According to God of Prompt on Twitter, current grant funding structures from organizations like NSF and DARPA mandate measurable progress on established safety metrics, driving researchers to prioritize benchmark scores over novel safety innovations (source: @godofprompt, Jan 14, 2026). This creates a cycle where new, potentially more effective AI safety metrics that are not easily quantifiable become unfundable, resulting in widespread optimization for existing benchmarks rather than substantive advancements. For AI industry stakeholders, this trend influences the allocation of resources and could limit true innovation in AI safety, emphasizing the need for funding models that reward qualitative as well as quantitative improvements.