If the numbers come from TypeSafe, how much of this can I actually trust?

The stated policy is to prioritise TypeSafe's official documentation and official evaluations as primary information while explicitly separating vendor claims from third-party measurements, and to check whether a claimed advantage is reachable with LLM logprobs or existing specialised logic. That is the right framing, but it means unverified claims are still in there, labelled by provenance in research.md rather than removed.

Is this a Jev demo or something I can call from code?

It is a research write-up. The deliverables are slides.md, presentation-script.md keyed to slides by stable ids, article.md written to be readable on its own, and research.md as the evidence ledger. Nothing here is a library or an API; if you want to run Jev, this is a briefing document, not an integration.

What happens to the published deck when I edit it?

Pushing to slides.md triggers GitHub Actions, which runs Marp to produce dist/site/index.html and deploys to GitHub Pages, at myokoym.github.io/misereru-slide-jev. Changes to the presentation script and article are also build targets. The raw presentation-script.md is never published, only its rendered form.

Can I take this content and put it in my own deck or repository?

The deck is generated from slides.md, which you can read, fork and rebuild. The one thing to be aware of is that AGENTS.md governs how this repository is maintained and it is strict about scope — updating slides, updating the material and reflecting a research finding are all treated as updating the canonical files in place, and creating a separate PPTX, a separate Markdown set, a separate repository or a different host is not done without an explicit request.

How do I update the research deck correctly?

Follow research.md first, then slides.md, then presentation-script.md or article.md if needed, and finally check the build and Pages deploy. AGENTS.md is the authority on the exact sync conditions.

What is the research.md file for?

It is the ledger of sources, third-party verification, reservations, and update history. Low-importance new information stays there rather than being added directly to the slides.

Which sources does the research privilege?

TypeSafe's official site, docs, and evals come first, with vendor performance claims kept separate from third-party measurements and with price, model version, rate limit, and access conditions dated.

CommunityJev Practices & Patterns

Misereru Slide Jev

A Japanese-language repository that continuously researches TypeSafe's Jev and System One models and organises the findings as Markdown slides, a presentation script and a standalone article.

See highlights

Open github.com

Stars: 0
Language: JavaScript
Published by: Community
Author: myokoym
Last verified: Oct 2026

myokoym/misereru-slide-jev

User repository on GitHub

View on GitHub

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

TypeSafe AI の Jev / System One Models を継続的に調査し、Markdownスライドとして整理するrepositoryです。. 調査更新の基本順序は research.md → slides.md → 必要に応じて presentation-script.md / article.md → build / Pages deploy確認です。詳細な同期条件と禁止事項はAGENTS.mdを正とします。. misereru-slide-writing: slides.md の生成・再構成・推敲。Jev資料は調査・共有を主目的とするため、原則として Reference mode を適用する. このrepositoryでは presentation-script.md を作成済みで、生成される目次を含む全slideを対象にしたcomplete scriptとして管理します。通常buildでは構造整合性を検査し、全slide分のcoverageを明示的に確認する場合は npm run build:script-complete を使います。.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: JavaScript

Highlights

AGENTS.md is the canonical rulebook, and a request to update the slides means updating the existing canonical files, not creating a new set.
Update order is research.md, then slides.md, then the script or article as needed, then build and verify the Pages deploy.
The research ledger keeps sources, third-party verification and reservations, and holds low-importance findings back from the slides.
The article is written to stand alone with its own structure, and is checked to have a single H1 and no raw HTML.

Watch out

This is a documentation and research repository, not code you call — there is no library, no SDK and no API surface to integrate with. Its stated policy is to treat TypeSafe's own documentation and evaluations as primary information while separating vendor performance claims from third-party measurements, which is the right posture but means the slides are downstream of upstream marketing in places the ledger has not yet independently checked. Price, model version, rate limit and access conditions are handled with explicit dates, so any figure you lift from a slide has a shelf life and should be read with the ledger's reference date. The tracked topics list includes items that are open questions rather than findings — jaggedness, SLA, privacy and data retention, regional differences, comparisons against logprobs-based and rule-based approaches — so treat those slides as a research agenda, not conclusions. Publication is automated: a push to slides.md runs GitHub Actions, Marp, then GitHub Pages. The raw presentation-script.md is deliberately not published, only the rendered version, and the deck is built from the misereru template. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-column-race

Live race rig that labels 1,000 real app reviews with Jev typed questions against Gemini JSON output on identical batches, then replays recordings and re-ranks rows locally without extra model calls.