Whiteboard: an open-source VS Code fork built for reviewing AI-written code

Whiteboard, a YC W26 project, is an MIT-licensed desktop app that turns a fork of Code OSS into a shared canvas where developers and coding agents design and review software together. It connects to existing agents like Claude Code or Codex and hands them an SDK to draw on an in-app canvas — sequence diagrams, entity-relationship diagrams, or quotes from the agent’s own trace — with each visualization clickable straight through to the underlying source. Because it’s built on Code OSS, it inherits VS Code keybindings and LSP support out of the box.

The standout piece is a semantic, AST-aware diff viewer written in Rust that cuts through noisy raw diffs: large added functions collapse into pseudocode, tests and docs are hidden by default, and the whole thing is customizable through a WASM plugin system. Agents can also query and link their own execution traces onto the canvas, so a reviewer can see which decisions were made autonomously and how they map back to the original requirements. Notably, the team vendors Code OSS wholesale rather than maintaining patches — partly to strip out the roughly 45% of stock VS Code they say is now Copilot — because agents handle patch-based forks poorly.

The tool runs entirely against local checkouts, and its anonymous telemetry excludes code, diffs, prompts, and model output and can be disabled. A hosted team version is planned but the project promises to stay self-hostable. Current limits are real: you can’t edit files, multi-repo reviews aren’t well supported, and shared reviews don’t sync updates. The framing reflects a growing thesis in AI-assisted development — that as agents write more of the code, the human bottleneck shifts from writing to understanding and reviewing it, and the editor should be rebuilt around that job.