GM just laid off IT workers to hire those with stronger AI skills

US car software, ya, I've never seen such trash in my life. This said I don't really have any complaints about my Hyundai software. It works, doesn't crash, and does what I want it too.

I kept reading about how bad Android Auto was for years but we finally bought a more modern used car and I can't believe they would ship that experience to customers. I had a week where I just had to unpair and re-pair Everytime I got in the car.

I would love to read about why that stuff is the way it is from the engineers, hmm that might be a good spelunking. I really must be missing something that makes it harder than I think it really should be.

Well, if all the training data for auto software was trash anyway, then you know what they say:

Carbage in, carbage out.

I used to have issues, I picked up an AA wireless dongle and adjusted the settings a little bit, and have had a pretty solid experience. YMMV

>automotive software is so bad I'd bet my life that Claude could do better

and they are!

> automotive software is so bad I'd bet my life that Claude could do better.

Leave it to Claude and you will.

The quality of software has overall decreased in the last decade. Testing is avoided, fixing bugs is done through redesigns.

I can't tell if this is sarcastic or not, but it seems insane to let the AI write the tests.

AI can't be held accountable, it shouldn't be writing the tests that determine whether car systems function correctly.

As someone who's worked on this kind of stuff at GM, I don't really get the exuberance in this particular space (not just the comment I'm responding to).

If you want 100% coverage, you just autogenerate the test cases. LLMs can't properly check MISRA requirements, so they're really just a layer on top existing automated checkers. Same for complexity metrics, it doesn't get merged if it violates that (or it's a vendor dependency you won't touch anyway).

If you care about the spirit of the rules, they're not that big a difference. If you don't care, there are already ways to do this. In either case they're an incremental change, not what I'd call a godsend.

Irony detected, but it's 2026, so nobody can be sure.

Yup thats exactly what they want

Cheaper younger people who dont think vibe coding is bad

Is this a good idea - probably not

Yea it's vibe/agentic coding. That's what about half of the current jobs are right now. It's really sad. Saw a vibe coder job today, 20bucks an hour

I mean agentic coding is a thing but anyone could learn that in a day or week. So the idea of throwing away people that could be the most productive with ai of course make no sense. But it's big corporations, not everything has to make sense. It's most likely a dressed up pure cost saving framed in 2026-lingo.

FWIW, I interpreted the article as saying they're not looking for vibe coding, but AI model development per se:

"...In practical terms, GM is looking for people who know how to build with AI from the ground up — designing the systems, training the models, and engineering the pipelines — not just use AI as a productivity tool."

You can just check any of the vacancies on their website to see what they're looking for.

Example: AI Agent Engineer (https://search-careers.gm.com/en/jobs/jr-202606937/ai-agent-...)

> We are seeking an AI Agent Engineer to design, build, and operationalize AI-powered agents that enhance employee productivity and decision-making in a complex enterprise environment. The ideal candidate combines strong AI/ML foundations, hands-on experience with agent frameworks, and a pragmatic approach to delivering business value in partnership with cross-functional teams.

Does not look like pure vibe-coding to me. More like developing wrappers over LLMs.

AI-ensbled engineers == any engineer without sufficient life experience to inherently tell that delegating huge swathes of implementation to a stochastic parrot is a dangerous and fundamentally unsafe idea. It's spun around "knowing how to use the tools" but somehow the whole part around "knowing when not to use them and refusing to use them in those contexts" is conveniently written out of the definition of AI enabled. I personally sed s_AI/ enabled/ engineers_engineers/ with/ unreasonable/ risk/ tolerance_g.

When the business field turns you into a pariah because you refuse to bow to the uninitiated's desires tends to be my red line for an exit. I will not do reasonably foreseeable net harmful work for an industry that won't take no for an answer. My definition for reasonably foreseeable sets the minimum bar for analysis of consequence at enumerating 3rd and 4th degree consequences minimum.

They don't want engineers. They want Operators for whom thinking isn't in the job description.

It's crazy that this still works for the companies. Anyone with any experience here knows whats happening. But the reporters can still just parrot whatever these companies say. Who cares if it's more lies, right?

They hire anew to lower salaries, while also gaining staff with "stronger AI skills."

Laid off ai workers are cheaper. I saw a job posting for a vibe coder, 20 bucks an hour.

I found that sort of thinking is no longer a part of corporate culture(in AU at least). As in, investing in your staff and planning for the future. Resources are meant to be used and abused for benefit and that's it, consequences be damned.

perhaps they tried that and the staff wasn't willing to be trained. have you ever worked at a company like that of that size?

Their tech positions were really the only common path to get continued raises and promotions starting from the bottom in GM for the past few decades. Most other positions only ever got hired from outside the company because internal hires expected raises for it, so to me this just looks like gm putting their tech department under the same self destructive hiring policies as the rest of the company.

I haven’t heard of a single software job in the last decade that offers anything resembling “training”. You sink or swim based on your previous knowledge, what you can glean from the codebase and coworkers, and how skilled you are at self-teaching.

A lot of people are un-trainable. I've worked with people who got angry when the Instructor showed up and the instructor wouldn't just give them the completed exercises, while most of the other people just sat on their laptops checking emails ignore the training all together. I've worked with people that want training on how to use Claude because they can't figure out "just type what you want in the box, and it does it for you." I've worked with people who were amazed when I googled for answers and were like "I should do that to!" (but didn't)

People will demand training, ignore it, and continue to be a drain on the company. There will be those people out there that have one very very very specific skill and that's all they want to do; "I remove people from Active directory that have names that start with A,B or C" or "I run this Ansible Playbook someone else wrote, and that's my entire job."

You won't be asked, your tax dollars will just be used to bail them out, either directly or through tariffs. Because democracy.

> And after seeing how they treat employees for decades

To say nothing of their cars.

That's the neat part, many don't need to worry about missing their kid's graduation because they can't afford neither kids nor the safety to have them (home ownership, etc).

"Cheaper" is indeed what you get when you reduce payroll.

This is only stupid if you value quality of outcome. If your sole metric is make stock price go up long enough to cash out this shit is brilliant.

Opens 50% of the time everytime

People who can use Claude and its peers effectively. AI skills = prompting and slop troubleshooting.

Why can’t US auto makers automate like China did?