Launches on September 20, 2026

A Claude Code skill of field notes on Jev: which question shape fits which job, how to test a use before trusting it, and the traps. Companion to TypeSafe's official skill.

Twelve field tests for TypeSafe's Jev model: calibration, guardrails, résumé screening, interview rubrics and its failure modes. Bring your own API key.

A pre-registered field trial of Jev (TypeSafe's judgment model) on a second brain and Claude Code history: 20 tests, bars written first, failures included, and the tools to repeat it.

SKYWARD: a 3D browser flight lab with a playable demo, bilingual docs, and live Jev control via Vercel AI Gateway.

登録済みプロフィールで Web フォームを自動入力する Chrome 拡張。入力欄とプロフィール項目の対応づけに typesafe-ai/jev の確率付き判定を使い、確信度つきで入力します（送信はしません）。

女友翻譯器 | 基於 TypeSafe AI 的 Jev 模型來幫你猜女友的真實想法

An unofficial Go SDK for TypeSafe AI's Jev model

Context-carrying graph retrieval with Jev: branching walks, reproducible ablations, and a visual replay.

A guarded MCP layer for Jev-powered browser decisions with isolated Playwright execution and explicit human approval.

Jev-backed guardrails for agent tool calls: library, native OpenCode/Hermes/DeepSeek Harness hooks, and a hosted eval-credit service.

Drive an Android phone from Claude Code: TypeSafe Jev picks which UI element to act on, the plugin validates and executes it.

Self-hosted project harness: code owns the loop, TypeSafe Jev owns the judgments, the LLM owns the content.

TypeScript demo using Jev as a decision layer for healthcare customer-support routing, urgency detection, and human escalation.

Benchmarking Jev (TypeSafe System One) on 5,000 real CDC survey respondents, with an interactive demo where every profile has a real model answer

Typed preflight for image generation. Jev decides whether a prompt may spend and which Gemini tier to use; code calls the image API only after that decision.

Local lab for exploring TypeSafe's Jev model on a frozen personal email corpus

Benchmarking TypeSafe's Jev against four rival prompt-injection detectors on 11,900 labelled prompts

flopcheck — paste a draft tweet, get 16 calibrated judgments from TypeSafe's Jev in one request

Codex-first CLI for bounded Jev judgments without echoing input state

Calibrated judgments for social posts, powered by Jev: relevance, quality, spam, kinds of post, how human a draft reads, and what someone keeps.

Typed client and benchmark harness for Jev, TypeSafe AI's System One decision model, through the Vercel AI Gateway. Measure accuracy, calibration and cost on your own data before you trust a threshold.

A small web app that identifies the predominant language in a text sample via Jev from TypeSafe AI

TypeSafe AI の Jev に水平思考クイズの出題者を任せた Web アプリ。Lambda Web Adapter + CloudFront + WAF の Terraform 一式

Chrome extension that scores how well a LinkedIn job matches your CV — and estimates your odds of passing screening — using the Jev decision model. Bilingual (ID/EN).

An experiment with semantic code linting using Jev from TypeSafe AI

Open-source Windows studio for Steam games: local AI host, editable avatars/chat, manual or JEV-assisted play, YouTube/Twitch OAuth and OBS streaming to YouTube/Twitch/X. Five-language UI; developer preview.

An LLM built without a language model — using TypeSafe Jev's non-generative AI for contextual response selection. Zero hallucination. Zero prompt injection.

Jev (TypeSafe AI の System One モデル) を地図に使ってみる実験場。音声指示で地理院地図Vector のレイヤーを重要度分類する。

Jev agent workstation.

Community experimental MCP server and Pi adapter for bounded TypeSafe Jev judgments

A tick-based ecology of mice, cats, traps and food where each animal's next move is judged by TypeSafe's Jev rather than decided by a rule.

Does Jev give the same answer twice? Repeatability and confidence-stability experiments on TypeSafe's Jev model — including a 52-dilemma moral reasoning stress test — via OpenRouter's Decisions API.

Before every NFL snap, a decision-only AI model (TypeSafe Jev) calls run or pass and go/punt/kick on fourth down, graded live against the coach.

End-to-end consumer demo for the JEV OAS Sentinel GitHub Action

TypeSafe AI の評価モデル Jev とローカルで対戦するオセロ

Self-hosted Outlook organization with TypeSafe Jev, custom categories, scoped Microsoft 365 access, and guided setup

Jev (TypeSafe System One) plays Pac-Man in real time: structured state in, one legal direction out. No screenshots, no fine-tuning.

用 typesafe-ai/jev 做决策引擎的 BTC/ETH 纸面交易台（每 5 秒评估，含买卖点与决策历史）

jev官方案例（中/eng）快速测试

Local JEV experiments: route decisions, Minesweeper solvers, and drone simulation

A UI over TypeSafe Jev: state and typed questions in, calibrated answers out.

Experiments with Jev (TypeSafe classifier via OpenRouter): an incident-triage agent with Jev as tool and gate, and a smart model router compared against an LLM router

TypeSafe Jev plays Flappy Bird via flappy-bird-gymnasium, spectated live over WebSocket/FastAPI

AI agent that plays Super Mario Bros with TypeSafe Jev - a typed decision model reading NES RAM instead of pixels, verified by emulator rollback search. Clears 1-1 to 1-4 and 2-1 for ~$0.02.

Is Jev (TypeSafe System One model) a decent preflop poker player? Reproducible benchmark vs 4 reference styles

JEV Prompt Injection Guardian is a prompt injection quarantine and risk-scoring system for LLMs powered by JEV (jev-1.13.0), TypeSafe AI's innovative "System One" model, as its primary engine. By eliminating ambiguous, long-winded natural language explanations, it instantly computes rigorously calibrated probability risk-scores.

Local observability proxy and playground for TypeSafe Jev

chrome-devtools-mcp plus one tool: Jev walks the browser to the spot, your agent makes one devtools call there.

把一段内容交给 TypeSafe AI 的 Jev 模型，得到带概率分布的结构化判断

Local Jev evaluation workbench: datasets, typed questions, threshold simulation and run comparison