OpenAI announced a deep collaboration with global networking giant Cisco, leveraging its advanced code AI model Codex to enhance the efficiency of enterprise software development and maintenance. Codex is capable of understanding natural language, generating new code, and fixing old code vulnerabilities, making it a core component in Cisco's enterprise software development process.

Cisco not only views Codex as an independent development tool but also deeply integrates it into production engineering workflows, enabling it to operate within large-scale multi-repository systems while meeting global enterprises' requirements for security and compliance. Ching Ho, a member of Cisco's engineering leadership team, stated that the biggest highlight of Codex is its "agency" capability, which makes it not just a code completion tool but also a powerful assistant for enterprise-level development.

In practice, Codex has demonstrated strong capabilities. It can connect multiple large code repositories, understand and reason through them, handle complex programming languages, and perform autonomous "compile-test-fix" loops based on command lines, seamlessly integrating into existing production environment workflows. Additionally, Codex can efficiently operate within existing audit, security, and governance frameworks.

Cisco engineers provided feedback on Codex, further improving its performance in workflow orchestration, security controls, and support for long-term engineering tasks. After using Codex, the Cisco team achieved significant results in several complex tasks. For example, by analyzing build logs and dependency graphs from more than 15 interconnected code repositories, Cisco successfully identified inefficiencies, reducing build times by approximately 20% and saving over 1,500 engineering hours per month. In terms of defect resolution, Cisco used Codex-CLI to implement large-scale automated fixes, shortening work that originally took weeks to just a few hours, increasing defect processing throughput by 10 to 15 times.

In terms of framework migration, when the Splunk team needed to migrate the user interface from React 18 to React 19, Codex handled most of the repetitive changes, compressing what would have taken weeks into just days, significantly improving work efficiency.

The collaboration between Cisco and OpenAI is not only a technological innovation but also an important milestone in the future of AI-assisted development and operations. Both parties will continue to maintain close cooperation, driving the development of enterprise-level AI-native engineering tasks.