Canvas Threads Give AI Agent Chats Context
Attaching conversations to specific objects on a visual canvas could make them easier to find, keep discussions focused and allow people to bring AI agents directly into the work.
Developed from a conversation between Pete Winn and Andy David
Chat works well as a way to talk to a computer, but it becomes difficult to manage when conversations grow long and sit alongside thousands of others. The problem isn’t chat itself. It’s knowing where one discussion should end and another should begin, particularly when important information becomes buried across multiple threads.
A canvas offers a possible solution by treating comments as chats. A conversation could be attached to the whole workspace, a particular applet or something inside that applet. This gives each thread a clear place and purpose, rather than adding everything to one long conversation. Because the comments are also chats, people could bring others into the discussion or send requests directly to an AI agent without leaving the canvas.
The difficult part is recognising when a conversation should branch. People often realise they’ve changed direction only after travelling well past the fork, leaving a discussion that deserved its own thread buried in the original. Attaching chats to canvas objects won’t eliminate that problem, but it provides a more natural way to organise conversations around the work itself. Someone can return to a particular object and pick up the discussion where they left off.