Evaluation
Benchmarks agent performance over time: PRs reviewed, accepted suggestions, rejection rate, and an acceptance trend by agent type.
Product
Insights turns every review into the numbers that matter: agent precision and acceptance, bugs caught before merge versus reported after, pick-up time, and time to merge.
| Updated | Baz Reviewer | Agents | User name | User interaction |
|---|---|---|---|---|
| 51 minutes ago | Removing `lightColor`/`darkColor` from the desktop `RecoloredMark` caus… | Logical Bugs | --- | LLukas Brandt / Unaddressed Baz Suggestion |
| 1 hour ago | `new Date(value).toLocaleDateString("en-US", { month: "short", day: "num… | Logical Bugs | baz-reviewer… | SSofia Keller / Addressed Baz Suggestion |
| 1 hour ago | `normalizeDate()` only trims/slices frontmatter, so `resource.publishedAt` … | Type Inconsisten… | --- | LLukas Brandt / Unaddressed Baz Suggestion |
| 1 hour ago | `employmentType` is built from freeform `Job.employmentType` via `toUp… | Type Inconsisten… | --- | LLukas Brandt / Unaddressed Baz Suggestion |
| 1 hour ago | `breadcrumbSchema` is built inline in both the docs page and `src/app/res… | Code Dedup and… | --- | LLukas Brandt / Unaddressed Baz Suggestion |
| 1 hour ago | `leadImage()` dereferences `image.src` without validating the media entry,… | Type Inconsisten… | --- | LLukas Brandt / Unaddressed Baz Suggestion |
| 1 hour ago | Lines 214–225 duplicate the `Finding` mapping in `findings/fetchers/ai_fin… | Code Dedup and… | baz-reviewer… | MMarco Rossi / Unaddressed Baz Suggestion |
| 1 hour ago | `impact` is inserted into `### {single_line_impact}` after only `\n`/`\r` norm… | Basic Security P… | --- | MMarco Rossi / Addressed Baz Suggestion |
| 2 hours ago | `CombineSuggestionsRequestItem`/`CombineSuggestionsResultItem` dro… | Logical Bugs | baz-reviewer… | MMarco Rossi / Addressed Baz Suggestion |
| 2 hours ago | `getCommentExcerpt` passes raw HTML/markdown into `CommentRender… | Basic Security P… | baz-reviewer… | AAnna Novak / Unaddressed Baz Suggestion |
What's inside
Filter every view by agent, user, and time period.
Benchmarks agent performance over time: PRs reviewed, accepted suggestions, rejection rate, and an acceptance trend by agent type.
Every agent in one table: accepted, unaddressed, and rejected suggestions, so you see which add value and which need tuning.
Logs each reply or reaction to a reviewer comment, so you see which reviewers help and where they confuse.
How Baz affects the stability of your codebase, opening with a carousel of notable bugs it caught and how developers reacted.
Bugs caught by the correctness reviewers against bugs reported after release, a read on post-release quality.
Median time to first review and median time from open to merge, with a pre-Baz baseline so the change since adoption is visible, not just the current number.
Evaluated PRs, ready-to-merge rate, the share merged without a human, an outcome breakdown, and throughput over time, the track record behind Merger's verdicts.
The shift
Metrics compute automatically from review activity, no setup. Sign up to see your numbers, or ask for a guided read.