Resolving issues during the training run led to a significant performance boost, which energized the team and improved morale.

Despite challenges, we had extensive planning for the run, starting a year before training began, which included careful sequencing of changes to minimize risks.

Post-launch, the team continuously monitored the run for improvements and correctness, balancing the need for data efficiency with the challenges of noisy signals.

The initial traction we received was surprising, leading us to iterate quickly based on user feedback and adapt our product accordingly.

We brought in customers early to test our product, which helped change the minds of skeptical team members when they saw real-time reactions from users.

Teams are becoming slightly more technical, which is exciting because working with strong technical teams makes the process more enjoyable.

We all had moments where we impressed each other, which contributed to our strong team dynamic despite occasional disappointments.

We've learned a lot through the process of training this model, discovering what scales well and what doesn't.

We shifted our guiding principle from 'think long-term' to 'think long-term, ship today,' empowering our team to take risks and experiment, which accelerated our execution significantly.