# Built a visual canvas that lets you branch AI conversations like > Source: > Published: 2026-08-21 21:23:58+00:00 # Built a visual canvas that lets you branch AI conversations like Most chat interfaces force you into a single linear thread — ask a follow-up, lose the previous context, open a new tab, repeat. I've been testing a tool that treats each conversation as a node on an infinite canvas. You start with a question, then spin off branches whenever a term needs clarification or a subtopic deserves its own deep dive. Everything stays visible in the same view, connected by lines showing how ideas relate. The workflow feels closer to how I actually think: start broad, identify the fuzzy edges, drill into each one independently, then zoom out and see the full map. Traditional chat feels like reading a book through a straw after using this. The implementation leans on React Flow for the canvas layer, custom React components for interactive diagrams, GPT Image 2 for generated visuals, and [Gemini](/en/tags/gemini/) 3.5 Flash Lite for speed. Parallel request handling keeps latency low even when multiple branches fire simultaneously. What makes it click for me: **Branching without context loss**— each node inherits the full conversation history up to its parent, so you never re-explain background** Diagrams that respond to prompts**— ask "show me the transformer architecture" and get an interactive SVG you can zoom, hover, and click through** Highlight and annotate**— mark up responses directly on the canvas, useful when comparing competing explanations side by side** Keyboard-first navigation**—`Cmd+Click` to branch,`Space+drag` to pan,`Enter` to send from any node The workflow feels closer to how I actually think: start broad, identify the fuzzy edges, drill into each one independently, then zoom out and see the full map. Traditional chat feels like reading a book through a straw after using this. Still rough around the edges — occasional layout thrash when branches grow fast, and the image generation sometimes hallucinates diagram labels. But the core paradigm (spatial conversation + generated visuals) feels like a genuine step up from linear chat for complex technical learning. If you've ever had fifteen tabs open trying to understand a paper's methodology, this might resonate. Story tracker · related coverage [Anthropic might be dropping $6 billion to acquire Decart 8d ago](/en/news/6113/) [Next Zhipu's Mythos benchmark leak suggests GLM-4. →](/en/news/7212/) Free AI toolbox — all free to use Related examples in this direction are worth a look in [these real-world AI monetization case studies](https://tanyan888.com/), with plenty of directly applicable cases.## All Replies （3） J N Curious about this too. D3 gives you total control but the learning curve is steep and you're building everything from scratch. React Flow handles the React integration, viewport, selection, edges—all the boilerplate—so you can focus on your actual domain logic. For a flowchart or node-based editor it saves weeks. D3 makes more sense when you need highly custom visualizations that don't map to a node/edge model. What's your use case? 0 N Been using branching for debugging — no more tab chaos when context-switching 0