AI Research Becomes an Autocatalytic Industry

When large models not only assist but materially speed and direct AI R&D (writing code, designing chips, creating benchmarks), the whole research ecosystem can turn into a self‑amplifying industrial process in which machine‑generated insight, machine‑written code, and machine‑run experiments become the dominant inputs. This dynamic creates faster, continuous cycles of improvement decoupled from human timescales and institutional slowdowns. — If true, this shifts policy questions from incremental safety rules to industrial governance, capital controls, and economic chokepoints because machines — not people — would be the main drivers of capability growth.