Diffs Don’t Scale: Why We Need Better Code Review Interfaces

Git diffs are fundamental to how we review and reason about code — but they're breaking down.

As teams adopt increasingly AI-assisted workflows, and the size and complexity of code changes grow, the old ways of reviewing code (just scrolling through textual diffs) don’t scale. Reviewers struggle to understand the full impact of a change, context is fragmented, and discussions often miss hidden dependencies.

In this talk, I’ll explore the limitations of current diff-based review interfaces and introduce a new way of thinking: diffgraphs — visual, contextual representations of how a codebase changes. These aren’t just graphs for the sake of it; they surface function-level connections, call flows, and implicit coupling between components in a PR, helping reviewers quickly grasp what really changed.

I’ll share learnings from building and deploying early versions of Vibinex (central repo: https://github.com/vibinex) — an open-source project to augment code review on GitHub — and show how we’re now doubling down on the diff-graph-generator CLI as a standalone FOSS tool. Along the way, we’ll cover real-world usage insights, community reactions (including a recent HN feature), and the road ahead.

This talk is a call to arms: let’s make code review better, together.

- Traditional git diff views are no longer sufficient for modern, AI-assisted development workflows. The attendee will be prompted to think of more ways to solve this problem apart from the ones proposed in the talk.

- Visual, contextual diffs (like diffgraphs) can dramatically improve how we review and reason about code changes.

- You’ll learn about open-source tools you can start using or contributing to, including diff-graph-generator, and see how to plug them into your workflow via CLI, browser, or VSCode.