Creating AI canvases

This guide walks you through creating an AI Canvas from start to finish, from attaching your knowledge to saving the response.

Before you begin

AI Canvas works best when you have entries in your knowledge base with marked bits (highlighted snippets or important sections). If you have not marked any bits yet, open an entry in the detail panel and highlight the passages you want to use as context.

Step-by-step walkthrough

Step 1: Open AI Canvas

Navigate to AI Canvas in the sidebar. The canvas opens with an empty prompt area and an attachment section.

Step 2: Attach marked bits

1. Click the Attach button at the top of the canvas
2. A panel appears showing your entries that have marked bits
3. Browse or search for the entries you want to include
4. Click on individual marked bits to attach them
5. Repeat for as many entries as you need

Attached bits appear as cards above the prompt area. Each card shows a preview of the content and the source entry title. You can remove any attachment by clicking the close button on its card.

What counts as a marked bit?

Content type	What gets attached
Text highlight	The selected passage from an article or note
Screenshot crop	A specific region you marked on an image
Conversation excerpt	A highlighted exchange from a ChatGPT or Claude conversation
Audio segment	A transcribed section from a voice recording
PDF selection	A highlighted passage from a document

Step 3: Write your prompt

In the prompt area, describe what you want the AI to produce. Good prompts are specific about:

• The task: What you want done (summarise, compare, draft, analyse, brainstorm)
• The format: How you want the output structured (bullet points, table, narrative, numbered list)
• The scope: How detailed the response should be (brief overview, comprehensive analysis, executive summary)

Example prompts by use case:

1. Research synthesis: "Summarise the key findings from these three articles. Identify where they agree and where they contradict each other."
2. Content drafting: "Using these notes as background, write a 500-word introduction for a blog post about knowledge management."
3. Decision support: "Based on these product reviews and comparison notes, create a recommendation table with scores out of 10 for each option."
4. Idea generation: "Combine the themes from these saved articles and suggest five novel approaches to the problem described in my project notes."

Step 4: Select a reasoning level

Choose how deeply the AI should think before responding. The reasoning level appears as a dropdown or selector near the prompt area.

Level	Behaviour	Best for
None	Direct response with no extended thinking	Simple tasks, quick summaries, straightforward questions
Low	Brief consideration before responding	Standard prompts, content drafting, basic analysis
Medium	Moderate deliberation with step-by-step thinking	Comparisons, multi-source synthesis, nuanced questions
High	Deep reasoning with thorough analysis	Complex analysis, strategic recommendations, detailed research

Higher reasoning levels produce more thoughtful responses but take longer to generate and use more processing credits. Start with Low or Medium for most tasks and increase to High when you need deeper analysis.

Step 5: Enable web search (optional)

Toggle Web search on if you want the AI to supplement your attached knowledge with current information from the web. This is useful when:

• Your saved content may be outdated
• You want to verify facts against current sources
• You need context that goes beyond what you have captured
• You are researching a fast-moving topic

When web search is enabled, the AI will cite web sources alongside your attached entries.

Step 6: Click Compose

Click the Compose button to submit your prompt. The AI processes your attachments and prompt, then generates a response below the prompt area.

Processing time depends on:

• The amount of attached content
• Your chosen reasoning level
• Whether web search is enabled

Working with responses

Reading the response

The AI's response appears directly below your prompt. It may include:

• Formatted text with headings, lists, and tables
• References to your attached entries
• Web citations (if web search was enabled)
• Code blocks (if relevant to your prompt)

Saving a response

To keep a response in your knowledge base:

1. Click Save below the response
2. The response is stored as a new entry
3. It includes a link back to the original prompt and attachments

Saved responses go through the same AI processing pipeline as any other entry, so they will receive a summary, classification, and relationship mapping.

Iterating on results

If the response is not quite what you need:

1. Adjust your prompt: Reword for clarity or add more specific instructions
2. Change the reasoning level: Try a higher level for more depth or a lower level for brevity
3. Add or remove attachments: Include more context or remove irrelevant bits
4. Toggle web search: Enable it for additional context or disable it to focus on your knowledge

You do not lose your previous response when you compose again. Each composition generates a new response, so you can compare iterations.

Managing your canvases

Viewing past canvases

Your previous AI Canvas sessions are accessible from the AI Canvas page. Scroll through your history to revisit past prompts and responses.

Organising canvas output

For best results, save valuable responses as entries and then:

• Add relevant topics to make them searchable
• Link them to related projects
• Mark important bits for use in future canvases

This creates a feedback loop where AI-generated content becomes part of your knowledge base, available as context for future prompts.

Common patterns

Here are effective ways to use AI Canvas in your workflow:

1. Weekly research digest: Attach all articles saved during the week and ask for a consolidated summary with key takeaways
2. Meeting preparation: Attach relevant project entries and recent notes, then ask for a briefing document
3. Competitive analysis: Attach entries about competitors and ask for a structured comparison
4. Learning consolidation: Attach course notes, tutorials, and articles on a topic, then ask for a study guide
5. Content repurposing: Attach a long article or report and ask for social media posts, summaries, or presentation slides

Troubleshooting

Problem	Solution
No marked bits available	Open entries in the detail panel and highlight important passages first
Response is too vague	Add more specific instructions to your prompt or increase the reasoning level
Response ignores attachments	Reference the attachments explicitly in your prompt
Processing takes too long	Reduce the number of attachments or lower the reasoning level
Web search results are irrelevant	Disable web search and rely on your attached knowledge instead