What things you decide to collect into a bucket is a high potency action for establishing context.

Most times computers can't extract that context. But humans can get the vibe, and LLMs can too.

If an LLM can give a collection of data a good title, that shows that the context established is clear.

Each incremental step of work should show incremental progress.

You guide the AI on what kinds of things to do by sorting it into buckets. The AI can figure out from context what kinds of things make sense the better you organize.

A "here are some items to add, which ones do you want to keep?" flow is easy to react to, gives you immediate feedback that your guidance was useful, allows steering it.