[jotux]

I'm ~20 years into my career, currently working on a relatively complex C++ codebase for a space application. Internally we have copilot (gpt 4 or claude 3.7) and a custom ChatGPT (4o/o3) client we can use. It's great for autocomplete, boilerplate stuff, and generally learning about a new topic but terrible for complicated or niche development (just outright fabricates stuff that doesn't exist or compile).

We got a batch (10+) of summer interns this summer, bright students from good schools, and they are absolutely glued to the AI tools. They're getting lots of code written with lots of tasks complete, but occasionally they write code that makes absolutely no sense and can't explain why they did it. Lots of very overcomplicated solutions to problems, and seemingly low comprehension of the code they're creating. It actually makes me feel pretty good about career viability long term -- people that have a deep understanding of the systems and code will need to be around to clean up the mess that is about to be created over the next 5-10 years.

[HisoMorow]

thats an actual hopeful answer cause it makes sense and its closer to what i ve been thinking it might be like... i d say that the real threat though is in the future cause AI will be do more stuff and better, hard to stick to predictions the further we go in the future though