Nuky logoNuky
Open source • v1.3.0 • Docker image

No judgments. Just signals.

Nuky helps you see what actually changed when the diff is too big to read — especially when the code was written by AI. It shows the shape and scale of change, so you can see what happened without drowning in lines of code.

Runs locally. No cloud. No data leaves your machine.

View on GitLabQuickstart

Open source on GitLab. Public Docker image. No gatekeeping.

Terminal
The diff is small. The behavior isn’t.

Code changes are happening in the dark.

AI can generate hundreds of lines of code in seconds. Humans still have to understand what changed. Traditional diffs show where text moved. Nuky shows how the system changed.

1
Diff too large to read
Hundreds of lines land at once.
2
Review without context
You can’t tell what actually shifted.
3
Behavior moved
BEHAVIOR_CHANGE=1
Nuky shows where the change concentrates.
New here?

Pick the path that matches the kind of change you’re staring at right now.

Quick startAI-generated changesLocal workflowHow to readNUKY_SUMMARY

What Nuky gives you

Clear snapshots of change you can trust — even when AI wrote the patch.

Structural change, not line-by-line diffs
See which parts of the codebase changed, how much they changed, and how those changes are distributed.
Signals without judgment
Nuky doesn’t review, block, or comment on your code. It shows signals — you decide what they mean.
Fast and consistent
The same change produces the same snapshot, every time. No randomness. No ‘AI mood swings’.

Nuky does not decide whether a change is good or bad.

It shows signals you can interpret on your own terms.

Dogfooded by design

Nuky is built using Nuky.

We rely on it daily to understand how the project changes as it evolves.

If the signals stop being useful to humans, we feel it immediately.

Roadmap

Focus: correctness first, adoption second.

Nuky evolves around helping humans reason about change — not automating decisions.

Across recent changes, Nuky observed that ~34% were AI-assisted — and flagged ~5% of those as high-risk structural changes that passed tests and review.

1
Shipped
Core change engine
  • Deterministic change engine: scan → diff.
  • Production language support: Go and Dart / Flutter.
  • Local-friendly outputs: canonical JSON + stable summaries.
2
Recently added
Workflow hardening
  • Explicit change classes: contract, behavior, refactor-only.
  • Deterministic summaries built for human review.
3
Planned
Next surface
  • VS Code extension to surface structural changes during review.
  • Earlier feedback before a change ships.

How Nuky works

Turning code into consistent snapshots of change.

Parse
Nuky reads your code and builds a structural view of it.
Compare
It compares snapshots to detect what changed and where.
Summarize
You get a compact, deterministic summary of the change — easy to inspect and reason about.

Each step makes changes comparable, even across large refactors.

NUKY IS

Shows the shape and scale of change, not just the diff.

  • A way to see the shape and scale of change
  • A tool for uncertain moments in development
  • Built for large refactors and AI-generated code

NUKY IS NOT

  • A code reviewer
  • A linter or QA tool
  • A gate that blocks your workflow

Run locally

docker pull registry.gitlab.com/hilly.bopper/nuky:v1.3.0
docker run --rm registry.gitlab.com/hilly.bopper/nuky:v1.3.0 --help
Registry (GitLab)Build instructions

“Text diffs show how code changed. Semantic diffs show what changed.”