Keep the evidence close.

Jev Fieldnotes is an independent developer reading guide, not a peer-reviewed paper or an official TypeSafe publication. We organize practical examples and link to the people who published them.

How to read the labels

• Official cookbook / documented pattern: a workflow described by TypeSafe. It still needs evaluation in your application.
• Community code: a project listed by its author or a source collection. Inclusion does not mean we executed or audited that project.
• Workflow idea: a useful direction to explore, not a shipped product claim.
• Research: an analysis or an author’s research report. Architecture hypotheses remain hypotheses.

The source review

All nine supplied sources were opened and visually inspected in Chrome on 25 September 2026. The 44-page Jev presentation was downloaded for local visual review. All 44 pages are now available in the on-page slide viewer, with the original PDF download and Drive source link. The slide content is preserved; corrections are noted separately.

One correction matters: the slide deck says independent benchmarks are absent. Benchmark Heaven now publishes JevBench with methods and versioned results. The reading guide therefore points to current evidence instead of repeating that slide as fact.

Laya’s report is useful for exploring an alternative, but its author-reported local latency and fine-tuned results are not a universal comparison with a hosted, zero-shot API. A typed answer can also be wrong; a fixed output format does not guarantee factual accuracy.

GitHub project directory and visuals

The project directory merges a 148-project community catalogue dated 19 September 2026 with public GitHub API results. New searches match Jev in project names and descriptions, with 100 results per page and GitHub’s 1,000-result search ceiling. Local filters search the loaded directory. Search results are discovery leads, not individually audited projects.

Each project labels whether its metrics are live or from the dated catalogue. Project details load the repository, original README and developer profile on demand. Public requests work without a token; rate limits may delay updates. Previously loaded data and the catalogue remain available.

Real demo preview frames come from the supplied Awesome Jev repository, with original creators credited beside the images. Other covers are original, clearly labeled workflow illustrations. They do not depict the actual project interface. Full visual credits and the source license are available.

Fresh feed and dated archive

The feed requests public GitHub repository search, Hacker News search via Algolia, and Bluesky public search. An optional X adapter uses the official recent-search endpoint when X_BEARER_TOKEN is set. Source failures are surfaced in the interface. We do not scrape your logged-in social accounts.

The server caches a refresh for five minutes and shares in-flight requests. The page polls while visible. This is periodically refreshed discovery, not a streaming firehose or complete internet coverage. Cached entries from a failed source are marked stale and expire after 24 hours.

Activity ordering uses log(1 + engagement) / (1 + ageHours / 24). GitHub uses lifetime stars and last push; social sources use likes or points and post time. These metrics are not directly equivalent, so use the per-source filters and Latest sort for more focused comparisons.

If a TypeSafe server key is configured, a bounded batch of up to 20 public feed entries gets Jev relevance probabilities. Those scores are advisory and do not hide entries automatically. Without a key, the site uses source queries and keyword matching.

The 74-demo archive is derived from the CC0 data in Awesome Jev Use Cases by Boulanouar Walid. Its likes are a 19 September 2026 snapshot, not current counts. Original posts and authors are linked. No creator videos or preview images are republished.

References

Build your own entry

This notebook stores source descriptions and use cases as typed data. Add a title, a concrete decision, steps, a limitation and a source link. Prefer a reproducible example over a performance claim without its measurement conditions.