The next quarter of effort at the frontier of applied AI is not about model capabilities. It's about 3 key things:

1. security, guardrails around data ingress/egress, and understanding risk in your system

2. balancing permissions with the right level of system autonomy and self-learning

3. utilizing local sandboxes/machines properly.

It doesn't matter how capable the base models are becoming. If you can't crack those issues, your solution is a demo, not an enterprise solution