Brief in progress — review the script and demo before recording.
This lesson is about seeing the development process through an AI lens: agents that handle development, manage specifications, and have the tools they need to verify their results, interact with a development version of the app, observe outcomes, access necessary information, and understand both the broad and narrow context of a product, including the history of implemented features. Agents also need to be proactive, work on schedules, and respond to external events.
- Data Sources: a look at the data sources agents need to work nearly autonomously, make decisions, and verify their results, not only during development.
- Templates: agents should not be constrained too much, but they can receive general guidance in the form of prompts, steering messages, specification or evidence templates, or rules for how work should be organized.
- Feedback Loops:
"The key lesson was that if an agent can't verify its own work, nothing else matters"— Lauren. We need to show what this means in practice: what loops are, how our thinking about developing features needs to change, and how we can use AI to deliver loops that were not possible before. - Direct Work: working with multiple agents at once is exhausting, and there are many reasons AI cannot yet be trusted. That changes when the points above are well organized and maintained by AI itself.
We may want to showcase scenarios such as:
- A video about how you develop Kody, showing how much you have delivered on your own, would be the strongest scenario here.