摘要
arXiv:2606.04025v1 Announce Type: cross Abstract: Dominant programming paradigms inherit an execution model optimised for a bygone era of a single human mind instructing a local machine, leaving contemporary systems burdened with historical path dependencies. When forced to host multi-dimensional, connectionist intelligence, this brittle assembly model fractures under the weight of a profound probabilistic-symbolic impedance mismatch. While contemporary Software 3.x frameworks attempt to patch the mismatch by encasing large language models (LLMs) in increasingly complicated external harnesses, this spiralling architectural complexity only compounds the carrying cost of static code assembly. To address the cause rather than the effects, this paper introduces Software 4.0 -- an autopoietic heterarchy of human intelligence, neural AI, and natively reflective symbolic substrate.