I'm like, how is that your recommendation? To your point, there are still some real risks here that are not being mitigated.

Well, I mean, look, it's outstanding technology because it's such...

And I do think that a lot of it boils down to a lack of understanding of where it should be placed.

So there's heaps of...

That's why I said I try to be as...

It's...

Less deliberate than it is.

I don't know if that's the best way, but I would love to hear how you would advise to consider choosing the right model for the right task and then also any token efficiency.