# 2026-09-24 — Public-evidence snapshot

## English

This snapshot establishes the public-evidence layer for the Jev Research Index. Papers, projects, and online materials remain separate collections with independent counts.

The initial seed counts below describe the first curated export; the later “Web-crawled catalogue restoration” section records the superseding scope used by the current manifest and generated site.

### Added

- Defined the canonical scope: Jev refers to TypeSafe AI’s System One typed probabilistic decision model.
- Added same-name exclusions for Japanese encephalitis virus and the Journal of Extracellular Vesicles.
- Preserved 12 arXiv papers and 17 software projects with canonical URLs, relationship labels, evidence notes, and verification dates (including the newly verified crash-narrative study and Visual-Jev repository).
- Added 13 public-material records covering the official announcement, a podcast interview, an independent field report, third-party tracking, an official X channel, selected X records, and public video demonstrations.
- Added bilingual titles and summaries where the source was primary or clearly attributable.
- Added generated Markdown tables to `README.md`; `npm run validate` now fails if the README catalogue and JSON snapshots diverge.
- Added a separate README review-queue table so unresolved TRAE AI and Xiaohongshu leads remain visible without being promoted to the main catalogue.
- Discovery-only directory batches are retained in audit data but excluded from the public catalogue, counts, downloads, and rendered tables.

### Evidence limits

- This is a dated seed snapshot, not a claim of complete internet coverage.
- X records that could not be fetched directly remain at probable evidence and retain the access limitation.
- GitHub stars, forks, issue counts, prices, and model versions are not stored as timeless facts; any later metric must include an `as_of` date and a source URL.
- A Xiaohongshu search lead and one unstable Chinese-language article remain in `data/pending_review.json`; neither is presented as an established record.

### Limited source fetches

- `source:typesafe-x`: the account-level page was available, but individual linked statuses were not consistently machine-readable.
- `source:x-matija-jev`: the original status was not directly machine-readable; the linked archive is retained as a secondary evidence source.
- These are record-level access limitations, not claims that the channels or posts do not exist; `state/last_run.json` therefore keeps `failedSources` empty.

### Maintenance

- `state/last_run.json` records the successful review window and sources checked.
- Future reviews should read the previous snapshot, preserve changes and exclusions, and run `npm run validate` before publication.

## 中文

本快照建立 Jev Research Index 的公开材料层。论文、项目与互联网材料继续分层保存，并分别统计。

下面的首批数量描述最初的精选导出；文末“恢复网络爬取记录”部分记录了当前 manifest 与生成站点采用的后续口径。

### 新增内容

- 明确 canonical scope：Jev 专指 TypeSafe AI 的 System One 类型化概率决策模型。
- 增加 Japanese encephalitis virus 与 Journal of Extracellular Vesicles 等同名排除规则。
- 保存 12 篇 arXiv 论文与 17 个软件项目，每条均有 canonical URL、关系标签、证据短注和核验日期（新增事故叙述研究与 Visual-Jev 仓库）。
- 增加 13 条公开材料，覆盖官方发布、播客访谈、独立实践报告、第三方追踪、官方 X 频道、部分 X 记录及公开视频演示。
- 在来源为一手或可明确归属时，补充中英文标题和摘要。
- 在 `README.md` 增加由 JSON 自动生成的论文、项目与公开材料表格；`npm run validate` 会检查 Markdown 目录与 JSON 快照是否同步。
- 在 README 单独列出待审队列，使 TRAE AI 与小红书未决线索保持可见，但不进入主目录。
- 目录型批量发现记录保留在审计数据中，但不进入公开目录、统计、下载或渲染表格。

### 证据限制

- 这是带日期的首批快照，不宣称覆盖全部互联网公开资料。
- 无法直接抓取的 X 原帖保留为 probable，并记录访问限制。
- GitHub stars、forks、issues、价格和模型版本不会作为无日期的固定事实保存；后续如加入动态指标，必须带 `as_of` 日期和来源 URL。
- 小红书搜索线索及一条不稳定的中文文章仍保存在 `data/pending_review.json`，未作为已确认记录展示。

### 来源访问限制

- `source:typesafe-x`：账号级页面可访问，但所链接的个别状态页在本轮未能稳定机器读取。
- `source:x-matija-jev`：原始状态页无法直接机器读取，目录保留了一个二次存档作为辅助证据。
- 这些是记录级访问限制，并不表示频道或帖子不存在，因此 `state/last_run.json` 的 `failedSources` 仍为空。

### 维护方式

- `state/last_run.json` 保存成功核验窗口及已检查来源。
- 后续更新应读取上一份快照、保留变更与排除项，并在发布前运行 `npm run validate`。

## 2026-09-24 — Project publication dates

### English

- Added a `published` field to every project record.
- For the 337 catalogue GitHub projects, the date is the repository creation date parsed from the public GitHub repository page and retained with `publishedType: github_repository_created` and `publishedSource`.
- Web-crawled discovery records are included in this count but retain their `probable` evidence level where the linked primary page was not independently checked.
- The project table and website metadata now display the date together with this qualification.

### 中文

- 为每个项目记录增加 `published` 字段。
- 目录中的 337 个 GitHub 项目日期来自公开 GitHub 仓库页面中的仓库创建日期，并保留 `publishedType: github_repository_created` 与 `publishedSource`。
- 网络爬取发现记录也计入该数量；如果对应一手页面尚未独立检查，则继续保留 `probable` 证据等级。
- 项目表格和网站条目元数据现在会显示日期及其含义说明。

## 2026-09-24 — Web-crawled catalogue restoration

### English

- Restored the web-crawled project and online-material records to the rendered catalogue so the index exposes the full discovery corpus: 12 papers, 337 projects, 108 public materials, and 31 sources.
- These records retain `probable` evidence where the item was found through a community index or crawl and the linked primary page was not independently checked. The expanded count is therefore a discovery-corpus size, not a count of independently verified implementations. In the current data, 320 projects and 95 online materials are linked to community-directory discovery sources.
- Repository creation dates are displayed as publication proxies for GitHub projects; they must not be interpreted as Jev integration dates, especially for repositories created before the 2026-09-15 launch.

### 中文

- 恢复网络爬取发现的项目与互联网材料到渲染目录，公开展示完整发现池：12 篇论文、337 个项目、108 条公开材料和 31 个来源。
- 如果记录只来自社区目录或网络爬取、尚未独立检查对应一手页面，则继续保留 `probable` 证据等级。扩展后的数量是发现池规模，不是独立核验实现的数量；当前数据中有 320 个项目和 95 条互联网材料链接到社区目录发现源。
- GitHub 项目显示的是仓库创建日这一公开日期代理，不能解释为 Jev 接入日；尤其不能把 2026-09-15 首发前创建的仓库写成首发前已使用 Jev。