THE JEV GUIDE · CURATED EDITION
Launch news, explanations, demos, and community projects around TypeSafe Jev, organized by topic. Every entry links back to its X source.
Curated selection · Not exhaustive · Updated manually
01 / EXPLORE
0 means awaiting a verified post
This post documents an experiment using the Jev model to solve a 3D bin packing problem, including computation time, number of Jev calls, and cost, noting that ignoring support constraints leads to floating items.
AbdelStark introduces LeJudge, the first experiment putting Jev in the loop of a JEPA world-model planner, judged by a decision model instead of an LLM. It imagines 300 futures and probes turn each into words.
This post shares a comparison and benchmark of decision models using Jev as a reference, with a link to the Jev Decision Index space on Hugging Face. Link
The author reports that averaging 8 permutations of Qwen outputs reduces KL error by ~79%, and all 720 permutations by ~97%, while Jev shows only slight improvement.
Inspired by the JEV hype, the author proposes testing JEV with MIT's Moral Machine dilemma test instead of chess.
Showcasing scRNAseq JEV application at a Boston hackathon, enabling faster, cheaper, more structured decisions for bioinformaticians and potentially serving as an RL signal. Link
The user notes that for identical resumes, Jev shows a slight preference for Black-sounding names and against Asian-sounding names, a very small but seemingly statistically significant difference.
The post argues that a Jev-like System One model is faster and cheaper, enables real-time interaction modeling, and has a drastically limited output space compared to LLMs, potentially revolutionizing human simulation.
Harrison Kinsley shares a small toy comparison of GLM 5.3 Flash with Jev versus OpenJev, noting the 4B OpenJev easily wins.
Running TypeSafeAI in ABM models to replicate the Gode & Sunder / Smith double-auction experiments with Jev agents: ZI-C agents reach equilibrium faster but with higher price variance, while Jev agents compress price variance and bid-ask spreads at the cost of slower equilibrium convergence.
According to the post, Jev has solved a long-standing open problem in mathematics.
Robert says Jev is now fully involved in the creation and further research with hybrid001.