OpenAI published an exploratory field report covering eight agent-assisted scientific computing projects, primarily in the life sciences, with five using Codex alone and three combining Codex with Claude Code. The report describes how scientists used coding agents to modernize scientific software for genomics and other data-rich fields, lowering the cost of engineering work and taking on tedious implementation tasks. It finds that the bottleneck has shifted to validating an agent’s output, which still depends on human judgment, and flags outstanding concerns such as responsibility and ownership for AI-assisted scientific software.
