AI joins the research team: code, experiments and human involvement

OpenAI’s September 6, 2026 account describes researchers delegating well-defined coding and experiment tasks to agents while continuing to guide design, building, running, analysis and communication.

Internal data record more code contributions and experiments. Those counts describe activity volume, not an equivalent increase in scientific discovery or research capability.

Over the preceding six months covered by the article, more than half of successful 4–8-hour tasks involved at least one human intervention. This observation is limited to that successful-task subset, duration and time window; it is not an overall automation rate.

The record describes changing responsibilities: people define tasks and continue guiding, while agents perform parts of coding and experimentation.