Reviewers got 80% of their afternoons back. Grid did the boring parts.
4.0 DIFFS
Review with context, not just code
Structural diffs that show *what* changed and *why*. AI suggestions inline, every time.
4.0
Diffs that explain themselves
See structural changes, not just text. Grid groups related edits and surfaces the intent.
session.ts
- export async function refresh(token) {
- const res = await fetch('/api/refresh')
- return res.json()
- }+ export async function refresh(token: Token) {
+ if (!isExpired(token)) return token
+ const res = await fetch('/api/refresh')
+ return res.json()
+ }
A
Adam commented
Consider extracting the token guard into a helper — used 3 times in this file.
C
Copilot resolved
Added guardToken() helper in commit 2.
C
Copilot 3 suggestions
Extracted token guard. Added unit tests. Removed unused import.
⌥
Codex 1 suggestion
Inferred return type from response shape — change applied.
Merge queue
- ✓ PR #4421 Ready
- ↻ PR #4419 Rebasing
- ○ PR #4418 Queued
-
4.1 Structural diffs
See what changed and why, not just lines. Related edits group together so review feels like reading a story.
-
4.2 Inline review
Comment on a line, a block, or a structural change. Threads stay attached to the code even after rebases.
-
4.3 Agent suggestions
Agents propose fixes inline. Accept with one keystroke; ignore with another. Reviewers focus on judgment, not typos.
-
4.4 Merge queue
Stack PRs in order, auto-rebase, and merge sequentially. No more rebase wars on a busy main branch.
Three steps to a clean merge
Step 01
Structural grouping
Grid clusters related changes — API + tests + docs land as one reviewable unit.
@Grid
create urgent issues
Step 02
Agent suggestions
Copilot proposes refactors and tests inline. Accept with one keystroke.
Bug
Perf
Design
Step 03
Merge queue
Stack PRs, auto-rebase, and merge in dependency order.
Copilot
Codex
Cursor
Reviews that respect everyone's time
Agent PRs land cleaner than half our human PRs.
Structural diffs found a regression in 90 seconds that would have taken a week.