“Reviews, refactoring, and one-off scripts are consistently useful. Well worth the money for that alone.”
A working programmer’s honest field notes, not another breathless thread. The models are genuinely good at review and mechanical grunt work. They fall apart the moment a task needs judgment or real verification. Useful as a tool under a human, useless as a replacement for one. That gap is the whole story, and it is the part the hype machine keeps skipping.