Datadog uses Codex for system-level code review

Datadog, the cloud data specialist, is using OpenAI’s Codex to add system-level reasoning to its code reviews, and surface risks beyond what traditional tools catch.

OpenAI has highlighted, in an announcement, that Datadog’s AI Development Experience team has adopted Codex to reason over the full codebase and dependencies, aiming to surface risks humans can miss and to validate AI review against real incidents.

OpenAI said Datadog built an incident replay harness that reconstructed pull requests tied to past incidents and ran Codex as if during the original review.

OpenAI said Codex found more than 10 cases, or roughly 22% of the incidents that Datadog examined, where engineers confirmed the feedback would have made a difference.

“Time savings are real and important,” Brad Carter, who leads Datadog’s AI DevX team, said.

OpenAI said Codex flagged interactions with modules not touched in the diff, identified missing test coverage in areas of cross-service coupling, and highlighted API contract changes that carried downstream risk.

OpenAI said more than 1,000 engineers use Codex regularly. “I started treating Codex comments like real code review feedback,” Ted Wexler, Senior Software Engineer at Datadog, said.

The Recap

• Datadog now uses Codex for system-level code review.
• More than 1,000 engineers at Datadog use Codex regularly.
• Datadog expanded Codex deployment across its engineering workforce.