Windsurf and copilot are pretty useful for me

If you're set on doing something related with coding then probably look into low level / microcontroller development, otherwise I'd try looking into adjacent spheres not directly related to software development in the common understanding.

The thing is, we can dislike and make fun of the way current LLMs work all we want but at the end of the day they will replace a lot of jobs, especially if those aren't tied to maintaining critical systems. Nobody's going to hire real ui / web designers when GPT-10 can shit out something visually identical (and may not even store your whole database in frontend anymore) in 5 seconds. Nobody would want to keep paying a team of backend developers to work on a system for 6 months when Claude 8 writes the same system a little bit worse after a few prompts. So what if nobody knows what happens in that code it wrote, it's there and it works for the moment so it's gotta be good.

We're already seeing products that are pretty much entirely vibe coded and they're being launched when these "ai agents" may decide to just ignore any orders and purge your prod database. They will eventually stop doing that (probably, some time), so the number of such products will only keep increasing to a point where companies will seriously consider outsourcing most if not all of their development to LLMs.
While the industry may not be able to completely get itself rid of human developers (especially at senior positions), it will likely run itself into a weird state of having no junior / mid developers that can grow into future seniors in the next decade because all those jobs were replaced by LLMs because it was cheaper that way for the moment.
So yeah, if you're studying CS now or planning to get into it I'd say there's a good risk of having a lot of trouble searching for a job once you get that degree. I hope I'm wrong but that's what I feel we're heading into even if (after?) current ai bubble bursts.

LLMs are really useful in development, but I kinda hate the fact that it started to get misused in companies that should have fences on code quality. In my previous company our managers start to push devs to use LLMs and even bought license for each of us to the point that the obsession of speed feature delivery over quality. An old codebase with a decent best practices applied for many years. Got jumbled into oblivion because the AI Agent decide to remove some of the key vital functions and follow its own 'best practices'. The result was the mess of blaming each other for pushing garbage code that looks 'good enough' and AI inveting functions that doesn't take all edge cases which should have taken by an existing library that is battle tested, taking the precedence of 'AI Code' is better is kinda damaging existing codebases. What supposed to be a fast iteration become a month long cleaning mess. I don't want to look like the guy who hates AI, I do use AI but I check my code to make sure it really does work. The whole industry at the moment is like going back to fundamentals, clean code rants and crusade is back again because the code quality today is really bad knowing that Coder LLms is probably trained with vast repositories that may include some that low quality projects. Well I kinda agree with sir that, LLMs may get better but Its currently plateaued atm.

That's why as a test engineer my job is safe for now, ai writes shit code and I find out where the shit broke and tell other people to fix it

chatgpt is this code valid? write some tests that always pass to prove it works

Is game development safe sir?

But then the user complains when software doesn't work right, I save the user from pain

Copilot should be GODDAMN ILLEGAL! thats what

AI IS MADE BY THE DEVIL!

was it ever safe

You might need a holiday brother.