# 2026-09-27 — Daily web review

## English

This review checked public arXiv records, GitHub repositories, project documentation, and dated public web material visible after the 2026-09-25 snapshot.

### Added

- Added five arXiv preprints: **Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding** (2609.27678), **Calibrated Decision Models for Autonomous Penetration-Testing Harnesses** (2609.28940), **JEV vs. LLMs as Rubric Judges** (2609.29769), **Jev-Mobile** (2609.30186), and **Jev in the Wild** (2609.30216).
- Added three high-signal open-source projects: Nokia Applied Research's **AnyJev**, **jev4j**, and **jev-rs**; and independently re-verified the existing **Jev Chat Jarvis** record from its primary repository. AnyJev and jev-rs are independent implementations or compatible runtimes; they are not the hosted TypeSafe model or its weights.
- Added two public materials: an archived X article by Sydney Runkle on Jev and LangGraph, and an independently operated production-integration guide from The Jev AI.

### Evidence limits

- arXiv records are primary bibliographic sources for titles, authors, submission dates, abstracts, and paper-level claims; all reported measurements remain bounded by the corresponding paper's protocol.
- Repository metadata and READMEs support project scope and publication dates. No live installation, API call, or benchmark was run for the new project records in this review.
- The ecosystem-analysis paper reports a larger GitHub corpus than this catalogue. Its project count is a study result and is not substituted for this repository's curated, evidence-labelled records.

### Current snapshot

The catalogue now contains **18 papers, 361 projects, 125 public materials, and 58 sources**.

## 中文

本轮核查了 2026-09-25 快照之后可见的 arXiv 记录、GitHub 仓库、项目文档和带日期的公开网络材料。

### 新增内容

- 新增 5 篇 arXiv 预印本：**Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding**（2609.27678）、**Calibrated Decision Models for Autonomous Penetration-Testing Harnesses**（2609.28940）、**JEV vs. LLMs as Rubric Judges**（2609.29769）、**Jev-Mobile**（2609.30186）和 **Jev in the Wild**（2609.30216）。
- 新增 3 个高信号开源项目：Nokia Applied Research 的 **AnyJev**、**jev4j**、**jev-rs**；并根据主仓库重新核验已有的 **Jev Chat Jarvis** 记录。AnyJev 与 jev-rs 属于独立实现或兼容运行时，不是 TypeSafe 托管 Jev 模型或其权重。
- 新增 2 条公开材料：Sydney Runkle 关于 Jev 与 LangGraph 的 X 文章存档，以及 The Jev AI 独立运营站点发布的生产集成指南。

### 证据边界

- arXiv 记录作为标题、作者、提交日期、摘要和论文级主张的一手书目来源；所有测量结果仍限定在对应论文的实验协议内。
- 仓库元数据与 README 支持项目范围和公开日期。本轮未对新增项目执行实时安装、API 调用或 benchmark。
- 生态分析论文报告的 GitHub 项目数大于本目录；该数字属于论文研究结果，不替代本仓库按证据等级维护的精选记录。

### 当前快照

目录现包含 **18 篇论文、361 个项目、125 条公开材料和 58 条来源**。