On October 21, OpenAI described work with PerturbAI and Scripps Research using many agents to analyze existing data from gene perturbations in mouse brains. The authors' repository shares the workflow and supporting materials and cites a preprint. This is computational interpretation of experimental data, not a human treatment study.

The announced 32 minutes concerns analysis, not producing the atlas, validating mechanisms or developing a drug. Agents divide work across gene perturbations and connect cellular changes with literature. Millions of cell records do not equal millions of independent animals or patients.

A possible use is a second pass over a large experiment: suggest a missed connection and expose its supporting data and publications. A biologist could then select a few testable hypotheses instead of manually examining every small signal.

Prerequisites are citation checking, independent reproduction and follow-up laboratory experiments. Our optimistic editorial horizon is 1–3 months for a research pilot using an available atlas and an initial shortlist. These materials cannot honestly date confirmation of a biological mechanism or treatment for neurodegeneration.