Lancements du 21 septembre 2026

L'assistant Android lit les conversations visibles via l'accessibilité et l'OCR local, puis affiche analyses et suggestions en surimpression. Jev évalue l'intention et les choix de réponse ; un modèle génératif rédige trois propositions que Jev classe.

AnyJev lit les probabilités de modèles ouverts pour répondre à des choix, questions binaires et scores sans rédiger. Cette approche indépendante s’inspire de Jev ; les têtes L2 ne modifient pas les poids du modèle de base. Utilisez L0 sans étiquettes ou ajustez une tête L2 par question avec transformers ou vLLM.

Rizzo Flow exécute des modèles ouverts localement et renvoie choix, probabilités, scores et nombres via une interface HTTP compatible Jev. Projet indépendant, il reproduit le principe d’interface de Jev, pas son architecture ni son entraînement propriétaires. Servez un modèle avec llama.cpp et connectez le code compatible TypeSafe au point d’accès local.

JevChat-Windows lit les conversations WeChat et KakaoTalk par OCR local et propose des réponses à côté du chat. Jev évalue intention et tension, puis classe trois propositions rédigées par un modèle de langage distinct. Configurez deux clés, pour décision et rédaction, sous Windows 10 1903+ ou Windows 11.

Le panneau flottant affiche intention, risque et réponses proposées, à copier ou insérer dans le champ de discussion. Utilisez un Mac Apple Silicon avec macOS 13 ou ultérieur. Autorisez l’enregistrement de l’écran pour lire les captures, et l’accessibilité pour lire du texte ou insérer une réponse ; choisissez les fournisseurs dans les réglages.

JevHarness fait écrire caractéristiques, questions et logique propres à une tâche par un modèle, avec amélioration facultative via récompenses et traces complètes. Explorez l’archive Pokémon avec les scripts web ou installez la compétence et décrivez une tâche pour créer un harnais.

Une tête de scoring renvoie des distributions pour contrôles, routage et notation, sans décoder de prose ni de réponse JSON. Le README compare AgentJev entraîné sur le benchmark à Jev 1.13.0 en zero-shot ; les conditions d'entraînement diffèrent.

Left-pad strings with TypeSafe AI's Jev. For reasons.

An open-source JEV-powered workbench for streaming data selection, quality evaluation, automatic LoRA training and held-out model evaluation.

An unofficial local visualizer for Jev (TypeSafe): a live view of every call your code makes. Not affiliated with TypeSafe AI.

A lightweight, native Swift 6 bridge integrating TypeSafe AI's Jev System One decision model into Apple's Foundation Models framework.

an open-source, from-first-principles reconstruction of the ideas behind TypeSafe AI's Jev, written in pytorch

Typed-judgment admission control at the traffic edge: three-layer prompt-injection and abuse filter for nginx/OpenResty, powered by TypeSafe Jev. Fail-open, cached, hot-reloadable.

5m BTC Up/Down Polymarket trading agent using Typesafe Jev as the decision layer & TUI

A Clash Royale bot with no trained policy: Jev (TypeSafe System One) makes every decision from the live game state

亚微秒级 System-1 模型，准确率服从高斯分布

Unofficial PoC: a BERT-style encoder decision engine behind a typed-decision (noul / choice / score) HTTP API. Not affiliated with TypeSafe.

Calibrate Jev questions against your own labels: tune criteria on labelled examples, confirm on a held-out set, get a verdict per question. Unofficial.

Laya (open source, local) vs TypeSafe Jev (API): two AI models race in Snake and fight in a Mortal-Kombat-style arena. Every move is a real model decision.

Jev shadcn-block generator

Score every test against a git diff with Jev, and emit the filter arguments vitest, node:test, Playwright, cargo test and go test already understand

微信里的话不知道怎么接？macOS 悬浮聊天助手：识别消息意图与沟通风险，GPT 生成多种话术，Jev 评估候选，一键填入微信。话我帮你想，发送你来定。

Run Jev-style typed decisions locally on your Mac with low RAM usage and fast responses

Software, by reflex. Say it, and the right small program runs: chosen by a calibrated classifier, run only when it is sure enough, and it asks before anything that cannot be undone. A CLI, a package manager and a TypeScript SDK: the first implementation of the idea.

A Minesweeper benchmark for LLM agents. One identical board, up to nine models in parallel, one clock, one tool layer.

Semantic "Did you mean?" for any CLI — wraps commands and uses TypeSafe's Jev to match subcommand typos by intent, not edit distance.

Craft Type-Safe Uncertainty-aware semantic pattern matching, control flow, and smart procedures for Effect DecisionModel and Jev

System One judgments (noul/choice/score) from any LLM in one prefill — a Rust, Jev-compatible /v1/systemone engine

Local, open-source Jev alternative: browser agent decisions with Laya (System One model) on your own machine. No cloud, no API key. Playwright/CDP, MCP-friendly.

A Swift package for typed decisions from language models (probabilities, choices, and scores), with support for local MLX models and the TypeSafe Jev API.

ego lite skill — each DOM step decided in ~0.4s, no LLM turn

ProphetLab's Texas Hold'em benchmark and playground for decision models: cash and SNG leaderboards, live replays, and bring-your-own-agent tables.

Hermes Agent plans, Jev (TypeSafe) picks bounded actions, Mineflayer executes: Minecraft with no screenshots or keypresses from a model. Includes the reproduction of rmalde/minecraft-agent's Ender Dragon run.

A local, offline System One server compatible with TypeSafe's Jev API. It answers typed yes/no, category and score questions with small open models.

jev_elixir is a small, idiomatic interface for making typed decisions with Jev. It exposes application-level functions such as Jev.ask?/2 and Jev.analyze/3, while provider adapters handle the external API wire format. The first release supports TypeSafe's System One API and OpenRouter's Decisions API. Both providers support all three System One qu

Route form submissions to the right people with Jev and AI SDK.

📏 Claude Code plugin that makes Claude follow your AGENTS.md: every reply and edit checked by TypeSafe Jev ✅

Zero-dependency System One decision harness: 5 semantic gates saving frontier AI agent tokens on trivial errors & doom loops. Python + TypeScript + Rust. MCP-compatible.

Site search that understands the question. Ranked by TypeSafe's Jev model.

Community-maintained JEV runner integration for Midscene Test

使用 TypeSafe Jev 在 X/Twitter 时间线上识别低质量、垃圾和广告帖子，并在文字区域显示醒目的半透明水印。

Trying to combine Jev with ontology

把对双休的支持，带进每一次购物。逛淘宝、京东时，JEV 实时猜测商品背后的工作制，疑似非双休直接盖上 PASS。原页生效，边逛边选。

Unofficial Go client for the TypeSafe AI System One API (Jev) — typed questions in, calibrated answers out.

Use local and hosted classifiers aka decision models aka Jev-like models, all through a single TypeSafe API

Confidence-gated incident routing with TypeSafe Jev

Claude Code plugin: asks TypeSafe Jev which installed skill fits each prompt and logs the answer (shadow-first). A working reference for the skill-suggestion cookbook on Claude Code — the README records why it is unlikely to help a strong model as a router.

Agent skill and CLI that turn natural language, an LLM prompt, or the code that runs one into a TypeSafe Jev decision: typed state, Choice/Score/Noul questions, and a runnable script

Type-safe Java 21 client for the TypeSafe AI Jev (System One) decision API

Give your AI JEV wings. A skill and plugin to build with TypeSafe Jev, upgrade LLM-only workflows and measure the result.