Really? Not challenging you, genuinely asking for more details.

If that is true, I think AI is counterproductive from the bean counter's standpoint.

Not counterproductive because people aren't just sitting back in the rest of the time while AI does work. They do more work. $3k per head on Claude is nothing if your devs get 2x more work done.

	▲	wg0 4 days ago \| parent [-]
		Do they review the code? Because in my experience using Claude Opus 4.6 generates code that would be buggy and the tests would be written agains that buggy code with wrong assumptions that certainly would pass with flying colors. It is only when you look closed you get to know what the hell has happened!