← Back to People
Co-authored research on automated code review as part of the Codex team, discussing how giving the reviewer repo-wide tools improves recall and precision.
2
Mentions
2
Articles
1
Outlets
Associated Topics
Companies Covered
Coverage Patterns
How media typically covers Maja Trębacz
Article Types
- research2
Media Angles
- technical2
Narrative Framing
- breakthrough2
Associated AI Models
GPT-5-Codex1GPT-5.1-Codex-Max1Codex1gpt-5-codex1gpt-5.1-codex-max1
Writes About
Appears Alongside
Articles
Most recent firstArticles Written
Maja Trębacz as author
A Practical Approach to Verifying Code at Scale
OpenAIresearchpositiveDec 2, 2025
OpenAI developed an agentic code reviewer with repo-wide tools and execution access that improves both recall and precision in detecting bugs and vulnerabilities in AI-generated code, prioritizing signal quality over exhaustive detection.
“Author of "A Practical Approach to Verifying Code at Scale" in OpenAI”
AI SafetyAI AgentsLLMsAI Products
Also Mentioned In
Referenced in coverage
OpenAI Breaks Down Techniques That Let Codex Use Repo Tools to Raise Review Accuracy
OpenAIresearchpositiveDec 3, 2025
OpenAI's agentic code reviewer trained on GPT-5-Codex and GPT-5.1-Codex-Max achieves higher review accuracy by giving the model repo-wide tools and execution access, prioritizing precision over recall to maintain developer trust and usability.
“Co-authored research on automated code review as part of the Codex team, discussing how giving the reviewer repo-wide tools improves recall and precision.”
AI SafetyAI AgentsFoundation ModelsAI Research