Why a canvas

Slashspace is an AI infinite canvas: a workbench for long, exploratory work with AI. Research, learning and writing rarely move in a straight line. You branch, backtrack, compare, and connect ideas that live far apart.

A chat thread holds one line of thought at a time. A canvas holds the whole shape of the work, and you decide what the AI reads at every step.

Where a chat thread breaks down

A single scrolling chat is fine for a quick question. It starts to fail when the work runs for hours or weeks:

• Tangents have nowhere to go. A side question lands in the main thread and every later answer reads around it. Skip it and the idea is gone.
• You lose sight of what the model is reading. The conversation grows until you can't see, or control, which parts of it shape the next answer.
• Your sources live somewhere else. Papers, pages and videos sit in other tabs and get pasted in piece by piece.
• Comparing means copy-paste. Trying a second approach or model means a fresh chat and pasting everything again.
• Weeks later you have a transcript, not a map. The reasoning is in there somewhere, under hundreds of messages.

What the canvas changes

Everything is a node. Conversations are Chat nodes and notes are Text nodes. A web page, a YouTube lecture, a PDF and an image each get their own node (Web, Post, Document, Image), and a Scratch pad holds a quick sketch.

Connections decide what each conversation sees. Connect a source to a chat and the chat reads it. Chains carry through: a chat reads everything connected upstream of it, and nothing else. A chat with nothing connected gets a summary of the whole canvas instead, or nothing at all when you switch it to Isolated context. See Connections and context.

Any answer can fork into a branch. Send your next question with Submit in new branch and it opens in a new chat node beside the original. The branch picks up the conversation so far, plus everything connected to it, and the original thread stays as it was. See Branching.

Sources sit next to the conversations that use them. Paste a link, drop a PDF onto the canvas, or type @ in a chat to mention a node and connect it.

Different models work side by side. Each chat node has its own model, so you can put the same question to Claude, GPT, Gemini and a local model and read the answers next to each other. Use Slashspace's hosted models, your Claude or ChatGPT subscription (Subscription providers), your own API keys, or local models through Ollama. See Choosing a model.

The AI can take on more of the work. In Agent mode, a chat can use your connectors and split a job across sub-agents, each working in its own chat node. In Orchestrator mode, you describe a goal and the AI builds the canvas: it creates and connects chat, text, web and post nodes, and hands research tasks to sub-agents.

Your canvases are files you own. Each canvas is saved in a folder on your computer. Messages go to the model provider you pick, and some features, such as reading web pages and documents or Orchestrator mode, process content on Slashspace's servers. Privacy and your data covers what leaves your computer.

What a session looks like

Researching a question across papers

You want to know whether a new method holds up. Drop four PDFs onto the canvas, paste links to two critical blog posts, connect them all to a chat, and ask where the sources disagree. When an answer cites a figure you doubt, select that sentence and type your question in the "Ask about this…" box: it opens in its own branch and the main thread stays on track. Later, connect a second chat to the two critiques only, for a reading the original papers don't shape.

Learning a subject from lectures

Paste the links to a YouTube lecture series and each becomes a Post node with its transcript. Put each week's lectures in a group and connect it to a chat that quizzes you. When a concept doesn't land, branch off for a slower explanation, or sketch your understanding in a connected Scratch pad and ask whether you've got it right. Next week, the canvas shows what you covered and where you got stuck.

Drafting a long article

Connect your interview transcripts, notes and reference pages to a chat where you work out the outline. Once it holds, fork three branches, one per angle you're weighing, and ask each for an opening. Every branch reads the same outline and sources, so the drafts differ only in angle. Copy the strongest opening, paste it onto the canvas as a Text node, and connect it to the chat where you write the rest.

Weighing a decision with several models

You're choosing between three vendors. Add each pricing page as a Web node, write your constraints in a Text node, and group them. Connect the group to three chats on different models: Claude through your subscription, GPT on your own OpenAI key, a local model through Ollama. Ask all three the same question, then connect them to a fourth chat and ask where they agree. The reasoning behind the decision stays on the canvas.

When a plain chat is enough

Not every question needs a canvas. A quick fact, a one-off rewrite or a short exchange you won't revisit works fine in one thread, and in Slashspace that's a single Chat node. The canvas asks a little more of you than a chat box, because you decide what connects to what. That pays off when the work spans many sources, several lines of thought, or more than one sitting.

Next

• Quickstart: install Slashspace, ask your first question, fork your first branch and connect your first source.
• Core concepts: the mental model behind nodes, connections, branches and chat modes.