For software development the latest models like opus are literally better at most coding then top tier humans.
Bullshit.
It may feel good to use these, but the data doesn't support that claim.
These things are putting out code that doesn't pass experienced programmer's definitions for quality or maintainability.
Where agentic AI runs in production, it is breaching data left and right and deleting production databases and sometimes also deleting the backups.
That you can get high quality code out of them, I won't argue against. I can, as well.
But these things aren't ready to replace programmers, and they actively make most programmers produce worse results in slightly longer timelines.
I'll tell you what is ready to replace programmers is fucking WordPress, lol. If you tell me WordPress is coming for my job, I won't argue.
Anything the AI can write perfectly could (and should) have just been built more maintainably on WordPress from existing plugins.
We're remixing excellent existing open source code, but we would be better off reading the documentation and using what already exists.