Most people are focused on bigger models.
I think the bigger opportunity is better agent infrastructure.
That’s why IronClaw caught my attention. The value isn’t in making agents “smarter”, it’s in making them more dependable, composable, and useful in real-world workflows.