Jev study of 2,170 GitHub repos finds attention-adoption gap

A new empirical map of Jev's public ecosystem, covering 2,170 GitHub projects as of September 22, 2026, finds that where the model actually ships and where the stars pile up are two different distributions.

The paper, Jev in the Wild, by Guoming Ling and Muen Xue of Sun Yat-sen University with Zijian Ye of the Chinese University of Hong Kong, describes Jev as "a fast, low-cost decision model that answers natural-language questions with choices, binary judgments, and scores." The abstract's headline result: "Public attention is concentrated in routing and interface agents and does not track project counts."

The gap is concrete. Routing and automation projects account for 41.4% of accumulated stars while representing 19.6% of the sampled repositories. Content-and-expert tasks and search-and-memory, the two largest categories, hold 17.8% and 17.6% of projects respectively but average only 31 and 37 stars per project. Attribute judgment appears in 77% of projects, scoring in 52%, action selection in 31%.

The authors position the work as informing "the design and evaluation of general-purpose decision models across diverse application contexts." The paper does not publish per-project accuracy or commercial revenue data, so what those stars reflect, mindshare or production usage, is left open.