The core of the warning is a comparison of magnitude, not of kind. Where a major exploit today might drain a single protocol or exchange, the fear is that AI agents could run many such attacks in parallel, at machine speed, making current losses look small by comparison. The post...