Dreamspace

Dreamspace rethinks what an LLM interface can be. Instead of the usual single chat column, it gives you a large two-dimensional canvas where every prompt becomes a node you can move, inspect, and connect. Running a prompt drops its generation onto the surface where you placed it, so exploring a dozen variations leaves you with a visible diagram of how you got from idea to output.

From there the tool is built for comparison and iteration. Clicking a node reveals its details, links show relationships between messages and prompt versions, and you can fork a prompt to experiment with different wording or a different model while keeping the original intact. Outputs can then be fed back in as context, which makes longer, multi-step reasoning chains practical to assemble and reason about.

The intended audience is prompt engineers, AI researchers, and writers who care about the mechanics of prompting rather than a quick answer. It is also useful for documenting why a particular prompt performs well, since the canvas keeps a visual record of every step and branch you tried.