Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design

Original Article ↗ Hacker News Discussion ↗

An open-source tool called Whiteboard aims to make AI-assisted software development more understandable by turning code changes into linked diagrams, semantic diffs, and visual “canvases” that sit alongside existing editors and agentic coding workflows. Commenters are broadly positive about the idea of a dedicated visualization layer for code review and architecture thinking, but question whether it adds yet another, potentially stale source of truth compared to markdown-based specs or existing IDEs. There is also active debate over positioning (IDE vs. canvas), platform support, and how such tools should evolve as agents take over more of the implementation work and humans focus on planning and review.

Product Scope & Positioning

  • Whiteboard is an open-source, local-only desktop tool (currently built on top of VS Code) to visualize and understand code changes, especially those produced by agents.
  • Initially marketed as an “IDE,” but many commenters objected because it cannot edit files; maintainers now frame it as a “canvas” or companion to existing IDEs/ADEs.
  • Intended as a complement to terminals, coding agents, and orchestrators, not a replacement.

Core Features & Use Cases

  • Visual diagrams linked directly to code and diffs; strong emphasis on navigation from diagrams/specs to implementation.
  • Semantic diff viewer and animated, “human-feeling” rendering are widely praised.
  • Main workflows: reviewing complex PRs, understanding agent-generated code, reducing cognitive load, and keeping a mental model of architecture.
  • Some see it as a single feature (an MCP + GUI) that could live inside another harness; others see standalone value.

Workflow & Agent Integration

  • Typical flow described: human plan → agent codes → Whiteboard session explains the implementation and trade-offs.
  • Longer term, they want a scratchpad/plan mode with versioned artifacts, shrinking the gap between design and implementation.
  • Tool is explicitly not tied to a specific agent vendor; integration with multiple agents (Claude, Copilot CLI, etc.) is in progress.

Platform, Packaging & Performance

  • macOS and Fedora builds exist; strong demand for Windows, Debian/Ubuntu (.deb), Arch (AUR), Nix, Flatpak, AppImage.
  • Electron/VS Code base makes the app large (~736MB); maintainers acknowledge this and plan a more native, slimmer rewrite.

Concerns & Skepticism

  • Multiple worries about creating an extra, drifting “source of truth” versus existing tools (Notion, Google Docs, Linear, markdown + Mermaid/PlantUML/C4).
  • Some fear hallucinated or inaccurate diagrams; maintainers stress all diagrams are code-linked to avoid this.
  • Broader unease about tools encouraging reliance on AI and distancing humans from code; others argue this tool actually promotes deeper understanding.

Future Directions & Monetization

  • Plans for a hosted, multiplayer/web product with tighter integration into specs, PR workflows, and a continuously updated “software map.”
  • Monetization likely via that hosted offering, while keeping the local tool open-source.