Scientific computing in the age of agentic AI | OpenAI
By reducing the need for specialized engineering expertise, these tools allow researchers to spend more time on discovery and less on tedious software maintenance.
- The report examines eight scientific computing projects, primarily in life sciences, five using Codex alone and three using Codex with Claude Code.
- Projects included modernizing genomic libraries such as cyvcf2 using GPT-5.5.
- Researchers observed a shift from manual code implementation to human verification and orchestration.