A preprint on arXiv proposes a "stateful research information infrastructure" for scientific discovery. According to the abstract, AI is already being applied to a widening range of research activities, including literature analysis, hypothesis generation, software development, and automated experimentation.

The abstract then begins to explain that these capabilities rely on something, but the text cuts off at that point. As a result, the full argument and the specifics of the proposed infrastructure are not available from the source alone.

What is clear is that the authors see a need for an infrastructure that maintains state — that is, one that preserves context and history across AI-assisted research tasks. The preprint is an announcement of this direction rather than a detailed specification, based on the excerpt provided.