Brief in progress — review the script and demo before recording.
Here we need to push the concept of a feedback loop as far as it can go, covering the major areas of a product and how we can safely connect to multiple data sources, including ones that come from a production environment. Besides explaining how we can do it, we should also show why we want to.
- Review: tests are useful, but agents can do much more — read logs and databases, interact with the app dynamically, take screenshots, and even record and label videos. This can be matched not only to the requirements of a given feature, but also to the broader product context.
- Verification: review needs hard evidence, which is tricky to get when multiple agents are working on the same project. We need to show that this is not just about the agents, but also about project setup and the development environment (often in the cloud).
- Quality: we need to show that the review process can focus on better product quality — catching bugs, error logs, or incidents faster. We can go well beyond code and think about customer support and feature requests, so we can deliver them faster.
- Improvements: a nearly autonomous verification process can be used not only to deliver reliable software, but also to explore user behavior analysis, the issues users run into, call transcripts, and more.
We may want to showcase scenarios such as:
- Similar to the previous lesson, we could focus on how you work with Kody and how you organize it so agents can work remotely, on multiple features at once, while also helping you make the entire product better with valuable insights. Note: we cannot just explain this; we need to show it.