Measuring what AI does to human wellbeing drew both funding and formal measurement frameworks in the same week.
Anthropic opened research grants for evaluating AI's impact on wellbeing, while arXiv preprints such as STONIC proposed layered contracts for profiling model values. IEEE Spectrum covered AI companion robots in the home. Five items in all: one official post, one trade report, three academic papers.
The evaluation axis is widening from capability to human effect, with funders and method-builders moving separately rather than together. The cluster sits under a developer-tools label, though, so whether it is one coherent thread is not yet established.
Which grant proposals are selected, and whether measurement contracts like STONIC are adopted as benchmarks elsewhere. Wellbeing metrics reaching developer toolchains would be the turn.