To me, this looks more like an iterative technology cycle. The frontier labs continue to improve the models, while the agents, tooling, infrastructure, and applications is evolving rapidly. So, as long as the feedback loop between model improvements and real-world adoption continues, short-term losses don't necessarily tell us much about the long-term trajectory.

The author never states what would constitute winning