Tessl says about 75% of its agent-written PRs now merge without human review
Original title75% of Tessl's PRs now merge without human review
AISummary
Tessl says about 75% of its agent-written pull requests now merge without a human correcting the agent or discussing the engineering. It attributes this to Tessl Code Review, which checks each PR against the team's own standards. Over one side-by-side week, 91% of its findings were real defects, against 68% for its previous AI reviewer, and review cost per PR fell about 35%.
AIWhy it matters
The post gives measured review quality and cost figures for an AI reviewer on a real monorepo, useful for teams weighing whether agent-written code can merge without human review.
Source: Tessl Blog · tessl.ioPublished · added here