/dev/fast

Developers and teams reviewing code changes today look at raw line diffs, which hide architectural intent and make it hard to understand why a change was made or whether an AI coding agent's decisions were sound. Long algorithms show up as walls of code rather than clear logic, and understanding the data flow or control flow of a change requires manually reconstructing it in your head. When reviewing agent-generated code, there is no easy way to trace whether the agent's actions match what was actually requested in the authoring conversation. This makes thorough review slow and architecture-level understanding of a diff difficult to achieve.

Whiteboard is a canvas application for reviewing diffs that renders changes as sequence diagrams, DB diagrams, and other architecture-level visualizations instead of raw line-by-line diffs. It uses an AST-based diffing library written in Rust to produce semantic diffs, showing long algorithms as pseudocode while always preserving function signatures, with the ability to expand folded code back to full source. It also supports auditing agent trajectories, letting a reviewer trace a quoted request from an authoring conversation to the specific commit and code change that implemented it. The product is fully extensible with plugins, and a proposed diff search API (parse, hydrate, postprocess) lets callers search hits, validate them against pinned Git blobs, and run plugins over selected comparisons. It runs on macOS, Windows, and Linux and is open source under the MIT license.