Scientific computing in the age of agentic AI

OpenAI published a field report on July 28 examining how scientists are using coding agents to modernize scientific software for genomics and other data-rich fields. The report compiles eight agent-assisted projects primarily in the life sciences, five using Codex alone and three using a combination of Codex and Codex and Claude Code. Contributors reported that agents significantly accelerated software development and maintenance, in some cases helping small teams take on work that would otherwise have required far more time or specialized engineering support. The projects ranged from routine maintenance and targeted optimization to large-scale language migrations and GPU-native redesigns. Researchers described a shift in their role from implementation to verification and orchestration, specifying what to build and defining how to measure correctness. The report identified validation of an AI agent's output as the primary bottleneck, noting that agents often expressed confidence even when their work contained clear errors. Human reviewers needed reliable ways to validate results using external references or measurable acceptance targets. Projects generally proceeded in stages using feedback-driven iterations, with resolving edge cases and subtle numerical differences taking the most work. Long-term stewardship and attribution remain essential, according to the report. Lower implementation costs can make it easier to produce many similar rewrites, fragmenting users and spreading expert attention. The case studies illustrate paths forward, including incorporating changes into original upstream projects or establishing new community stewardship with a clear owner and credible maintenance plan.