Branching prompts instead of linear chat

Most LLM interfaces assume thinking happens in a straight line.

prompt → response → refine → refine

But when I explore ideas with AI, it usually branches.

One prompt leads to multiple directions, and it's hard to compare them in a single chat thread.

So I built a small workspace where prompts branch like a tree.

You can explore different reasoning paths and continue from the most interesting one.

I'm curious what HN thinks about node-based interfaces for prompting compared to traditional chat UIs.