@nblr @bert_hubert By now, I 100% embrace LLMs. I use Claude Code daily. I love the new way to work.
I don't have to write mundane shit anymore that I copy from docs or Stack overflow anyway like a basic auth authentication for the rest API I am implementing right now.
I don't need to write the stupid CSS and HTML for the overview dashboard or login form of the service.

These are all tasks that don't even count towards programming for me anymore. They are annoying chores that I can now delegate to my never complaining and incredibly cheap coworker called Claude.

What I can concentrate on are the meetings with coworkers where we discuss and plan how we (re)design that (old or new) distributed set of micro services that searches through petabytes of malware samples to let analysts find if we have ever seen the shit they have on their table right now. We can concentrate on the real problems.

And even if I let an LLM generate core code for the project, I like to think that I am senior enough to recognize when it produces unmaintainable bullshit and then just correct it in a way it doesn't anymore. I've seen and done enough things in the last 20 years to judge what good and maintainable code looks like and what could bite me in the future.

The only thing is that you need to force yourself to learn that new programming language and get proficient at it no matter if you generate the bulk of the code written in it in the future. You can't stop learning ever. But that has always been the case in the IT landscape.

I'll be honest: I absolutely wouldn't want to be a Junior right now. Getting to the place I am right now took 20 years of grinding manually. Doing that and getting started while being replaced by a bunch of LLM agents is going to be hard. But then again, for half of the time I didn't even get paid to do that like a Junior wouldn't get paid now because they are just replaced by an LLM.

Also when we seniors meet and discuss a system architecture, why not include a junior as well? They can bring in ideas, concerns and thoughts about the system design and that stays no matter if using an LLM or not. They learn that way too.

For me it all boils down to: It's a tool. It's here. It will stay, even if the bubble bursts. The toothpaste left the tube. Other people are using it, so I need to use it as well. Maybe I am morally flexible enough to not give a fuck. For me it changed the way I work and I 100% like the new way more than the old one.

Edit: Another thing: The sunken costs of writing code are just gone. I can prototype a looot of stuff all day long and just try things out. It's shit? In the bin it goes! I only spent 1-2 hours on it and the wasted time and energy is negligible compared to the few days I had to spend to get familiar with that new framework or to flesh out that idea that turned out to be garbage before having code assistants.
I am trying so many new things right now and churn out code, it's wild. And I like to believe the code is fairly okay. That might be a skill issue and maybe I've always been a bad programmer in general, I will admit that.