Launches on September 24, 2026

Utilizes Jev (System One model) for chess gameplay.

Decisões tipadas em milissegundos com o Jev (System One da TypeSafe) via OpenRouter — state + perguntas noul/choice/score entram, decisões calibradas saem. Validação pesada baseada nos conceitos do modelo, bandas auto/hitl/abstain, deteção de prompt injection e servidor MCP. Zero dependências (Node ≥20).

Open-source single-file tools for AI agent builders, powered by Jev (TypeSafe AI's decision model): a loop detector that catches runaway agents and a task→tool router. No build. Use free: jevtypesafeai.com

Small air traffic control sim for a single runway (27). Give radio calls in plain English, or hand the frequency to Jev, TypeSafe's decision model, and watch it sequence, space and land the traffic.

Open-source single-file browser games where you play against Jev — TypeSafe AI's decision model drives the enemies/teammate, one real typed decision per move. No build. Play free: jevtypesafeai.com

Run a Pokémon Champions battle arena locally with Jev, TypeSafe, and Pokémon Showdown.

Automatic model + effort routing for coding agents, decided by TypeSafe's Jev

Benchmark TypeSafe's Jev (System One model) vs Claude on structured decisions: accuracy, latency, calibration, cost

面向语义决策模型的中文言下之意评测集：100 道伴侣对话 Choice 题，支持本地 ONNX、Jev 官方 API 与内网模型对比

Agente de browser para o Claude Code: jev-ultrafast + Jev (TypeSafe) com guardas de domínio, paragem antes de ações irreversíveis, verificação do resultado e custo por tarefa.

Claude Code skill: find searches where Google shows the wrong one of your pages, from Search Console, judged by TypeSafe's Jev model. Under 1 cent per run.

Claude Code skill: see how ChatGPT treats your brand across real buyer questions, read by TypeSafe's Jev model. About $0.20 per run.

Simple demo of TypeSafe AI - JEV

Independent, unofficial CLI for TypeSafe AI's System One API

Evaluate cloud spend against a budget and gate CI with Jev (approve, warn, block, or review).

Experimental TypeSafe Jev model routing for Codex Desktop: per-task model and reasoning selection, complex-task planning, subagent dispatch, and fallback.

Fight beside a Jev-controlled AI teammate that decides its own move (attack/defend/follow/retreat) each round. Single-file, no build. Play free: jevtypesafeai.com/games/jev-companion-arena

Jev-guided transcript compaction for Claude Code. Rust rewrite of tamaratran/fast-jev-compaction.

Jev-assisted Computer Use for Japanese UI, TypeSafe and Vercel AI Gateway, with portable Agent Skills

Voice-controlled Mac computer use on TypeSafe Jev: speech to typed decisions to verified accessibility actions, no screenshots to a big model

JEV action selection at 10 Hz meets original Chromium Dino physics. 13,724 points, 1080p60 recording, and auditable evidence.

Experimental macOS menu-bar assistant using TypeSafe Jev to recommend Codex models from drafts.

Claude Code plugin: per-prompt reasoning-effort routing with TypeSafe Jev

How far can a small model go? Testing TypeSafe's Jev on 1,352 cases against large LLMs: 98% accuracy, ~200x cheaper than Claude Opus 5.5

Jev (TypeSafe AI) plays FOOTSIES and Street Fighter III: 3rd Strike in real time from live game state

Classify failing tests as regression, flaky, environment, or unknown. Jev decides; results are never masked or auto-rerun.

Calling TypeSafe's Jev from Java with a hand-typed API contract. Small app to demo triage, zero-shot classification of commands from MASSIVE's 60 intents. And live batch classification with confidence-based routing.

Play poker, liar's dice, prisoner's dilemma and more against Jev, TypeSafe's System One model. It reads the situation and returns a probability for each legal move: a live mixed strategy. Bilingual EN/中文, runs in the browser. Educational, no real money.

Google Apps Script that classifies Gmail messages with Jev, TypeSafe's decision-only AI model. Runs on a time-driven trigger, sends each email's sender, subject, and trimmed body to Jev (via TypeSafe, OpenRouter, or Vercel AI Gateway), and applies Gmail labels based on typed Choice, Score, and Noul answers with confidence thresholds.

Name, classify and search graphify communities with TypeSafe Jev

Jev (TypeSafe) as the decision layer for Claude Code: permission gate, skill routing, conditional instructions, effort-tiered subagents, context packs. Tested PoC with evals.

Fast, cheap, typed decisions for AI agents — stdlib-only Python client for TypeSafe's Jev decision model via OpenRouter, with a design method, reusable question sets, examples, and tests.

Bridge that lets Claude Code verify an iOS app on a simulator, with TypeSafe's Jev choosing each step (planning stage)

Reproducible Jev benchmark on Klondike Draw-1 with baselines, decision traces, latency metrics, and replay viewer

Test bench for TypeSafe AI's Jev evaluation model on Vercel AI Gateway

TypeSafe AI の Jev を Ruby で試した実践例（感情分析・CloudTrailのリスク評価・フォームのリアルタイム検証）

Typed-decision models (Jev API, Laya 421M) vs Qwen3.5-35B on 12,000 real U.S. federal IT solicitations, graded against actual reseller quotes: accuracy, calibration, Wilson-bounded auto-accept cutoffs, cost.

⚡ JEV vs LAYA · Next-Gen AI Tetris Duel Benchmark Platform (TypeSafe Jev vs Local GPU ModernBERT) with React-Bits UI/UX

Servidor MCP que expõe o Jev (TypeSafe System One) como ferramenta de julgamento tipado para agentes: Noul, Choice e Score com validação por regra.

Route Issues and PRs to the best AI model using typed TypeSafe Jev decisions in GitHub Actions.

Codex plugin that uses TypeSafe Jev to choose subagent models, with scoped pstack integration.

Pick the right Claude model (haiku/sonnet/opus) per task using TypeSafe's Jev, with a Claude Code subagent hook

Claude Code per-turn model routing with TypeSafe Jev through an OpenRouter API key (fork of gargpratyush/jev-router)

Interactive Jev demo: Noul vs yes/no Choice disagree (TypeSafe jev-1.13 jaggedness)

You drive Pac-Man; Jev — TypeSafe AI's decision model — drives the ghosts hunting you, one real typed decision per turn. Single-file, no build. Play free: jevtypesafeai.com/games/jev-pacman

Extracting repeating patterns from noisy sequences with Jev (TypeSafe System One) — code proposes, model disposes

pi extension: TypeSafe Jev picks the model for each turn, only from models pi can use

pi extension: TypeSafe Jev decisions as tools (search, rerank, supervise, pick) and hooks (skill select, compaction)

Evaluate pull request risk with Jev and expose review depth plus recommended checks to CI.

C++ trading system leveraging TypeSafe Jev.