misereru-slide-jev
Japanese research deck, article and source ledger tracking TypeSafe's Jev and System One models over time.
This repository is mostly documentation or write-up rather than runnable code.
Overview
misereru-slide-jev is a continuously updated Japanese-language research deck on TypeSafe AI's Jev and System One models, built from the misereru Marp template and published to GitHub Pages as slides, a presentation script and a standalone article. A research.md ledger keeps sources, third-party verification, caveats and update history.
Its stated policy is to prefer official TypeSafe sources, separate vendor claims from independent measurements, date prices and rate limits, and not to confuse type-valid output with semantically correct judgments. It tracks model updates, price and latency, independent accuracy and calibration checks, practical uses in routing, browser automation, RAG and guardrails, and comparisons with LLM logprobs or rule-based approaches. It also ships agent skills and an AGENTS.md so an AI agent can maintain the deck. Useful for Japanese readers wanting a sourced overview.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Studies whether Jev's typed decisions are accurate, calibrated and better than logprob or rule alternatives(inferred) | — |
How it works
- Record sources and findings in research.md.
- Update slides.md, then the script and article as needed.
- GitHub Actions builds with Marp and deploys to Pages.
Strengths
- Separates vendor claims from third-party measurements.
- Dated tracking of price, latency and access conditions.
- Published as slides, script and article.
Caveats
- Japanese only.
- A research document, not software that makes decisions.
- Accuracy depends on how current the ledger is (base date 2026-09-21).
Getting started
Read the published slides at https://myokoym.github.io/misereru-slide-jev/.At a glance
- Type
- Project
- Status
- discussion
- Language
- Markdown
- Runtime
- Marp, GitHub Pages
- Has code
- no
- Has tests
- no
Use cases and integrations
How curated lists describe it
Ongoing Japanese research deck on Jev and System One models, maintained as Markdown slides.
hellogumbo/awesome-jev
Frequently asked questions
What is this?
A maintained Japanese slide deck and article summarising research on Jev and System One.
Does it run Jev?
No, it compiles and cites sources and measurements.
Can an AI agent update it?
Yes, AGENTS.md and bundled skills define how an agent should keep it current.
Related projects
Bilingual research report on what Jev is, how its claims compare with independent evidence, and the open-source alternatives.
ProjectresearchLong-form analysis of what Jev is, why it is fast, how much of the hype holds up, and how hard it is to reproduce.
ProjectdiscussionUnofficial Jev information hub site with access guide, pricing comparison, a BYOK playground and an auto-updated news timeline.
AppproductionPythonDatawhale's Chinese Jev course: notebooks on the three primitives, translated recipes, evaluation and local Laya fine-tuning.
ArticleproductionPythonTypeSafe's launch post for System One models and Jev, covering the interface, RLCD training, pricing, demos and caveats.
ArticleproductionLatent Space podcast episode with TypeSafe's Diogo Almeida on RLCD, intelligence per dollar, reliability and post-chat interfaces.
Articlediscussion