Rethinking AI Workflows with Moodboard-Style Threads

I use ChatGPT and Claude extensively across research and development work. While large language models are incredibly capable, the linear chat interface becomes a real bottleneck once a project spans multiple days.

Digging up a key insight from yesterday means scrolling through endless history or asking the assistant to resurface it. That re-generation usually comes back slightly different, burns tokens, and sometimes fails entirely if the message already slipped outside the context window.

For bigger initiatives you often juggle several sub-topics at once. Keeping those discussions separate would help retain focus, preserve only what matters in context, and give you a better overview of the project.

Inspired by mood boards and collaborative canvases like Miro or FigJam, I built a quick prototype that lets you arrange conversational artifacts spatially. Each thread lives in its own tile so you can zoom in, annotate, or branch without losing the bigger picture.

I spun it up in Lovable over a single weekend - the tool was a perfect fit for hacking together an interactive proof of concept with minimal friction.

This kind of rapid prototyping feels essential for modern product managers. Instead of spec-ing a static mock for design, you can assemble a working experience in a fraction of the time and bring it straight into user conversations.

Take a look and share your thoughts: canvasgpt.novykov.com. Could this pattern reshape brainstorming, planning, and creative exploration with AI?