Tessl ships Verifiers GA and reports 75% of its PRs merge without human review
Tessl announced that its Verifiers feature is generally available: plain-text repository rules are checked on every code change by an LLM judge, returning pass or fail and runnable locally, in pre-commit, or as a CI gate. The default judge is openai/gpt-6-luna, with Jev offered as an alternative. In a separate post, Tessl says about 75% of PRs at its team, which opens over 1,200 agent-written PRs weekly, merged without human correction, and that 91% of Tessl Code Review findings were real defects versus 68% for its previous AI reviewer, at roughly 35% lower review cost per PR; the figures are vendor-reported.
Tessl argues verifiers fill a gap between deterministic linters and general AI code review, and says its code review allowed most of its own agent-written PRs to merge without human correction.
ARTICLE
Verifiers: enforce your repo's rules on every agent change
Enforce repo rules on every agent change with Tessl Verifiers. Automate checks beyond linting to catch issues early. Learn how to implement them.
#Tessl#Tessl Verifiers#Tessl Code Review#Jev#OpenAI#GPT-6 Luna