← All updates
Release 3 min read

Reasoning You Can Trace Back to Your Board

Review source excerpts, distinguish observations from hypotheses and questions, and keep control of pending AI actions.

Start with the source

Open the Visual Reasoning Partner with Ctrl+Shift+R and run an audit. Each finding shows its type. Expand Board sources to read the note text or connection label that prompted it, then click a source to focus the relevant ideas.

The excerpts retain the board text from when the finding was created. If a source is removed, its focus button is disabled. A finding without a valid reference is labelled as uncited; an unsupported Observation is treated as a Hypothesis.

Suggestions remain open to review

Challenge mode presents possible alternatives as hypotheses, while Socratic mode generates questions. Saved gap and next-step notes retain explicit Reviews source links to the ideas behind them.

The Board structure score describes the graph. It does not establish factual accuracy. A source link shows the context behind a suggestion, not proof that its interpretation is correct.

Clearer control over AI actions

Cancel a pending request or close its panel to prevent a late result from being added. Completed results stay visible until you choose View board. If the board changes, rerun the reasoning review before applying its old suggestions.

Tidy still applies its layout directly. It preserves locked items, existing framed groups, and colours, and makes room for different element sizes.

Know what the assistant reads

Canvio reads element text, relation labels and types, and map pins. It does not interpret images or handwriting, or independently check facts against external sources.

Context is limited to 80 elements, 120 connections, 240 text characters per element, and 40 pins per map. Selected elements are prioritized. The interface reports the included context and identifies local fallback when server AI is unavailable.