The interesting differentiation happening in the AI Agent track right now—many projects are still in the stage of papers and concepts, discussing ideas on paper, but some teams have already run the Agent workflows and are actually handling real scenarios. This is the key to success or failure in this cycle.

The gap between "having ideas" and "being truly usable" is much larger than it appears. Just telling stories isn't enough; you need to deliver something that can run. Projects that have already deployed and operated their Agent systems will have a growing advantage over competitors still writing technical whitepapers.

This is also why the market will gradually淘汰 teams that only have visions but no products. Actual execution capability is always the strongest moat.