The core idea
Linear chat is fine until the work stops being linear. Research splits into sub-questions. Debugging needs a second opinion. Writing needs a fact-check side path that should not leak back into the draft. In a normal chat app, all of that gets tangled together. NodePad gives that work a shape. A workspace is a canvas with its own title, default model, instructions, and set of branches. Any message can become the starting point for a new branch. You can reference a specific message from elsewhere on the canvas with@, or you can merge several source messages into one new draft when the next turn needs to synthesize across them.
In practice
Research
Keep separate threads for separate questions. Compare findings side by side. Pull only the messages you want into the next synthesis turn.
Engineering
Branch from a diagnosis message, ask another model for a second read, and merge the useful pieces before you change code or infrastructure.
Writing
Draft in one branch, fact-check in another, then reference or merge the result back into the main draft without dragging the whole research thread with it.
What lives on the canvas
Conversation nodes
Your messages and the assistant’s replies — the units of AI interaction.
Edges
Continuation, fork, merge, and reference edges show how messages relate.
Text notes
Freeform annotations for you, not for the model.
Attachments
Files travel with the message that used them and carry forward into branches and merges.
Continue
Quickstart
See the product flow in seven steps.
The canvas
How the workspace surface holds branches, notes, and edges together.
Forking
Why and how to branch from a specific message.