Heimdl and Codex

Codex reviews the change, Heimdl checks the tests behind it

Two different questions

Codex asks

“Does this change look right?”

Heimdl asks

“If this change were wrong, would a test fail?”

Across the 1 pull request Codex reviewed, the tests missed 27% of the changes

documenso/documenso #3355

HighAvailabilityInvalid timezone shows an error with no message93changes68caught22survived3uncovered4tests suggestedteam-analytics-range.ts:46survived- { message: 'Invalid analytics timezone' }+ {}Suggested test, rejects invalid timezones with correct error message

expect(() =>
  resolveTeamAnalyticsRange({ range: '30d', timezone: 'Not/A_Zone', now: new Date('2025-01-15T12:00:00.000Z') }),
).toThrowError(/Invalid analytics timezone/);

Running both

Keep Codex for review, add Heimdl for proof your tests hold the change

Get a free audit All case studies

Other comparisons

CodeRabbit Greptile cubic Devin Veria GitHub Copilot

Missed, the share of changes on the pull requests a reviewer reviewed that no test caught, survived plus uncovered

From Heimdl’s published audits of public pull requests, a reviewer is listed where it commented on the pull request