The design of 6G is moving faster than it can be built, according to a new arXiv paper. AI-driven search already proposes physical-layer algorithms at machine speed, but the abstract notes that taking a single one onto embedded baseband silicon remains a specialist task. That silicon imposes constraints that are not yet automated, creating a bottleneck between algorithmic innovation and practical hardware.
The paper frames this as a challenge for agentic AI: rather than stopping at algorithm discovery, future systems should extend to verified baseband silicon. The authors argue for a more integrated approach, where AI handles not just exploration but also the implementation and verification steps that currently require human expertise.
Because only the abstract is available, the specific methods are not detailed here. But the central observation is clear: the 6G physical layer will need AI that can carry ideas all the way to silicon, not just generate them.