miguelacevedo

love it!

Same thing was said about Character AI, Cohere, etc. Companies started by the authors of the "Attention is all you need" paper. Look how that went... the team gets thrown around a lot as if it's the magic bullet. There is no magic bullet.

wild.

That helps, but people tend to ask about uses they can already imagine. The hard part is discovering what’s possible beyond that.

nice catch haha

really cool!

The AI will auto-select tools if your request has enough context for it to pick up on the relevant tools. That said, sometimes you want to think in terms of "tool first" rather than prompting on its own.

Thanks for the video rec! Engelbart was heavily influenced by the 1945 essay "As We May Think". We agree with the general principles from the essay i.e. cognitive augmentation over automation and the focus on extending our cognitive ability.

We just want to make sure we're grounded in real-world practical utility because it's easy for ambitious interface ideas to remain demos. We have a lot more exciting, weird, and strange ideas to come!

We had a couple of other weird ideas that we discarded as they were not as intuitive. ie you could drag in files and tools along with the idea of what you want, and Marble would come up with 3 task cards for you to select from (we proactively included different files or tools that could complement your goal/request).

Another idea was a "preview" step before running a task, which used dummy data to mock up a visual of what the output from a task could look like. This allowed you to shape the structure of the task results before even running it. It ended up being confusing as it was hard to tell real vs fake/simulated dummy results.

Something we will be keeping, though, is the ability to have an artifact opened i.e., a spreadsheet, Slides, PDF... and select or interact with any part of the document, and Marble proactively suggests ways to extend that document with the context of what you selected/highlighted, and the purpose of the workspace/the work task. It will look like a tool selection wheel used in video games.

Yes, I think the good comparison here is games with keyboard shortcuts + graphical interface. The interface helps with discovery, but it's complemented by the need for speed through shortcuts.

We see this playing a role in things like tool selection, ie setting up shortcuts similar to gaming macros for quick selections.

To clarify, we meant "strange" from the perspective of a new user, not that terminals are bad. GUIs made common actions visible and discoverable: you could point, click, and drag instead of already knowing the command and syntax.

The key design tension is whether the main space should be the asset you are working on or the work being delegated.

For a lot of tasks, I don’t think you will need to stay inside the document or app the whole time. But I can also imagine starting from an asset you already have open ie a PDF, spreadsheet, etc and bringing agents and tools into that context, rather than creating a separate task and attaching the asset again.

With that in mind, I agree that fanning out into disconnected task outputs is not enough. Sometimes the asset itself should become the center of the workspace, with the delegated work happening around it.

I think a better example is how Marble makes tools visible and easy to select for each task instead of requiring you to remember what exists or describing everything through the prompt.

Marble keeps the tools you use most close at hand and will surface other useful ones for the job. Combined with a birds eye view of multiple tasks running, it makes agent work easier to set up and manage. The goal is to minimize cognitive overhead in day to day tasks (both in specifying what you want, and managing after you ran something)