Edugolr/sverige-framatPRlive source
HUMAN ASSURANCE FOR AI-GENERATED CODE

Review the change, not the diff.

AI can produce a pull request in seconds. L.I.N.S. turns that change into a small set of evidence-backed questions that still require human judgment.

Traditional review Reconstruct the change from the diff Files, hunks and context first.
L.I.N.S. Start with the decisions that matter Questions, consequences and exact evidence on demand.
THE HUMAN LAYER

AI prepares the case. You make the judgment.

01Understand the change

See what becomes true and why it matters before reconstructing implementation details.

02Verify what matters

Every important claim stays traceable to the exact changed source.

03Own the judgment

L.I.N.S. can expose uncertainty and disagreement. The reviewer owns the decision and the words sent to the author.

Shared evidence. Independent judgment.

Other reviewers’ conclusions stay hidden until you form your own. If reviewers disagree, L.I.N.S. makes the difference visible instead of averaging it away.

YOUR REVIEWS

Start your next review

Open the pull request that needs your attention, or browse another one.

GITHUB · CONNECTING Connecting to GitHub…
CHANGE OVERVIEW

What will this change make true?

CHANGE IN BRIEF

SYSTEM CONTEXT

REVIEW QUESTIONS
Questions that require human judgment
Shared evidence. Independent judgment.
No authorization judgment is recorded until you make it.
AUTHORIZATION QUESTION

ASSURANCE ARGUMENTWhat the current evidence says
WHAT BECOMES TRUE

CONSEQUENCE

BEHAVIOR PATHChanged execution path

Every step is grounded in changed source
Follow the causal path. Use source only where you need to verify it.
YOUR JUDGMENTIs the evidence enough to decide?
—
Choose review languageWe’ll remember your latest choice on this browser. You can change it anytime without rebuilding the review.
Review model history

Immutable analyzer snapshots for this exact pull request revision.

GitHub review preview

This is the complete author-facing package. Nothing below is sent automatically.

Attach finding to codeSelect one changed line, or a range within the same diff hunk.
No source selected.
Capture findingRough is fine. Polish comes later.
PR-level finding
What do you want from the author?