The latest arXiv postings show AI research scattered across distinct frontiers. One paper proposes learning-to-optimize as a missing layer in AI-native communication networks, arguing that autonomous resource management and zero-touch operation demand more than conventional optimization. Another steps back from technical design to frame AI as a driver of economic, environmental, geopolitical, and social transformation, cautioning that treating it purely as a breakthrough misses broader structural shifts.
Two other contributions explore human-AI interaction from very different angles. One describes a techno-artistic project that reanimates a 1980s East German typewriter with a modern language model, blending media archaeology and speculative design. A second develops "adaptive complementarity," a framework for choosing interaction architecture that improves current performance while reshaping the capabilities future performance depends on.
A fifth paper examines industrial roll-outs of AI development tools, finding that while organizations expect large productivity gains, developers' actual experiences reveal a gap between promise and practice. The sources share a common interest in AI's role in complex systems, but they otherwise diverge in method and scope. Readers should treat them as separate contributions rather than a unified research agenda.