Disclaimer: I care a lot about the problems being solved than the nth degree of how.
That said, although we make a ton of use of agentic development at work we’re still making the system design and architecture choices by hand.
There are certainly some team members who miss doing things like figuring out how to solve a puzzle, but we’ve had a ton more time to focus on system stability and performance.
I would argue that there’s a third class of engineering: the splashy YOLO. We have one of those. The boss’s boss loves the output, but the implemetaions are dog shit. We now have a personal atlassian token committed to the repo and said grand boss isn’t technical enough to see why that’s an issue.
CIA_chatbot@lemmy.world 2 weeks ago
I can add to this, I work in a space where cybersecurity compliance is mandated. I’m a software architect. Slop coded apps combined with AI literally brute forcing vulnerabilities leads to a shitty outcomes for everyone.
Two carpenters can build a table. One has uneven legs, isn’t stained or sealed, and has a crack in the middle. The other is measured properly, the joints and seams are well sanded, and the table has been oiled and sealed. Both “solve the problem”. Only one of them is actually reliable and useable. (And no, the AI didn’t make the good one)