to do accurately.
3. “Secure” requires a formal definition of your threat model. A system proven secure against one attacker may fall to side-channel attacks, hardware bugs, or threats you didn’t model.

This is the longest footnote I’ve ever written. Sorry.
 
joec said:
And AI is only going to make this worse, as folks stop looking at their code entirely and just trust the machine to "vibe code" it.

AI's immaturity, combined with good old fashioned human laziness, is going to lead to a massive disaster or three in the coming years.
Click to expand...
You’re right.

Also, I’m not entirely convinced we weren’t headed there anyway; remember the massive boon of coding bootcamp grads? There’s nothing wrong with that and I’m all for it, but Opus 4.5 legitimately produces better code than most bootcamp grads.

Someone else still has to read it though, which is why experienced engineers still have jobs.

What I worry about for the future isn’t that our AI overlords will be writing all the code, but that nobody experienced will be left to judge it, since there will be fewer and fewer people learning true computer science.

Anyway, a topic for another thread.
 
You must log in or register to reply here.
Back
Top