Post

I will say, the initial tests it wrote did require some finessing to get working properly. I do not think AI like this is a replacement for senior devs, and I still think companies trying to replace junior devs with LLMs are short sighted. I manually review and massage every change Claude Code

Suggests and that's very easy to do with the command line tools they provide. Nothing goes into my work that I've not personally looked at, verified, and understand. And someone vibe coding does not have the skill set to do that, software engineers do. Tools like Claude are just prediction engines

And the need for high quality SWEs especially for projects dealing with sensitive areas like finance or cryptography, need that oversight. Furthermore, there is a human element to SWEs in understanding intent and empathizing with users that I don't think we would ever truly get in AI.

0 reposts0 likes

No replies yet.