Most inference UIs that I've come across pretty much just give us a chat-like interface to toy around with models in a single visual conversation thread.<p>Given the fact that we are limited to seeing only one output at a time, it's kind of hard to compare outputs from different models, adjustments made to the prompting, and sampler settings.<p>But even when keeping the generation parameters the same (e.g., to test for reliability in the output) and just going for multiple passes, there is no easy way to have a side-by-side comparison to keep track of the outputs from the multiple "rounds".<p>I really like excalidraw and thinking visually, so that got me thinking: why not build a sort of "open-world" playground to just place nodes that represent system, user, and assistant messages on a big canvas? Since they all fit on the screen, we have this easy side-by-side comparison that is basically unlimited in the number of direct side-by-side threads at any level. Meaning one user input could lead to two assistant outputs that in turn introduce new sub-branches that can be tracked visually.