This is a placeholder essay body for Agents Need Boundary Objects, Not Better Prompts. Replace it with your own writing. The structure below mirrors the shape of a long-form piece: an opening, a few sections with headings, a pull-quote, and a closing.
The problem
Most tools are designed for a single person sitting alone at a desk. But most meaningful work happens between people. The interface rarely acknowledges that anyone else exists. Presence indicators are the usual exception, and they are often the first feature to be switched off.
When we design as if the user is alone, we optimise for throughput and ignore the slower, messier coordination work that actually determines whether a project succeeds. The research on computer-supported cooperative work has said this for decades.
A small example
Consider a shared document. The cursor of a collaborator is a tiny signal, but it changes how you write. You slow down near them. You leave a comment instead of overwriting. The interface made someone else visible, and your behaviour changed.
Design is not just how it looks and feels. Design is how it works for the people around you.
What we could do instead
Three ideas keep resurfacing in my notes:
- Make other people ambiently visible, without demanding attention.
- Prefer reversible actions, so coordination mistakes are cheap.
- Give groups a shared history they can point at.
Where this breaks down
None of these are free. Every signal you add is a small tax on attention, and the budget is finite. A useful test: remove the signal for a week and see whether anyone asks for it back. I wrote about that budget in Quiet Interfaces: Designing Software That Knows When to Shut Up
On notification fatigue, ambient signals, and the case for tools that respect your attention
Closing thoughts
I don't think this is solved. I think it is under-explored, which is a much more interesting state for a problem to be in.