Sowing seeds

Recently I’ve been hearing a lot about “loops” as the new focus for AI use. We used to talk about prompt engineering, then context engineering, and now the focus seems to be loop engineering.

My AI research process has settled into a well-worn pattern. Often it starts with a video, podcast, or article. I drop the text into the chat window and start talking about the ideas I encountered in the source. At some point, I feel like I’ve found the core concepts I want to explore and I trigger my research-prompts skill. This generates about 15 prompts that I feed into three new chat windows five prompts at a time. So I end up with about 30k words of research into the topic that I then need summarize with NotebookLM. The podcast becomes a podcast again, but I also have a pile of research I can dig into or use as context.

Recently, I’ve added a new step – creating a limited bibliography that I request after the research is done.

Can you pick out a few sources (available online) that you think I would be most interested in reading? Create a limited bibliography that explains why I might find each source interesting given the direction of this research.

I usually end up with 20 to 25 different sources I can then pick through and explore, and then I can follow the same research process, sowing seeds for the next batch of research. The research loop works as a context-building machine.