Conversation control
Annotate AI Responses Without Rewriting the Prompt
Annotations let you attach a note to an exact passage in an AI response. Your note stays visible beside the selected text and becomes part of the context for later requests. You can clarify a fact, reject an assumption, or preserve a decision without restating the whole discussion.
Correct the passage, not the whole conversation
Long chats accumulate decisions, exceptions, and corrections. A normal follow-up prompt puts the correction several messages away from the claim it belongs to. An annotation keeps the two together.
This is useful when the response is mostly right but one sentence needs a caveat. It also works as a small editorial layer: mark the wording to preserve, add missing context, or record why a suggestion was rejected.
- 1Select text in an AI response.
- 2Choose Add Annotation from the selection menu.
- 3Write the correction, instruction, or context attached to that passage.
- 4Continue the conversation and refer back to the annotated material when needed.
Use annotations as durable working context
Annotations are saved with the conversation and sync through iCloud with the rest of your Cumbersome data. They are especially useful for research, editing, planning, and other work where the same constraint should survive several rounds of discussion.
What to know before you use it
- An annotation does not prove that the selected claim is true. It records your context for the model.
- Providers still have context limits. Compact or start a new conversation when a thread becomes unwieldy.
- The annotation is included when relevant context is sent to the AI provider selected for the conversation.