Hacker News Reader

Comment by pwthornton | Hacker News Reader

Tesla is a hardware company not known for stellar software. Both its self-driving/driver assist capabilities are suspect, and the functionality of the giant tablet UI could be better. You hear a lot of requests to be able to use Apple CarPlay in Teslas.

Teslas are probably better than the vast majority of car companies, particularly the UI of the touch controls and system, but it still seems weird to me to have engineers from a company perform a review in an area that they aren't known for. Honestly, I would assume Twitter has better software engineers.

View on HN

For those who don't remember, one example of just how badly they can fuck up their software, the operating system on Teslas for a number of years was based on a default Ubuntu that wrote logs which destroyed the SSD. Exceeded the total write wear leveling endurance after just 1-2 years of use of the car. Required total motherboard replacement on a great number of cars, because of course the ssd was permanently soldered down.

Yeah, this did happen to my S. It was covered out-of-warranty because the TSA said it would be a good idea if Tesla covered it out-of-warranty. Tesla concurred.

https://www.tesla.com/support/8gb-emmc-recall-frequently-ask...

A decent article on this: https://www.tomshardware.com/news/flash-memory-wear-killing-...

We have this example as a "don't do this" example for excessive logging in our engineering guidelines.

This was such a basic, newbie mistake that everyone I knew who had ever worked even minimally in embedded couldn't believe they were that incompetent and inexperienced.

Telecom systems and ISP people were also astonished, as it is quite normal for core switching and routing equipment to have a very limited write endurance solid state boot and OS storage device. For a long time very serious $300,000+ routers booted off compact flash cards, back in the day.

Yep. I worked on those myself and participated in the usual pre-DVT destructive testing of our embedded flash devices specifically to count core dumps and config commits that the device could survive..

“Better software engineers” to anyone who does software engineering is a joke. It’s an totally unquantifiable measure, and some people use dumb metrics like leetcode capabilities as some IQ test.

Sshhh, don't give them any ideas.

There absolutely _are_ better and worse software engineers.

There may not be an objective metric to discern a difference between two people of similar skill, but for sure it is easy to tell John Carmack from a junior developer with two years of experience.

Some Tesla engineers will be better than Twitter engineers and vice versa, but the Tesla engineers aren't there as a bastion of better software engineering.

They're a "trusted" external party that Elon can use to get an unbiased assessment of what's going on internally at twitter and get a better picture of what's valuable and what's not. If you ask a senior engineer at Twitter the same questions you'll get a very different set of answers, not necessarily wrong answers but a different perspective. Unless Elon's an idiot he'll be doing both.

I think they are there to figure out "the lay of the land". I don't think they'd be assessing code line by line. Just there to figure out how everything is connected, what the ongoing projects are, what is on fire etc. done by people Musk personally knows and presumably trusts.

Absolutely, I don’t view Tesla as a place top tier talent goes. Twitter much more so.

This seems ridiculous to me. Self driving cars are cutting edge tech. Social media algos are not for the most part, certainly not Twitters. Top talent goes to wherever cutting edge developments are. If I wanted to participate in developing the future I would choose Tesla over Twitter any day.

Top tier talent generally goes where they get compensated the best

This exactly!

That said, if you can't solve a random leetcode hard problem on a whiteboard in 40 mins, and you don't mind working 60+ hours a week, and you are still not a doofus, Tesla will probably hire you! :)

What self-driving car are you talking about? I’d bet Twitter has had more success in computational neuroscience with their structured tweets than Tesla has had with broken roads.

I bet they dont.

Twitter does some cutting edge ai research, they are unique (just like meta and google and some other firms) in that the challenges they face at the scale they do are unique, and need unique solutions. If you think "social media algos" are a) not cutting edge or b)all Twitter does technically, then you are incredibly misinformed. It's not just Twitter, Google and Meta are also doing incredible research in AI and are among the biggest contributers at any platform you choose (conference papers, libraries, datasets, architectures)

Reply on news.ycombinator.com